update to v5.0.0 (#1308)

* build(deps-dev): bump black from 22.12.0 to 23.1.0

Bumps [black](https://github.com/psf/black) from 22.12.0 to 23.1.0.
- [Release notes](https://github.com/psf/black/releases)
- [Changelog](https://github.com/psf/black/blob/main/CHANGES.md)
- [Commits](https://github.com/psf/black/compare/22.12.0...23.1.0)

---
updated-dependencies:
- dependency-name: black
  dependency-type: direct:development
  update-type: version-update:semver-major
...

Signed-off-by: dependabot[bot] <support@github.com>

* reformat black 23.1.0

---------

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

.devcontainer/Dockerfile

@@ -0,0 +1,21 @@
+# See here for image contents: https://github.com/microsoft/vscode-dev-containers/tree/v0.233.0/containers/python-3/.devcontainer/base.Dockerfile
+
+# [Choice] Python version (use -bullseye variants on local arm64/Apple Silicon): 3, 3.10, 3.9, 3.8, 3.7, 3.6, 3-bullseye, 3.10-bullseye, 3.9-bullseye, 3.8-bullseye, 3.7-bullseye, 3.6-bullseye, 3-buster, 3.10-buster, 3.9-buster, 3.8-buster, 3.7-buster, 3.6-buster
+ARG VARIANT="3.10-bullseye"
+FROM mcr.microsoft.com/vscode/devcontainers/python:0-${VARIANT}
+
+# [Choice] Node.js version: none, lts/*, 16, 14, 12, 10
+ARG NODE_VERSION="none"
+RUN if [ "${NODE_VERSION}" != "none" ]; then su vscode -c "umask 0002 && . /usr/local/share/nvm/nvm.sh && nvm install ${NODE_VERSION} 2>&1"; fi
+
+# [Optional] If your pip requirements rarely change, uncomment this section to add them to the image.
+# COPY requirements.txt /tmp/pip-tmp/
+# RUN pip3 --disable-pip-version-check --no-cache-dir install -r /tmp/pip-tmp/requirements.txt \
+#    && rm -rf /tmp/pip-tmp
+
+# [Optional] Uncomment this section to install additional OS packages.
+# RUN apt-get update && export DEBIAN_FRONTEND=noninteractive \
+#     && apt-get -y install --no-install-recommends <your-package-list-here>
+
+# [Optional] Uncomment this line to install global node packages.
+# RUN su vscode -c "source /usr/local/share/nvm/nvm.sh && npm install -g <your-package-here>" 2>&1

.devcontainer/devcontainer.json

@@ -0,0 +1,51 @@
+// For format details, see https://aka.ms/devcontainer.json. For config options, see the README at:
+// https://github.com/microsoft/vscode-dev-containers/tree/v0.233.0/containers/python-3
+{
+	"name": "Python 3",
+	"build": {
+		"dockerfile": "Dockerfile",
+		"context": "..",
+		"args": { 
+			// Update 'VARIANT' to pick a Python version: 3, 3.10, 3.9, 3.8, 3.7, 3.6
+			// Append -bullseye or -buster to pin to an OS version.
+			// Use -bullseye variants on local on arm64/Apple Silicon.
+			"VARIANT": "3.10",
+			// Options
+			"NODE_VERSION": "none"
+		}
+	},
+
+	// Set *default* container specific settings.json values on container create.
+	"settings": { 
+		"python.defaultInterpreterPath": "/usr/local/bin/python",
+		"python.linting.enabled": true,
+		"python.linting.pylintEnabled": true,
+		"python.formatting.autopep8Path": "/usr/local/py-utils/bin/autopep8",
+		"python.formatting.blackPath": "/usr/local/py-utils/bin/black",
+		"python.formatting.yapfPath": "/usr/local/py-utils/bin/yapf",
+		"python.linting.banditPath": "/usr/local/py-utils/bin/bandit",
+		"python.linting.flake8Path": "/usr/local/py-utils/bin/flake8",
+		"python.linting.mypyPath": "/usr/local/py-utils/bin/mypy",
+		"python.linting.pycodestylePath": "/usr/local/py-utils/bin/pycodestyle",
+		"python.linting.pydocstylePath": "/usr/local/py-utils/bin/pydocstyle",
+		"python.linting.pylintPath": "/usr/local/py-utils/bin/pylint"
+	},
+
+	// Add the IDs of extensions you want installed when the container is created.
+	"extensions": [
+		"ms-python.python",
+		"ms-python.vscode-pylance"
+	],
+
+	// Use 'forwardPorts' to make a list of ports inside the container available locally.
+	// "forwardPorts": [],
+
+	// Use 'postCreateCommand' to run commands after the container is created.
+	"postCreateCommand": "git submodule update --init && pip3 install --user -e .[dev]",
+
+	// Comment out to connect as root instead. More info: https://aka.ms/vscode-remote/containers/non-root.
+	"remoteUser": "vscode",
+	"features": {
+		"git": "latest"
+	}
+}

.github/CONTRIBUTING.md

@@ -2,7 +2,7 @@
 
 First off, thanks for taking the time to contribute!
 
-The following is a set of guidelines for contributing to capa and its packages, which are hosted in the [FireEye Organization](https://github.com/fireeye) on GitHub. These are mostly guidelines, not rules. Use your best judgment, and feel free to propose changes to this document in a pull request.
+The following is a set of guidelines for contributing to capa and its packages, which are hosted in the [Mandiant Organization](https://github.com/mandiant) on GitHub. These are mostly guidelines, not rules. Use your best judgment, and feel free to propose changes to this document in a pull request.
 
 #### Table Of Contents
 
@@ -31,10 +31,10 @@ This project and everyone participating in it is governed by the [Capa Code of C
 
 ### Capa and its repositories
 
-We host the capa project as three Github repositories:
-  - [capa](https://github.com/fireeye/capa)
-  - [capa-rules](https://github.com/fireeye/capa-rules)
-  - [capa-testfiles](https://github.com/fireeye/capa-testfiles)
+We host the capa project as three GitHub repositories:
+  - [capa](https://github.com/mandiant/capa)
+  - [capa-rules](https://github.com/mandiant/capa-rules)
+  - [capa-testfiles](https://github.com/mandiant/capa-testfiles)
   
 The command line tools, logic engine, and other Python source code are found in the `capa` repository.
 This is the repository to fork when you want to enhance the features, performance, or user interface of capa.
@@ -54,7 +54,7 @@ These are files you'll need in order to run the linter (in `--thorough` mode) an
 ### Design Decisions
 
 When we make a significant decision in how we maintain the project and what we can or cannot support,
- we will document it in the [capa issues tracker](https://github.com/fireeye/capa/issues).
+ we will document it in the [capa issues tracker](https://github.com/mandiant/capa/issues).
 This is the best place review our discussions about what/how/why we do things in the project.
 If you have a question, check to see if it is documented there.
 If it is *not* documented there, or you can't find an answer, please open a issue.
@@ -78,7 +78,7 @@ Fill out [the required template](./ISSUE_TEMPLATE/bug_report.md),
 #### Before Submitting A Bug Report
 
 * **Determine [which repository the problem should be reported in](#capa-and-its-repositories)**.
-* **Perform a [cursory search](https://github.com/fireeye/capa/issues?q=is%3Aissue)** to see if the problem has already been reported. If it has **and the issue is still open**, add a comment to the existing issue instead of opening a new one.
+* **Perform a [cursory search](https://github.com/mandiant/capa/issues?q=is%3Aissue)** to see if the problem has already been reported. If it has **and the issue is still open**, add a comment to the existing issue instead of opening a new one.
 
 #### How Do I Submit A (Good) Bug Report?
 
@@ -101,7 +101,7 @@ Explain the problem and include additional details to help maintainers reproduce
 Provide more context by answering these questions:
 
 * **Did the problem start happening recently** (e.g. after updating to a new version of capa) or was this always a problem?
-* If the problem started happening recently, **can you reproduce the problem in an older version of capa?** What's the most recent version in which the problem doesn't happen? You can download older versions of capa from [the releases page](https://github.com/fireeye/capa/releases).
+* If the problem started happening recently, **can you reproduce the problem in an older version of capa?** What's the most recent version in which the problem doesn't happen? You can download older versions of capa from [the releases page](https://github.com/mandiant/capa/releases).
 * **Can you reliably reproduce the issue?** If not, provide details about how often the problem happens and under which conditions it normally happens.
 * If the problem is related to working with files (e.g. opening and editing files), **does the problem happen for all files and projects or only some?** Does the problem happen only when working with local or remote files (e.g. on network drives), with files of a specific type (e.g. only JavaScript or Python files), with large files or files with very long lines, or with files in a specific encoding? Is there anything else special about the files you are using?
 
@@ -119,7 +119,7 @@ Before creating enhancement suggestions, please check [this list](#before-submit
 #### Before Submitting An Enhancement Suggestion
 
 * **Determine [which repository the enhancement should be suggested in](#capa-and-its-repositories).**
-* **Perform a [cursory search](https://github.com/fireeye/capa/issues?q=is%3Aissue)** to see if the enhancement has already been suggested. If it has, add a comment to the existing issue instead of opening a new one.
+* **Perform a [cursory search](https://github.com/mandiant/capa/issues?q=is%3Aissue)** to see if the enhancement has already been suggested. If it has, add a comment to the existing issue instead of opening a new one.
 
 #### How Do I Submit A (Good) Enhancement Suggestion?
 
@@ -138,15 +138,15 @@ Enhancement suggestions are tracked as [GitHub issues](https://guides.github.com
 
 Unsure where to begin contributing to capa? You can start by looking through these `good-first-issue` and `rule-idea` issues:
 
-* [good-first-issue](https://github.com/fireeye/capa/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22) - issues which should only require a few lines of code, and a test or two.
-* [rule-idea](https://github.com/fireeye/capa-rules/issues?q=is%3Aissue+is%3Aopen+label%3A%22rule+idea%22) - issues that describe potential new rule ideas.
+* [good-first-issue](https://github.com/mandiant/capa/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22) - issues which should only require a few lines of code, and a test or two.
+* [rule-idea](https://github.com/mandiant/capa-rules/issues?q=is%3Aissue+is%3Aopen+label%3A%22rule+idea%22) - issues that describe potential new rule ideas.
 
 Both issue lists are sorted by total number of comments. While not perfect, number of comments is a reasonable proxy for impact a given change will have.
 
 #### Local development
 
 capa and all its resources can be developed locally. 
-For instructions on how to do this, see the "Method 3" section of the [installation guide](https://github.com/fireeye/capa/blob/master/doc/installation.md).
+For instructions on how to do this, see the "Method 3" section of the [installation guide](https://github.com/mandiant/capa/blob/master/doc/installation.md).
 
 ### Pull Requests
 
@@ -190,8 +190,8 @@ Our CI pipeline will reformat and enforce the Python styleguide.
 
 All (non-nursery) capa rules must:
 
-  1. pass the [linter](https://github.com/fireeye/capa/blob/master/scripts/lint.py), and
-  2. be formatted with [capafmt](https://github.com/fireeye/capa/blob/master/scripts/capafmt.py)
+  1. pass the [linter](https://github.com/mandiant/capa/blob/master/scripts/lint.py), and
+  2. be formatted with [capafmt](https://github.com/mandiant/capa/blob/master/scripts/capafmt.py)
   
 This ensures that all rules meet the same minimum level of quality and are structured in a consistent way.
 Our CI pipeline will reformat and enforce the capa rules styleguide.

.github/ISSUE_TEMPLATE/bug_report.md

@@ -5,16 +5,16 @@ about: Create a report to help us improve
 ---
 <!--
 # Is your bug report related to capa rules (for example a false positive)?
-We use submodules to separate code, rules and test data. If your issue is related to capa rules, please report it at https://github.com/fireeye/capa-rules/issues.
+We use submodules to separate code, rules and test data. If your issue is related to capa rules, please report it at https://github.com/mandiant/capa-rules/issues.
 
 # Have you checked that your issue isn't already filed?
-Please search if there is a similar issue at https://github.com/fireeye/capa/issues. If there is already a similar issue, please add more details there instead of opening a new one.
+Please search if there is a similar issue at https://github.com/mandiant/capa/issues. If there is already a similar issue, please add more details there instead of opening a new one.
 
 # Have you read capa's Code of Conduct?
-By filing an Issue, you are expected to comply with it, including treating everyone with respect: https://github.com/fireeye/capa/blob/master/.github/CODE_OF_CONDUCT.md
+By filing an Issue, you are expected to comply with it, including treating everyone with respect: https://github.com/mandiant/capa/blob/master/.github/CODE_OF_CONDUCT.md
 
 # Have you read capa's CONTRIBUTING guide?
-It contains helpful information about how to contribute to capa. Check https://github.com/fireeye/capa/blob/master/.github/CONTRIBUTING.md#reporting-bugs
+It contains helpful information about how to contribute to capa. Check https://github.com/mandiant/capa/blob/master/.github/CONTRIBUTING.md#reporting-bugs
 -->
 
 ### Description

.github/ISSUE_TEMPLATE/feature_request.md

@@ -5,16 +5,16 @@ about: Suggest an idea for capa
 ---
 <!--
 # Is your issue related to capa rules (for example an idea for a new rule)?
-We use submodules to separate code, rules and test data. If your issue is related to capa rules, please report it at https://github.com/fireeye/capa-rules/issues.
+We use submodules to separate code, rules and test data. If your issue is related to capa rules, please report it at https://github.com/mandiant/capa-rules/issues.
 
 # Have you checked that your issue isn't already filed?
-Please search if there is a similar issue at https://github.com/fireeye/capa/issues. If there is already a similar issue, please add more details there instead of opening a new one.
+Please search if there is a similar issue at https://github.com/mandiant/capa/issues. If there is already a similar issue, please add more details there instead of opening a new one.
 
 # Have you read capa's Code of Conduct?
-By filing an Issue, you are expected to comply with it, including treating everyone with respect: https://github.com/fireeye/capa/blob/master/.github/CODE_OF_CONDUCT.md
+By filing an Issue, you are expected to comply with it, including treating everyone with respect: https://github.com/mandiant/capa/blob/master/.github/CODE_OF_CONDUCT.md
 
 # Have you read capa's CONTRIBUTING guide?
-It contains helpful information about how to contribute to capa. Check https://github.com/fireeye/capa/blob/master/.github/CONTRIBUTING.md#suggesting-enhancements
+It contains helpful information about how to contribute to capa. Check https://github.com/mandiant/capa/blob/master/.github/CONTRIBUTING.md#suggesting-enhancements
 -->
 
 ### Summary

.github/mypy/mypy.ini

@@ -21,9 +21,6 @@ ignore_missing_imports = True
 [mypy-flirt.*]
 ignore_missing_imports = True
 
-[mypy-smda.*]
-ignore_missing_imports = True
-
 [mypy-lief.*]
 ignore_missing_imports = True
 
@@ -48,6 +45,9 @@ ignore_missing_imports = True
 [mypy-ida_bytes.*]
 ignore_missing_imports = True
 
+[mypy-ida_nalt.*]
+ignore_missing_imports = True
+
 [mypy-ida_kernwin.*]
 ignore_missing_imports = True
 
@@ -60,6 +60,9 @@ ignore_missing_imports = True
 [mypy-ida_loader.*]
 ignore_missing_imports = True
 
+[mypy-ida_segment.*]
+ignore_missing_imports = True
+
 [mypy-PyQt5.*]
 ignore_missing_imports = True
 
@@ -74,3 +77,9 @@ ignore_missing_imports = True
 
 [mypy-elftools.*]
 ignore_missing_imports = True
+
+[mypy-dncil.*]
+ignore_missing_imports = True
+
+[mypy-netnode.*]
+ignore_missing_imports = True

.github/pull_request_template.md

@@ -3,7 +3,7 @@
 Thank you for contributing to capa! <3
 
 Please read capa's CONTRIBUTING guide if you haven't done so already.
-It contains helpful information about how to contribute to capa. Check https://github.com/fireeye/capa/blob/master/.github/CONTRIBUTING.md
+It contains helpful information about how to contribute to capa. Check https://github.com/mandiant/capa/blob/master/.github/CONTRIBUTING.md
 
 Please describe the changes in this pull request (PR). Include your motivation and context to help us review.
 

.github/pyinstaller/hooks/hook-smda.py

@@ -1,5 +0,0 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
-import PyInstaller.utils.hooks
-
-# ref: https://groups.google.com/g/pyinstaller/c/amWi0-66uZI/m/miPoKfWjBAAJ
-binaries = PyInstaller.utils.hooks.collect_dynamic_libs("capstone")

.github/pyinstaller/hooks/hook-vivisect.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 
 from PyInstaller.utils.hooks import copy_metadata
 

.github/pyinstaller/pyinstaller.spec

@@ -1,56 +1,40 @@
 # -*- mode: python -*-
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 import os.path
 import subprocess
 
 import wcwidth
 
 
-# when invoking pyinstaller from the project root,
-# this gets run from the project root.
-with open('./capa/version.py', 'wb') as f:
-    # git output will look like:
-    #
-    #     tags/v1.0.0-0-g3af38dc
-    #         ------- tag
-    #                 - commits since
-    #                   g------- git hash fragment
-    version = (subprocess.check_output(["git", "describe", "--always", "--tags", "--long"])
-               .decode("utf-8")
-               .strip()
-               .replace("tags/", ""))
-    f.write(("__version__ = '%s'" % version).encode("utf-8"))
-
 a = Analysis(
     # when invoking pyinstaller from the project root,
     # this gets invoked from the directory of the spec file,
     # i.e. ./.github/pyinstaller
-    ['../../capa/main.py'],
-    pathex=['capa'],
+    ["../../capa/main.py"],
+    pathex=["capa"],
     binaries=None,
     datas=[
         # when invoking pyinstaller from the project root,
         # this gets invoked from the directory of the spec file,
         # i.e. ./.github/pyinstaller
-        ('../../rules', 'rules'),
-        ('../../sigs', 'sigs'),
-
+        ("../../rules", "rules"),
+        ("../../sigs", "sigs"),
+        ("../../cache", "cache"),
         # capa.render.default uses tabulate that depends on wcwidth.
         # it seems wcwidth uses a json file `version.json`
         # and this doesn't get picked up by pyinstaller automatically.
         # so we manually embed the wcwidth resources here.
         #
         # ref: https://stackoverflow.com/a/62278462/87207
-        (os.path.dirname(wcwidth.__file__), 'wcwidth')
+        (os.path.dirname(wcwidth.__file__), "wcwidth"),
     ],
     # when invoking pyinstaller from the project root,
     # this gets run from the project root.
-    hookspath=['.github/pyinstaller/hooks'],
+    hookspath=[".github/pyinstaller/hooks"],
     runtime_hooks=None,
     excludes=[
         # ignore packages that would otherwise be bundled with the .exe.
         # review: build/pyinstaller/xref-pyinstaller.html
-
         # we don't do any GUI stuff, so ignore these modules
         "tkinter",
         "_tkinter",
@@ -60,7 +44,6 @@ a = Analysis(
         # since we don't spawn a notebook, we can safely remove these.
         "IPython",
         "ipywidgets",
-
         # these are pulled in by networkx
         # but we don't need to compute the strongly connected components.
         "numpy",
@@ -68,7 +51,6 @@ a = Analysis(
         "matplotlib",
         "pandas",
         "pytest",
-
         # deps from viv that we don't use.
         # this duplicates the entries in `hook-vivisect`,
         # but works better this way.
@@ -78,35 +60,35 @@ a = Analysis(
         "PyQt5",
         "qt5",
         "pyqtwebengine",
-        "pyasn1"
-    ])
+        "pyasn1",
+    ],
+)
 
-a.binaries = a.binaries - TOC([
- ('tcl85.dll', None, None),
- ('tk85.dll', None, None),
- ('_tkinter', None, None)])
+a.binaries = a.binaries - TOC([("tcl85.dll", None, None), ("tk85.dll", None, None), ("_tkinter", None, None)])
 
 pyz = PYZ(a.pure, a.zipped_data)
 
-exe = EXE(pyz,
-          a.scripts,
-          a.binaries,
-          a.zipfiles,
-          a.datas,
-          exclude_binaries=False,
-          name='capa',
-          icon='logo.ico',
-          debug=False,
-          strip=None,
-          upx=True,
-          console=True )
+exe = EXE(
+    pyz,
+    a.scripts,
+    a.binaries,
+    a.zipfiles,
+    a.datas,
+    exclude_binaries=False,
+    name="capa",
+    icon="logo.ico",
+    debug=False,
+    strip=None,
+    upx=True,
+    console=True,
+)
 
 # enable the following to debug the contents of the .exe
 #
-#coll = COLLECT(exe,
+# coll = COLLECT(exe,
 #               a.binaries,
 #               a.zipfiles,
 #               a.datas,
 #               strip=None,
 #               upx=True,
-#               name='capa-dat')
+#               name='capa-dat')

.github/workflows/build.yml

@@ -1,8 +1,8 @@
 name: build
 
 on:
-  push:
-    branches: [master]
+  pull_request:
+    branches: [ master ]
   release:
     types: [edited, published]
 
@@ -11,6 +11,8 @@ jobs:
     name: PyInstaller for ${{ matrix.os }}
     runs-on: ${{ matrix.os }}
     strategy:
+      # set to false for debugging
+      fail-fast: true
       matrix:
         include:
           - os: ubuntu-18.04
@@ -20,41 +22,42 @@ jobs:
           - os: windows-2019
             artifact_name: capa.exe
             asset_name: windows
-          - os: macos-10.15
+          - os: macos-11
+            # use older macOS for assumed better portability
             artifact_name: capa
             asset_name: macos
     steps:
       - name: Checkout capa
-        uses: actions/checkout@v2
+        uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
         with:
           submodules: true
       # using Python 3.8 to support running across multiple operating systems including Windows 7
       - name: Set up Python 3.8
-        uses: actions/setup-python@v2
+        uses: actions/setup-python@d27e3f3d7c64b4bbf8e4abfb9b63b83e846e0435 # v4.5.0
         with:
           python-version: 3.8
       - if: matrix.os == 'ubuntu-18.04'
         run: sudo apt-get install -y libyaml-dev
-      - name: Install PyInstaller
-        run: pip install 'pyinstaller==4.2'
-      - name: Install capa
-        run: pip install -e .
+      - name: Upgrade pip, setuptools
+        run: python -m pip install --upgrade pip setuptools
+      - name: Install capa with build requirements
+        run: pip install -e .[build]
+      - name: Cache the rule set
+        run: python ./scripts/cache-ruleset.py ./rules/ ./cache/
       - name: Build standalone executable
-        run: pyinstaller .github/pyinstaller/pyinstaller.spec
+        run: pyinstaller --log-level DEBUG .github/pyinstaller/pyinstaller.spec
       - name: Does it run (PE)?
         run: dist/capa "tests/data/Practical Malware Analysis Lab 01-01.dll_"
       - name: Does it run (Shellcode)?
         run: dist/capa "tests/data/499c2a85f6e8142c3f48d4251c9c7cd6.raw32"
       - name: Does it run (ELF)?
         run: dist/capa "tests/data/7351f8a40c5450557b24622417fc478d.elf_"
-      - uses: actions/upload-artifact@v2
+      - uses: actions/upload-artifact@0b7f8abb1508181956e8e162db84b466c27e18ce # v3.1.2
         with:
           name: ${{ matrix.asset_name }}
           path: dist/${{ matrix.artifact_name }}
 
   test_run:
-    # test that binaries run on push to master
-    if: github.event_name == 'push'
     name: Test run on ${{ matrix.os }}
     runs-on: ${{ matrix.os }}
     needs: [build]
@@ -68,16 +71,16 @@ jobs:
           - os: ubuntu-20.04
             artifact_name: capa
             asset_name: linux
-          - os: windows-2016
+          - os: windows-2022
             artifact_name: capa.exe
             asset_name: windows
     steps:
       - name: Download ${{ matrix.asset_name }}
-        uses: actions/download-artifact@v2
+        uses: actions/download-artifact@9bc31d5ccc31df68ecc42ccf4149144866c47d8a # v3.0.2
         with:
           name: ${{ matrix.asset_name }}
       - name: Set executable flag
-        if: matrix.os != 'windows-2016'
+        if: matrix.os != 'windows-2022'
         run: chmod +x ${{ matrix.artifact_name }}
       - name: Run capa
         run: ./${{ matrix.artifact_name }} -h
@@ -99,7 +102,7 @@ jobs:
             artifact_name: capa
     steps:
       - name: Download ${{ matrix.asset_name }}
-        uses: actions/download-artifact@v2
+        uses: actions/download-artifact@9bc31d5ccc31df68ecc42ccf4149144866c47d8a # v3.0.2
         with:
           name: ${{ matrix.asset_name }}
       - name: Set executable flag
@@ -109,7 +112,7 @@ jobs:
       - name: Zip ${{ matrix.artifact_name }} into ${{ env.zip_name }}
         run: zip ${{ env.zip_name }} ${{ matrix.artifact_name }}
       - name: Upload ${{ env.zip_name }} to GH Release
-        uses: svenstaro/upload-release-action@v2
+        uses: svenstaro/upload-release-action@2728235f7dc9ff598bd86ce3c274b74f802d2208 # v2
         with:
           repo_token: ${{ secrets.GITHUB_TOKEN}}
           file: ${{ env.zip_name }}

.github/workflows/changelog.yml

@@ -17,7 +17,7 @@ jobs:
     steps:
     - name: Get changed files
       id: files
-      uses: Ana06/get-changed-files@v1.2
+      uses: Ana06/get-changed-files@e0c398b7065a8d84700c471b6afc4116d1ba4e96 # v2.2.0
     - name: check changelog updated
       id: changelog_updated
       env:
@@ -27,14 +27,14 @@ jobs:
         echo $FILES | grep -qF 'CHANGELOG.md' || echo $PR_BODY | grep -qiF "$NO_CHANGELOG"
     - name: Reject pull request if no CHANGELOG update
       if: ${{ always() && steps.changelog_updated.outcome == 'failure' }}
-      uses: Ana06/automatic-pull-request-review@v0.1.0
+      uses: Ana06/automatic-pull-request-review@0cf4e8a17ba79344ed3fdd7fed6dd0311d08a9d4 # v0.1.0
       with:
         repo-token: ${{ secrets.GITHUB_TOKEN }}
         event: REQUEST_CHANGES
         body: "Please add bug fixes, new features, breaking changes and anything else you think is worthwhile mentioning to the `master (unreleased)` section of CHANGELOG.md. If no CHANGELOG update is needed add the following to the PR description: `${{ env.NO_CHANGELOG }}`"
         allow_duplicate: false
     - name: Dismiss previous review if CHANGELOG update
-      uses: Ana06/automatic-pull-request-review@v0.1.0
+      uses: Ana06/automatic-pull-request-review@0cf4e8a17ba79344ed3fdd7fed6dd0311d08a9d4 # v0.1.0
       with:
         repo-token: ${{ secrets.GITHUB_TOKEN }}
         event: DISMISS

.github/workflows/publish.yml

@@ -11,11 +11,11 @@ jobs:
   deploy:
     runs-on: ubuntu-20.04
     steps:
-      - uses: actions/checkout@v2
+      - uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
       - name: Set up Python
-        uses: actions/setup-python@v2
+        uses: actions/setup-python@d27e3f3d7c64b4bbf8e4abfb9b63b83e846e0435 # v4.5.0
         with:
-          python-version: '3.6'
+          python-version: '3.7'
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip
@@ -27,4 +27,3 @@ jobs:
         run: |
           python setup.py sdist bdist_wheel
           twine upload --skip-existing dist/*
-

.github/workflows/scorecard.yml

@@ -0,0 +1,72 @@
+# This workflow uses actions that are not certified by GitHub. They are provided
+# by a third-party and are governed by separate terms of service, privacy
+# policy, and support documentation.
+
+name: Scorecard supply-chain security
+on:
+  # For Branch-Protection check. Only the default branch is supported. See
+  # https://github.com/ossf/scorecard/blob/main/docs/checks.md#branch-protection
+  branch_protection_rule:
+  # To guarantee Maintained check is occasionally updated. See
+  # https://github.com/ossf/scorecard/blob/main/docs/checks.md#maintained
+  schedule:
+    - cron: '43 4 * * 3'
+  push:
+    branches: [ "master" ]
+
+# Declare default permissions as read only.
+permissions: read-all
+
+jobs:
+  analysis:
+    name: Scorecard analysis
+    runs-on: ubuntu-latest
+    permissions:
+      # Needed to upload the results to code-scanning dashboard.
+      security-events: write
+      # Needed to publish results and get a badge (see publish_results below).
+      id-token: write
+      # Uncomment the permissions below if installing in a private repository.
+      # contents: read
+      # actions: read
+
+    steps:
+      - name: "Checkout code"
+        uses: actions/checkout@93ea575cb5d8a053eaa0ac8fa3b40d7e05a33cc8 # v3.1.0
+        with:
+          persist-credentials: false
+
+      - name: "Run analysis"
+        uses: ossf/scorecard-action@99c53751e09b9529366343771cc321ec74e9bd3d # v2.0.6
+        with:
+          results_file: results.sarif
+          results_format: sarif
+          # (Optional) "write" PAT token. Uncomment the `repo_token` line below if:
+          # - you want to enable the Branch-Protection check on a *public* repository, or
+          # - you are installing Scorecard on a *private* repository
+          # To create the PAT, follow the steps in https://github.com/ossf/scorecard-action#authentication-with-pat.
+          # repo_token: ${{ secrets.SCORECARD_TOKEN }}
+
+          # Public repositories:
+          #   - Publish results to OpenSSF REST API for easy access by consumers
+          #   - Allows the repository to include the Scorecard badge.
+          #   - See https://github.com/ossf/scorecard-action#publishing-results.
+          # For private repositories:
+          #   - `publish_results` will always be set to `false`, regardless
+          #     of the value entered here.
+          publish_results: true
+
+      # Upload the results as artifacts (optional). Commenting out will disable uploads of run results in SARIF
+      # format to the repository Actions tab.
+      - name: "Upload artifact"
+        uses: actions/upload-artifact@3cea5372237819ed00197afe530f5a7ea3e805c8 # v3.1.0
+        with:
+          name: SARIF file
+          path: results.sarif
+          retention-days: 5
+
+      # Upload the results to GitHub's code scanning dashboard.
+      - name: "Upload to code-scanning"
+        uses: github/codeql-action/upload-sarif@807578363a7869ca324a79039e6db9c843e0e100 # v2.1.27
+        with:
+          sarif_file: results.sarif

.github/workflows/tag.yml

@@ -10,20 +10,21 @@ jobs:
     runs-on: ubuntu-20.04
     steps:
     - name: Checkout capa-rules
-      uses: actions/checkout@v2
+      uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
       with:
-        repository: fireeye/capa-rules
+        repository: mandiant/capa-rules
         token: ${{ secrets.CAPA_TOKEN }}
     - name: Tag capa-rules
       run: |
         # user information is needed to create annotated tags (with a message)
-        git config user.email 'capa-dev@fireeye.com'
+        git config user.email 'capa-dev@mandiant.com'
         git config user.name 'Capa Bot'
         name=${{ github.event.release.tag_name }}
-        git tag $name -m "https://github.com/fireeye/capa/releases/$name"
+        git tag $name -m "https://github.com/mandiant/capa/releases/$name"
+        # TODO update branch name-major=${name%%.*}
     - name: Push tag to capa-rules
-      uses: ad-m/github-push-action@master
+      uses: ad-m/github-push-action@0fafdd62b84042d49ec0cb92d9cac7f7ce4ec79e # master
       with:
-        repository: fireeye/capa-rules
+        repository: mandiant/capa-rules
         github_token: ${{ secrets.CAPA_TOKEN }}
         tags: true

.github/workflows/tests.yml

@@ -15,7 +15,7 @@ jobs:
     runs-on: ubuntu-20.04
     steps:
     - name: Checkout capa
-      uses: actions/checkout@v2
+      uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
     # The sync GH action in capa-rules relies on a single '- *$' in the CHANGELOG file
     - name: Ensure CHANGELOG has '- *$'
       run: |
@@ -26,31 +26,33 @@ jobs:
     runs-on: ubuntu-20.04
     steps:
     - name: Checkout capa
-      uses: actions/checkout@v2
+      uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
     - name: Set up Python 3.8
-      uses: actions/setup-python@v2
+      uses: actions/setup-python@d27e3f3d7c64b4bbf8e4abfb9b63b83e846e0435 # v4.5.0
       with:
-        python-version: 3.8
+        python-version: "3.8"
     - name: Install dependencies
       run: pip install -e .[dev]
     - name: Lint with isort
       run: isort --profile black --length-sort --line-width 120 -c .
     - name: Lint with black
       run: black -l 120 --check .
+    - name: Lint with pycodestyle
+      run: pycodestyle --show-source capa/ scripts/ tests/
     - name: Check types with mypy
-      run: mypy --config-file .github/mypy/mypy.ini capa/ scripts/ tests/
+      run: mypy --config-file .github/mypy/mypy.ini --check-untyped-defs capa/ scripts/ tests/
 
   rule_linter:
     runs-on: ubuntu-20.04
     steps:
     - name: Checkout capa with submodules
-      uses: actions/checkout@v2
+      uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
       with:
-        submodules: true
+        submodules: recursive
     - name: Set up Python 3.8
-      uses: actions/setup-python@v2
+      uses: actions/setup-python@d27e3f3d7c64b4bbf8e4abfb9b63b83e846e0435 # v4.5.0
       with:
-        python-version: 3.8
+        python-version: "3.8"
     - name: Install capa
       run: pip install -e .
     - name: Run rule linter
@@ -63,22 +65,22 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-20.04, windows-2019, macos-10.15]
+        os: [ubuntu-20.04, windows-2019, macos-11]
         # across all operating systems
-        python-version: [3.6, 3.9]
+        python-version: ["3.7", "3.11"]
         include:
           # on Ubuntu run these as well
           - os: ubuntu-20.04
-            python-version: 3.7
+            python-version: "3.8"
           - os: ubuntu-20.04
-            python-version: 3.8
+            python-version: "3.9"
     steps:
     - name: Checkout capa with submodules
-      uses: actions/checkout@v2
+      uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
       with:
-        submodules: true
+        submodules: recursive
     - name: Set up Python ${{ matrix.python-version }}
-      uses: actions/setup-python@v2
+      uses: actions/setup-python@d27e3f3d7c64b4bbf8e4abfb9b63b83e846e0435 # v4.5.0
       with:
         python-version: ${{ matrix.python-version }}
     - name: Install pyyaml

.gitignore

@@ -114,3 +114,11 @@ venv.bak/
 isort-output.log
 black-output.log
 rule-linter-output.log
+.vscode
+scripts/perf/*.txt
+scripts/perf/*.svg
+scripts/perf/*.zip
+.direnv
+.envrc
+.DS_Store
+*/.DS_Store

CHANGELOG.md

@@ -1,6 +1,5 @@
 # Change Log
 
-
 ## master (unreleased)
 
 ### New Features
@@ -18,10 +17,407 @@
 ### Development
 
 ### Raw diffs
-- [capa <release>...master](https://github.com/fireeye/capa/compare/v3.0.2...master)
-- [capa-rules <release>...master](https://github.com/fireeye/capa-rules/compare/v3.0.2...master)
+- [capa v5.0.0...master](https://github.com/mandiant/capa/compare/v5.0.0...master)
+- [capa-rules v5.0.0...master](https://github.com/mandiant/capa-rules/compare/v5.0.0...master)
+
+
+## v5.0.0 (2023-02-08)
+This capa version comes with major improvements and additions to better handle .NET binaries. To showcase this we've updated and added over 30 .NET rules.  
+
+Additionally, capa now caches its rule set for better performance. The capa explorer also caches its analysis results, so that multiple IDA Pro or plugin invocations don't need to repeat the same analysis.
+
+We have removed the SMDA backend and changed the program return codes to be positive numbers.
+
+Other improvements to highlight include better ELF OS detection, various rendering bug fixes, and enhancements to the feature extraction. We've also added support for Python 3.11.
+
+Thanks for all the support, especially to @jsoref, @bkojusner, @edeca, @richardweiss80, @joren485, @ryantxu1, @mwilliams31, @anushkavirgaonkar, @MalwareMechanic, @Still34, @dzbeck, @johnk3r, and everyone else who submitted bugs and provided feedback!
+
+### New Features
+- verify rule metadata format on load #1160 @mr-tz
+- dotnet: emit property features #1168 @anushkavirgaonkar
+- dotnet: emit API features for objects created via the newobj instruction #1186 @mike-hunhoff
+- dotnet: emit API features for generic methods #1231 @mike-hunhoff
+- Python 3.11 support #1192 @williballenthin
+- dotnet: emit calls to/from MethodDef methods #1236 @mike-hunhoff
+- dotnet: emit namespace/class features for ldvirtftn/ldftn instructions #1241 @mike-hunhoff
+- dotnet: emit namespace/class features for type references #1242 @mike-hunhoff
+- dotnet: extract dotnet and pe format #1187 @mr-tz
+- don't render all library rule matches in vverbose output #1174 @mr-tz
+- cache the rule set across invocations for better performance #1212 @williballenthin
+- update ATT&CK/MBC data for linting #1297 @mr-tz
+
+### Breaking Changes
+- remove SMDA backend #1062 @williballenthin
+- error return codes are now positive numbers #1269 @mr-tz
+
+### New Rules (77)
+
+- collection/use-dotnet-library-sharpclipboard @johnk3r
+- data-manipulation/encryption/aes/use-dotnet-library-encryptdecryptutils @johnk3r
+- data-manipulation/json/use-dotnet-library-newtonsoftjson @johnk3r
+- data-manipulation/svg/use-dotnet-library-sharpvectors @johnk3r
+- executable/resource/embed-dependencies-as-resources-using-fodycostura @johnk3r @mr-tz
+- communication/ftp/send/send-file-using-ftp michael.hunhof@mandiant.com anushka.virgaonkar@mandiant.com
+- nursery/extract-zip-archive anushka.virgaonkar@mandiant.com
+- nursery/allocate-unmanaged-memory-in-dotnet michael.hunhoff@mandiant.com
+- nursery/check-file-extension-in-dotnet michael.hunhoff@mandiant.com
+- nursery/decode-data-using-base64-in-dotnet michael.hunhoff@mandiant.com
+- nursery/deserialize-json-in-dotnet michael.hunhoff@mandiant.com
+- nursery/find-data-using-regex-in-dotnet michael.hunhoff@mandiant.com
+- nursery/generate-random-filename-in-dotnet michael.hunhoff@mandiant.com
+- nursery/get-os-version-in-dotnet michael.hunhoff@mandiant.com
+- nursery/load-xml-in-dotnet michael.hunhoff@mandiant.com
+- nursery/manipulate-unmanaged-memory-in-dotnet michael.hunhoff@mandiant.com
+- nursery/save-image-in-dotnet michael.hunhoff@mandiant.com
+- nursery/send-email-in-dotnet michael.hunhoff@mandiant.com
+- nursery/serialize-json-in-dotnet michael.hunhoff@mandiant.com
+- nursery/set-http-user-agent-in-dotnet michael.hunhoff@mandiant.com
+- nursery/compile-csharp-in-dotnet michael.hunhoff@mandiant.com
+- nursery/compile-visual-basic-in-dotnet michael.hunhoff@mandiant.com
+- nursery/compress-data-using-gzip-in-dotnet michael.hunhoff@mandiant.com
+- nursery/execute-sqlite-statement-in-dotnet michael.hunhoff@mandiant.com
+- nursery/execute-via-asynchronous-task-in-dotnet michael.hunhoff@mandiant.com
+- nursery/execute-via-timer-in-dotnet michael.hunhoff@mandiant.com
+- nursery/execute-wmi-query-in-dotnet michael.hunhoff@mandiant.com
+- nursery/manipulate-network-credentials-in-dotnet michael.hunhoff@mandiant.com
+- nursery/encrypt-data-using-aes william.ballenthin@mandiant.com Ivan Kwiatkowski (@JusticeRage)
+- host-interaction/uac/bypass/bypass-uac-via-rpc david.cannings@pwc.com david@edeca.net
+- nursery/check-for-vm-using-instruction-vpcext richard.weiss@mandiant.com
+- nursery/get-windows-directory-from-kuser_shared_data david.cannings@pwc.com
+- nursery/encrypt-data-using-openssl-dsa Ana06
+- nursery/encrypt-data-using-openssl-ecdsa Ana06
+- nursery/encrypt-data-using-openssl-rsa Ana06
+- runtime/dotnet/execute-via-dotnet-startup-hook william.ballenthin@mandiant.com
+- host-interaction/console/manipulate-console-buffer william.ballenthin@mandiant.com michael.hunhoff@mandiant.com
+- nursery/access-wmi-data-in-dotnet michael.hunhoff@mandiant.com
+- nursery/allocate-unmanaged-memory-via-dotnet michael.hunhoff@mandiant.com
+- nursery/generate-random-bytes-in-dotnet michael.hunhoff@mandiant.com
+- nursery/manipulate-console-window michael.hunhoff@mandiant.com
+- nursery/obfuscated-with-koivm michael.hunhoff@mandiant.com
+- nursery/implement-com-dll moritz.raabe@mandiant.com
+- nursery/linked-against-libsodium @mr-tz
+- compiler/nuitka/compiled-with-nuitka @williballenthin
+- nursery/authenticate-data-with-md5-mac william.ballenthin@mandiant.com
+- nursery/resolve-function-by-djb2-hash still@teamt5.org
+- host-interaction/mutex/create-semaphore-on-linux @ramen0x3f
+- host-interaction/mutex/lock-semaphore-on-linux @ramen0x3f
+- host-interaction/mutex/unlock-semaphore-on-linux @ramen0x3f
+- data-manipulation/hashing/sha384/hash-data-using-sha384 william.ballenthin@mandiant.com
+- data-manipulation/hashing/sha512/hash-data-using-sha512 william.ballenthin@mandiant.com
+- nursery/decode-data-using-url-encoding michael.hunhoff@mandiant.com
+- nursery/manipulate-user-privileges michael.hunhoff@mandiant.com
+- lib/get-os-version @mr-tz
+- nursery/decrypt-data-using-tea william.ballenthin@mandiant.com
+- nursery/encrypt-data-using-tea william.ballenthin@mandiant.com
+- nursery/hash-data-using-whirlpool william.ballenthin@mandiant.com
+- nursery/reference-base58-string william.ballenthin@mandiant.com
+- communication/mailslot/create-mailslot william.ballenthin@mandiant.com
+- executable/resource/access-dotnet-resource @mr-tz
+- linking/static/linked-against-cpp-standard-library @mr-tz
+- data-manipulation/compression/compress-data-using-lzo david@edeca.net david.cannings@pwc.com
+- data-manipulation/compression/decompress-data-using-lzo david@edeca.net david.cannings@pwc.com
+- communication/socket/tcp/create-tcp-socket-via-raw-afd-driver william.ballenthin@mandiant.com
+- host-interaction/process/map-section-object william.ballenthin@mandiant.com
+- lib/create-or-open-section-object william.ballenthin@mandiant.com
+- load-code/dotnet/execute-dotnet-assembly-via-clr-host blas.kojusner@mandiant.com
+- load-code/execute-vbscript-javascript-or-jscript-in-memory blas.kojusner@mandiant.com
+- host-interaction/file-system/reference-absolute-stream-path-on-windows blas.kojusner@mandiant.com
+- nursery/generate-method-via-reflection-in-dotnet michael.hunhoff@mandiant.com
+- nursery/unmanaged-call-via-dynamic-pinvoke-in-dotnet michael.hunhoff@mandiant.com
+
+### Bug Fixes
+- render: convert feature attributes to aliased dictionary for vverbose #1152 @mike-hunhoff
+- decouple Token dependency / extractor and features #1139 @mr-tz
+- update pydantic model to guarantee type coercion #1176 @mike-hunhoff
+- do not overwrite version in version.py during PyInstaller build #1169 @mr-tz
+- render: fix vverbose rendering of offsets #1215 @williballenthin
+- elf: better detect OS via GLIBC ABI version needed and dependencies #1221 @williballenthin
+- dotnet: address unhandled exceptions with improved type checking #1230 @mike-hunhoff
+- fix import-to-ida script formatting #1208 @williballenthin
+- render: fix verbose rendering of scopes #1263 @williballenthin
+- rules: better detect invalid rules #1282 @williballenthin
+- show-features: better render strings with embedded whitespace #1267 @williballenthin
+- handle vivisect bug around strings at instruction level, use min length 4 #1271 @williballenthin @mr-tz
+- extractor: guard against invalid "calls from" features #1177 @mr-tz
+- extractor: add format to global features #1258 @mr-tz
+- extractor: discover all strings with length >= 4 #1280 @mr-tz
+- extractor: don't extract byte features for strings #1293 @mr-tz
+
+### capa explorer IDA Pro plugin
+- fix: display instruction items #1154 @mr-tz
+- fix: accept only plaintext pasted content #1194 @williballenthin
+- fix: UnboundLocalError #1217 @williballenthin
+- extractor: add support for COFF files and extern functions #1223 @mike-hunhoff
+- doc: improve error messaging and documentation related to capa rule set #1249 @mike-hunhoff
+- fix: assume 32-bit displacement for offsets #1250 @mike-hunhoff
+- generator: refactor caching and matching #1251 @mike-hunhoff
+- fix: improve exception handling to prevent IDA from locking up when errors occur #1262 @mike-hunhoff
+- verify rule metadata using Pydantic #1167 @mr-tz
+- extractor: make read consistent with file object behavior #1254 @mr-tz
+- fix: UnboundLocalError x2 #1302 @mike-hunhoff
+- cache capa results across IDA sessions #1279 @mr-tz
+
+### Raw diffs
+- [capa v4.0.1...v5.0.0](https://github.com/mandiant/capa/compare/v4.0.1...v5.0.0)
+- [capa-rules v4.0.1...v5.0.0](https://github.com/mandiant/capa-rules/compare/v4.0.1...v5.0.0)
+
+
+## v4.0.1 (2022-08-15)
+Some rules contained invalid metadata fields that caused an error when rendering rule hits. We've updated all rules and enhanced the rule linter to catch such issues.
+
+### New Rules (1)
+
+- anti-analysis/obfuscation/obfuscated-with-vs-obfuscation jakub.jozwiak@mandiant.com
+
+
+### Bug Fixes
+- linter: use pydantic to validate rule metadata #1141 @mike-hunhoff
+- build binaries using PyInstaller no longer overwrites functions in version.py #1136 @mr-tz
+
+### Raw diffs
+- [capa v4.0.0...v4.0.1](https://github.com/mandiant/capa/compare/v4.0.0...v4.0.1)
+- [capa-rules v4.0.0...v4.0.1](https://github.com/mandiant/capa-rules/compare/v4.0.0...v4.0.1)
+
+## v4.0.0 (2022-08-10)
+Version 4 adds support for analyzing .NET executables. capa will autodetect .NET modules, or you can explicitly invoke the new feature extractor via `--format dotnet`. We've also extended the rule syntax for .NET features including `namespace` and `class`.
+
+Additionally, new `instruction` scope and `operand` features enable users to create more explicit rules. These features are not backwards compatible. We removed the previously used `/x32` and `/x64` flavors of number and operand features.
+
+We updated 49 existing rules and added 22 new rules leveraging these new features and characteristics to detect capabilities seen in .NET malware.
+
+More breaking changes include updates to the JSON results document, freeze file format schema (now format version v2), and the internal handling of addresses.
+
+Thanks for all the support, especially to @htnhan, @jtothej, @sara-rn, @anushkavirgaonkar, and @_re_fox!
+
+*Deprecation warning: v4.0 will be the last capa version to support the SMDA backend.*
+
+### New Features
+
+ - add new scope "instruction" for matching mnemonics and operands #767 @williballenthin
+ - add new feature "operand[{0, 1, 2}].number" for matching instruction operand immediate values #767 @williballenthin
+ - add new feature "operand[{0, 1, 2}].offset" for matching instruction operand offsets #767 @williballenthin
+ - extract additional offset/number features in certain circumstances #320 @williballenthin
+ - add detection and basic feature extraction for dotnet #987 @mr-tz, @mike-hunhoff, @williballenthin
+ - add file string extraction for dotnet files #1012 @mike-hunhoff
+ - add file function-name extraction for dotnet files #1015 @mike-hunhoff
+ - add unmanaged call characteristic for dotnet files #1023 @mike-hunhoff
+ - add mixed mode characteristic feature extraction for dotnet files #1024 @mike-hunhoff
+ - emit class and namespace features for dotnet files #1030 @mike-hunhoff
+ - render: support Addresses that aren't simple integers, like .NET token+offset #981 @williballenthin
+ - document rule tags and branches #1006 @williballenthin, @mr-tz
+
+### Breaking Changes
+
+  - instruction scope and operand feature are new and are not backwards compatible with older versions of capa
+  - Python 3.7 is now the minimum supported Python version #866 @williballenthin
+  - remove /x32 and /x64 flavors of number and operand features #932 @williballenthin
+  - the tool now accepts multiple paths to rules, and JSON doc updated accordingly @williballenthin
+  - extractors must use handles to identify functions/basic blocks/instructions #981 @williballenthin
+  - the freeze file format schema was updated, including format version bump to v2 #986 @williballenthin
+
+Deprecation notice: as described in [#937](https://github.com/mandiant/capa/issues/937), we plan to remove the SMDA backend for v5. If you rely on this backend, please reach out so we can discuss extending the support for SMDA or transitioning your workflow to use vivisect.
+
+### New Rules (30)
+
+- data-manipulation/encryption/aes/manually-build-aes-constants huynh.t.nhan@gmail.com
+- nursery/get-process-image-filename michael.hunhoff@mandiant.com
+- compiler/v/compiled-with-v jakub.jozwiak@mandiant.com
+- compiler/zig/compiled-with-zig jakub.jozwiak@mandiant.com
+- anti-analysis/packer/huan/packed-with-huan jakub.jozwiak@mandiant.com
+- internal/limitation/file/internal-dotnet-file-limitation william.ballenthin@mandiant.com
+- nursery/get-os-information-via-kuser_shared_data @mr-tz
+- load-code/pe/resolve-function-by-parsing-PE-exports @sara-rn
+- anti-analysis/packer/huan/packed-with-huan jakub.jozwiak@mandiant.com
+- nursery/execute-dotnet-assembly anushka.virgaonkar@mandiant.com
+- nursery/invoke-dotnet-assembly-method anushka.virgaonkar@mandiant.com
+- collection/screenshot/capture-screenshot-via-keybd-event @_re_fox
+- collection/browser/gather-chrome-based-browser-login-information @_re_fox
+- nursery/power-down-monitor michael.hunhoff@mandiant.com
+- nursery/hash-data-using-aphash @_re_fox
+- nursery/hash-data-using-jshash @_re_fox
+- host-interaction/file-system/files/list/enumerate-files-on-windows moritz.raabe@mandiant.com anushka.virgaonkar@mandiant.com
+- nursery/check-clipboard-data anushka.virgaonkar@mandiant.com
+- nursery/clear-clipboard-data anushka.virgaonkar@mandiant.com
+- nursery/compile-dotnet-assembly anushka.virgaonkar@mandiant.com
+- nursery/create-process-via-wmi anushka.virgaonkar@mandiant.com
+- nursery/display-service-notification-message-box anushka.virgaonkar@mandiant.com
+- nursery/find-process-by-name anushka.virgaonkar@mandiant.com
+- nursery/generate-random-numbers-in-dotnet anushka.virgaonkar@mandiant.com
+- nursery/send-keystrokes anushka.virgaonkar@mandiant.com
+- nursery/send-request-in-dotnet anushka.virgaonakr@mandiant.com
+- nursery/terminate-process-by-name-in-dotnet anushka.virgaonkar@mandiant.com
+- nursery/hash-data-using-rshash @_re_fox
+- persistence/authentication-process/act-as-credential-manager-dll jakub.jozwiak@mandiant.com
+- persistence/authentication-process/act-as-password-filter-dll jakub.jozwiak@mandiant.com
+
+### Bug Fixes
+- improve handling _ prefix compile/link artifact #924 @mike-hunhoff
+- better detect OS in ELF samples #988 @williballenthin
+- display number feature zero in vverbose #1097 @mike-hunhoff
+
+### capa explorer IDA Pro plugin
+- improve file format extraction #918 @mike-hunhoff
+- remove decorators added by IDA to ELF imports #919 @mike-hunhoff
+- bug fixes for Address abstraction #1091 @mike-hunhoff
+
+### Development
+
+### Raw diffs
+- [capa v3.2.0...v4.0.0](https://github.com/mandiant/capa/compare/v3.2.0...master)
+- [capa-rules v3.2.0...v4.0.0](https://github.com/mandiant/capa-rules/compare/v3.2.0...master)
+
+## v3.2.1 (2022-06-06)
+This out-of-band release bumps the SMDA dependency version to enable installation on Python 3.10.
+
+### Bug Fixes
+
+- update SMDA dependency @mike-hunhoff #922
+
+### Raw diffs
+- [capa v3.2.0...v3.2.1](https://github.com/mandiant/capa/compare/v3.2.0...v3.2.1)
+- [capa-rules v3.2.0...v3.2.1](https://github.com/mandiant/capa-rules/compare/v3.2.0...v3.2.1)
+
+## v3.2.0 (2022-03-03)
+This release adds a new characteristic `characteristic: call $+5` enabling users to create more explicit rules. The linter now also validates ATT&CK and MBC categories. Additionally, many dependencies, including the vivisect backend, have been updated.
+
+One rule has been added and many more have been improved.
+
+Thanks for all the support, especially to @kn0wl3dge and first time contributor @uckelman-sf!
+
+### New Features
+
+- linter: validate ATT&CK/MBC categories and IDs #103 @kn0wl3dge
+- extractor: add characteristic "call $+5" feature #366 @kn0wl3dge
+
+### New Rules (1)
+
+- anti-analysis/obfuscation/obfuscated-with-advobfuscator jakub.jozwiak@mandiant.com
+
+### Bug Fixes
+
+- remove typing package as a requirement for Python 3.7+ compatibility #901 @uckelman-sf
+- elf: fix OS detection for Linux kernel modules #867 @williballenthin
+
+### Raw diffs
+- [capa v3.1.0...v3.2.0](https://github.com/mandiant/capa/compare/v3.1.0...v3.2.0)
+- [capa-rules v3.1.0...v3.2.0](https://github.com/mandiant/capa-rules/compare/v3.1.0...v3.2.0)
+
+## v3.1.0 (2022-01-10)
+This release improves the performance of capa while also adding 23 new rules and many code quality enhancements. We profiled capa's CPU usage and optimized the way that it matches rules, such as by short circuiting when appropriate. According to our testing, the matching phase is approximately 66% faster than v3.0.3! We also added support for Python 3.10, aarch64 builds, and additional MAEC metadata in the rule headers.
   
-## v3.0.2
+This release adds 23 new rules, including nine by Jakub Jozwiak of Mandiant. @ryantxu1 and @dzbeck updated the ATT&CK and MBC mappings for many rules. Thank you!
+  
+And as always, welcome first time contributors!
+
+  - @kn0wl3dge
+  - @jtothej
+  - @cl30
+  
+
+### New Features
+
+- engine: short circuit logic nodes for better performance #824 @williballenthin
+- engine: add optimizer the order faster nodes first #829 @williballenthin
+- engine: optimize rule evaluation by skipping rules that can't match #830 @williballenthin
+- support python 3.10 #816 @williballenthin
+- support aarch64 #683 @williballenthin
+- rules: support maec/malware-family meta #841 @mr-tz
+- engine: better type annotations/exhaustiveness checking #839 @cl30
+
+### Breaking Changes: None
+
+### New Rules (23)
+
+- nursery/delete-windows-backup-catalog michael.hunhoff@mandiant.com
+- nursery/disable-automatic-windows-recovery-features michael.hunhoff@mandiant.com
+- nursery/capture-webcam-video @johnk3r
+- nursery/create-registry-key-via-stdregprov michael.hunhoff@mandiant.com
+- nursery/delete-registry-key-via-stdregprov michael.hunhoff@mandiant.com
+- nursery/delete-registry-value-via-stdregprov michael.hunhoff@mandiant.com
+- nursery/query-or-enumerate-registry-key-via-stdregprov michael.hunhoff@mandiant.com
+- nursery/query-or-enumerate-registry-value-via-stdregprov michael.hunhoff@mandiant.com
+- nursery/set-registry-value-via-stdregprov michael.hunhoff@mandiant.com
+- data-manipulation/compression/decompress-data-using-ucl jakub.jozwiak@mandiant.com
+- linking/static/wolfcrypt/linked-against-wolfcrypt jakub.jozwiak@mandiant.com
+- linking/static/wolfssl/linked-against-wolfssl jakub.jozwiak@mandiant.com
+- anti-analysis/packer/pespin/packed-with-pespin jakub.jozwiak@mandiant.com
+- load-code/shellcode/execute-shellcode-via-windows-fibers jakub.jozwiak@mandiant.com
+- load-code/shellcode/execute-shellcode-via-enumuilanguages jakub.jozwiak@mandiant.com
+- anti-analysis/packer/themida/packed-with-themida william.ballenthin@mandiant.com
+- load-code/shellcode/execute-shellcode-via-createthreadpoolwait jakub.jozwiak@mandiant.com
+- host-interaction/process/inject/inject-shellcode-using-a-file-mapping-object jakub.jozwiak@mandiant.com
+- load-code/shellcode/execute-shellcode-via-copyfile2 jakub.jozwiak@mandiant.com
+- malware-family/plugx/match-known-plugx-module still@teamt5.org
+
+### Rule Changes
+
+  - update ATT&CK mappings by @ryantxu1
+  - update ATT&CK and MBC mappings by @dzbeck
+  - aplib detection by @cdong1012
+  - golang runtime detection by @stevemk14eber
+
+### Bug Fixes
+
+- fix circular import error #825 @williballenthin
+- fix smda negative number extraction #430 @kn0wl3dge
+
+### capa explorer IDA Pro plugin
+
+- pin supported versions to >= 7.4 and < 8.0 #849 @mike-hunhoff
+
+### Development
+
+- add profiling infrastructure #828 @williballenthin
+- linter: detect shellcode extension #820 @mr-tz
+- show features script: add backend flag #430 @kn0wl3dge
+
+### Raw diffs
+- [capa v3.0.3...v3.1.0](https://github.com/mandiant/capa/compare/v3.0.3...v3.1.0)
+- [capa-rules v3.0.3...v3.1.0](https://github.com/mandiant/capa-rules/compare/v3.0.3...v3.1.0)
+
+
+## v3.0.3 (2021-10-27)
+
+This is primarily a rule maintenance release:
+  - eight new rules, including all relevant techniques from [ATT&CK v10](https://medium.com/mitre-attack/introducing-attack-v10-7743870b37e3), and
+  - two rules removed, due to the prevalence of false positives
+
+We've also tweaked the status codes returned by capa.exe to be more specific and added a bit more metadata to the JSON output format.
+ 
+As always, welcome first time contributors!
+  - still@teamt5.org
+  - zander.work@mandiant.com
+                                                                                                     
+
+### New Features
+
+- show in which function a BB match is #130 @williballenthin
+- main: exit with unique error codes when bailing #802 @williballenthin
+
+### New Rules (8)
+
+- nursery/resolve-function-by-fnv-1a-hash still@teamt5.org
+- data-manipulation/encryption/encrypt-data-using-memfrob-from-glibc zander.work@mandiant.com
+- collection/group-policy/discover-group-policy-via-gpresult william.ballenthin@mandiant.com
+- host-interaction/bootloader/manipulate-safe-mode-programs william.ballenthin@mandiant.com
+- nursery/enable-safe-mode-boot william.ballenthin@mandiant.com
+- persistence/iis/persist-via-iis-module william.ballenthin@mandiant.com
+- persistence/iis/persist-via-isapi-extension william.ballenthin@mandiant.com
+- targeting/language/identify-system-language-via-api william.ballenthin@mandiant.com
+
+## Removed rules (2)
+- load-code/pe/parse-pe-exports: too many false positives in unrelated structure accesses
+- anti-analysis/anti-vm/vm-detection/execute-anti-vm-instructions: too many false positives in junk code
+
+### Bug Fixes
+
+- update references from FireEye to Mandiant
+
+### Raw diffs
+- [capa v3.0.2...v3.0.3](https://github.com/fireeye/capa/compare/v3.0.2...v3.0.3)
+- [capa-rules v3.0.2...v3.0.3](https://github.com/fireeye/capa-rules/compare/v3.0.2...v3.0.3)
+  
+## v3.0.2 (2021-09-28)
   
 This release fixes an issue with the standalone executables built with PyInstaller when running capa against ELF files.
 
@@ -30,8 +426,8 @@ This release fixes an issue with the standalone executables built with PyInstall
 - fix bug in PyInstaller config preventing ELF analysis #795 @mr-tz
 
 ### Raw diffs
-- [capa v3.0.1...master](https://github.com/fireeye/capa/compare/v3.0.1...v3.0.2)
-- [capa-rules v3.0.1...master](https://github.com/fireeye/capa-rules/compare/v3.0.1...v3.0.2)
+- [capa v3.0.1...v3.0.2](https://github.com/fireeye/capa/compare/v3.0.1...v3.0.2)
+- [capa-rules v3.0.1...v3.0.2](https://github.com/fireeye/capa-rules/compare/v3.0.1...v3.0.2)
 
 ## v3.0.1 (2021-09-27)
 
@@ -83,29 +479,29 @@ Also, welcome first time contributors:
 ### New Rules (80)
 
 - collection/webcam/capture-webcam-image @johnk3r
-- nursery/list-drag-and-drop-files michael.hunhoff@fireeye.com
-- nursery/monitor-clipboard-content michael.hunhoff@fireeye.com
-- nursery/monitor-local-ipv4-address-changes michael.hunhoff@fireeye.com
-- nursery/load-windows-common-language-runtime michael.hunhoff@fireeye.com
-- nursery/resize-volume-shadow-copy-storage michael.hunhoff@fireeye.com
-- nursery/add-user-account-group michael.hunhoff@fireeye.com
-- nursery/add-user-account-to-group michael.hunhoff@fireeye.com
-- nursery/add-user-account michael.hunhoff@fireeye.com
-- nursery/change-user-account-password michael.hunhoff@fireeye.com
-- nursery/delete-user-account-from-group michael.hunhoff@fireeye.com
-- nursery/delete-user-account-group michael.hunhoff@fireeye.com
-- nursery/delete-user-account michael.hunhoff@fireeye.com
-- nursery/list-domain-servers michael.hunhoff@fireeye.com
-- nursery/list-groups-for-user-account michael.hunhoff@fireeye.com
-- nursery/list-user-account-groups michael.hunhoff@fireeye.com
-- nursery/list-user-accounts-for-group michael.hunhoff@fireeye.com
-- nursery/list-user-accounts michael.hunhoff@fireeye.com
-- nursery/parse-url michael.hunhoff@fireeye.com
-- nursery/register-raw-input-devices michael.hunhoff@fireeye.com
-- anti-analysis/packer/gopacker/packed-with-gopacker jared.wilson@fireeye.com
+- nursery/list-drag-and-drop-files michael.hunhoff@mandiant.com
+- nursery/monitor-clipboard-content michael.hunhoff@mandiant.com
+- nursery/monitor-local-ipv4-address-changes michael.hunhoff@mandiant.com
+- nursery/load-windows-common-language-runtime michael.hunhoff@mandiant.com
+- nursery/resize-volume-shadow-copy-storage michael.hunhoff@mandiant.com
+- nursery/add-user-account-group michael.hunhoff@mandiant.com
+- nursery/add-user-account-to-group michael.hunhoff@mandiant.com
+- nursery/add-user-account michael.hunhoff@mandiant.com
+- nursery/change-user-account-password michael.hunhoff@mandiant.com
+- nursery/delete-user-account-from-group michael.hunhoff@mandiant.com
+- nursery/delete-user-account-group michael.hunhoff@mandiant.com
+- nursery/delete-user-account michael.hunhoff@mandiant.com
+- nursery/list-domain-servers michael.hunhoff@mandiant.com
+- nursery/list-groups-for-user-account michael.hunhoff@mandiant.com
+- nursery/list-user-account-groups michael.hunhoff@mandiant.com
+- nursery/list-user-accounts-for-group michael.hunhoff@mandiant.com
+- nursery/list-user-accounts michael.hunhoff@mandiant.com
+- nursery/parse-url michael.hunhoff@mandiant.com
+- nursery/register-raw-input-devices michael.hunhoff@mandiant.com
+- anti-analysis/packer/gopacker/packed-with-gopacker jared.wilson@mandiant.com
 - host-interaction/driver/create-device-object @mr-tz
 - host-interaction/process/create/execute-command @mr-tz
-- data-manipulation/encryption/create-new-key-via-cryptacquirecontext chuong.dong@fireeye.com
+- data-manipulation/encryption/create-new-key-via-cryptacquirecontext chuong.dong@mandiant.com
 - host-interaction/log/clfs/append-data-to-clfs-log-container blaine.stancill@mandiant.com
 - host-interaction/log/clfs/read-data-from-clfs-log-container blaine.stancill@mandiant.com
 - data-manipulation/encryption/hc-128/encrypt-data-using-hc-128-via-wolfssl blaine.stancill@mandiant.com
@@ -127,15 +523,15 @@ Also, welcome first time contributors:
 - persistence/persist-via-shell-profile-or-rc-file joakim@intezer.com
 - persistence/service/persist-via-rc-script joakim@intezer.com
 - collection/get-current-user-on-linux joakim@intezer.com
-- collection/network/get-mac-address-on-windows moritz.raabe@fireeye.com
-- host-interaction/file-system/read/read-file-on-linux moritz.raabe@fireeye.com joakim@intezer.com
-- host-interaction/file-system/read/read-file-on-windows moritz.raabe@fireeye.com
-- host-interaction/file-system/write/write-file-on-windows william.ballenthin@fireeye.com
-- host-interaction/os/info/get-system-information-on-windows moritz.raabe@fireeye.com joakim@intezer.com
-- host-interaction/process/create/create-process-on-windows moritz.raabe@fireeye.com
-- linking/runtime-linking/link-function-at-runtime-on-windows moritz.raabe@fireeye.com
+- collection/network/get-mac-address-on-windows moritz.raabe@mandiant.com
+- host-interaction/file-system/read/read-file-on-linux moritz.raabe@mandiant.com joakim@intezer.com
+- host-interaction/file-system/read/read-file-on-windows moritz.raabe@mandiant.com
+- host-interaction/file-system/write/write-file-on-windows william.ballenthin@mandiant.com
+- host-interaction/os/info/get-system-information-on-windows moritz.raabe@mandiant.com joakim@intezer.com
+- host-interaction/process/create/create-process-on-windows moritz.raabe@mandiant.com
+- linking/runtime-linking/link-function-at-runtime-on-windows moritz.raabe@mandiant.com
 - nursery/create-process-on-linux joakim@intezer.com
-- nursery/enumerate-files-on-linux william.ballenthin@fireeye.com
+- nursery/enumerate-files-on-linux william.ballenthin@mandiant.com
 - nursery/get-mac-address-on-linux joakim@intezer.com
 - nursery/get-system-information-on-linux joakim@intezer.com
 - nursery/link-function-at-runtime-on-linux joakim@intezer.com
@@ -167,8 +563,8 @@ Also, welcome first time contributors:
 ### Development
 
 ### Raw diffs
-- [capa v2.0.0...v3.0.0](https://github.com/fireeye/capa/compare/v2.0.0...v3.0.0)
-- [capa-rules v2.0.0...v3.0.0](https://github.com/fireeye/capa-rules/compare/v2.0.0...v3.0.0)
+- [capa v2.0.0...v3.0.0](https://github.com/mandiant/capa/compare/v2.0.0...v3.0.0)
+- [capa-rules v2.0.0...v3.0.0](https://github.com/mandiant/capa-rules/compare/v2.0.0...v3.0.0)
 
 
 ## v2.0.0 (2021-07-19)
@@ -187,7 +583,7 @@ A huge thanks to everyone who submitted issues, provided feedback, and contribut
 
 ### New Features
 
-- rules: update ATT&CK and MBC mappings https://github.com/fireeye/capa-rules/pull/317 @williballenthin
+- rules: update ATT&CK and MBC mappings https://github.com/mandiant/capa-rules/pull/317 @williballenthin
 - main: use FLIRT signatures to identify and ignore library code #446 @williballenthin
 - tests: update test cases and caching #545 @mr-tz
 - scripts: capa2yara.py convert capa rules to YARA rules #561 @ruppde
@@ -275,34 +671,34 @@ A huge thanks to everyone who submitted issues, provided feedback, and contribut
 - nursery/run-in-container @williballenthin
 - persistence/registry/appinitdlls/disable-appinit_dlls-code-signature-enforcement @williballenthin
 - collection/password-manager/steal-keepass-passwords-using-keefarce @Ana06
-- host-interaction/network/connectivity/check-internet-connectivity-via-wininet matthew.williams@fireeye.com michael.hunhoff@fireeye.com
+- host-interaction/network/connectivity/check-internet-connectivity-via-wininet matthew.williams@mandiant.com michael.hunhoff@mandiant.com
 - nursery/create-bits-job @mr-tz
 - nursery/execute-syscall-instruction @kulinacs @mr-tz
-- nursery/connect-to-wmi-namespace-via-wbemlocator michael.hunhoff@fireeye.com
+- nursery/connect-to-wmi-namespace-via-wbemlocator michael.hunhoff@mandiant.com
 - anti-analysis/obfuscation/obfuscated-with-callobfuscator johnk3r
 - executable/installer/inno-setup/packaged-as-an-inno-setup-installer awillia2@cisco.com
 - data-manipulation/hashing/djb2/hash-data-using-djb2 awillia2@cisco.com
-- data-manipulation/encoding/base64/decode-data-using-base64-via-dword-translation-table gilbert.elliot@fireeye.com
-- nursery/list-tcp-connections-and-listeners michael.hunhoff@fireeye.com
-- nursery/list-udp-connections-and-listeners michael.hunhoff@fireeye.com
-- nursery/log-keystrokes-via-raw-input-data michael.hunhoff@fireeye.com
-- nursery/register-http-server-url michael.hunhoff@fireeye.com
-- internal/limitation/file/internal-autoit-file-limitation.yml william.ballenthin@fireeye.com
-- internal/limitation/file/internal-dotnet-file-limitation.yml william.ballenthin@fireeye.com
-- internal/limitation/file/internal-installer-file-limitation.yml william.ballenthin@fireeye.com
-- internal/limitation/file/internal-packer-file-limitation.yml william.ballenthin@fireeye.com
+- data-manipulation/encoding/base64/decode-data-using-base64-via-dword-translation-table gilbert.elliot@mandiant.com
+- nursery/list-tcp-connections-and-listeners michael.hunhoff@mandiant.com
+- nursery/list-udp-connections-and-listeners michael.hunhoff@mandiant.com
+- nursery/log-keystrokes-via-raw-input-data michael.hunhoff@mandiant.com
+- nursery/register-http-server-url michael.hunhoff@mandiant.com
+- internal/limitation/file/internal-autoit-file-limitation.yml william.ballenthin@mandiant.com
+- internal/limitation/file/internal-dotnet-file-limitation.yml william.ballenthin@mandiant.com
+- internal/limitation/file/internal-installer-file-limitation.yml william.ballenthin@mandiant.com
+- internal/limitation/file/internal-packer-file-limitation.yml william.ballenthin@mandiant.com
 - host-interaction/network/domain/enumerate-domain-computers-via-ldap awillia2@cisco.com
 - host-interaction/network/domain/get-domain-controller-name awillia2@cisco.com
 - internal/limitation/file/internal-visual-basic-file-limitation @mr-tz
-- data-manipulation/hashing/md5/hash-data-with-md5 moritz.raabe@fireeye.com
+- data-manipulation/hashing/md5/hash-data-with-md5 moritz.raabe@mandiant.com
 - compiler/autohotkey/compiled-with-autohotkey awillia2@cisco.com
 - internal/limitation/file/internal-autohotkey-file-limitation @mr-tz
-- host-interaction/process/dump/create-process-memory-minidump michael.hunhoff@fireeye.com
-- nursery/get-storage-device-properties michael.hunhoff@fireeye.com
-- nursery/execute-shell-command-via-windows-remote-management michael.hunhoff@fireeye.com
-- nursery/get-token-privileges michael.hunhoff@fireeye.com
-- nursery/prompt-user-for-credentials michael.hunhoff@fireeye.com
-- nursery/spoof-parent-pid michael.hunhoff@fireeye.com
+- host-interaction/process/dump/create-process-memory-minidump michael.hunhoff@mandiant.com
+- nursery/get-storage-device-properties michael.hunhoff@mandiant.com
+- nursery/execute-shell-command-via-windows-remote-management michael.hunhoff@mandiant.com
+- nursery/get-token-privileges michael.hunhoff@mandiant.com
+- nursery/prompt-user-for-credentials michael.hunhoff@mandiant.com
+- nursery/spoof-parent-pid michael.hunhoff@mandiant.com
 
 ### Bug Fixes
 
@@ -322,8 +718,8 @@ A huge thanks to everyone who submitted issues, provided feedback, and contribut
 ### Development
 
 - ci: add capa release link to capa-rules tag #517 @Ana06
-- ci, changelog: update `New Rules` section in CHANGELOG automatically https://github.com/fireeye/capa-rules/pull/374 #549 #604 @Ana06
-- ci, changelog: support multiple author in sync GH https://github.com/fireeye/capa-rules/pull/378 @Ana06
+- ci, changelog: update `New Rules` section in CHANGELOG automatically https://github.com/mandiant/capa-rules/pull/374 #549 #604 @Ana06
+- ci, changelog: support multiple author in sync GH https://github.com/mandiant/capa-rules/pull/378 @Ana06
 - ci, lint: check statements for single child statements #563 @mr-tz
 - ci: reject PRs without CHANGELOG update to ensure CHANGELOG is kept up-to-date #584 @Ana06
 - ci: test that scripts run #660 @mr-tz
@@ -331,8 +727,8 @@ A huge thanks to everyone who submitted issues, provided feedback, and contribut
 ### Raw diffs
 
 <!-- The diff uses v1.6.1 because master doesn't include v1.6.2 and v1.6.3 -->
-- [capa v1.6.1...v2.0.0](https://github.com/fireeye/capa/compare/v1.6.1...v2.0.0)
-- [capa-rules v1.6.1...v2.0.0](https://github.com/fireeye/capa-rules/compare/v1.6.1...v2.0.0)
+- [capa v1.6.1...v2.0.0](https://github.com/mandiant/capa/compare/v1.6.1...v2.0.0)
+- [capa-rules v1.6.1...v2.0.0](https://github.com/mandiant/capa-rules/compare/v1.6.1...v2.0.0)
 
 
 ## v1.6.3 (2021-04-29)
@@ -345,7 +741,7 @@ This release adds IDA 7.6 support to capa.
 
 ### Raw diffs
 
-  - [capa v1.6.2...v1.6.3](https://github.com/fireeye/capa/compare/v1.6.2...v1.6.3)
+  - [capa v1.6.2...v1.6.3](https://github.com/mandiant/capa/compare/v1.6.2...v1.6.3)
 
 
 ## v1.6.2 (2021-04-13)
@@ -358,7 +754,7 @@ This release backports a fix to capa 1.6: The Windows binary was built with Pyth
 
 ### Raw diffs
 
-  - [capa v1.6.1...v1.6.2](https://github.com/fireeye/capa/compare/v1.6.1...v1.6.2)
+  - [capa v1.6.1...v1.6.2](https://github.com/mandiant/capa/compare/v1.6.1...v1.6.2)
 
 
 ## v1.6.1 (2021-04-07)
@@ -426,8 +822,8 @@ This release includes several bug fixes, such as a vivisect issue that prevented
 
 ### Raw diffs
 
-  - [capa v1.6.0...v1.6.1](https://github.com/fireeye/capa/compare/v1.6.0...v1.6.1)
-  - [capa-rules v1.6.0...v1.6.1](https://github.com/fireeye/capa-rules/compare/v1.6.0...v1.6.1)
+  - [capa v1.6.0...v1.6.1](https://github.com/mandiant/capa/compare/v1.6.0...v1.6.1)
+  - [capa-rules v1.6.0...v1.6.1](https://github.com/mandiant/capa-rules/compare/v1.6.0...v1.6.1)
 
 
 ## v1.6.0 (2021-03-09)
@@ -436,7 +832,7 @@ This release adds the capa explorer rule generator plugin for IDA Pro, vivisect
 
 ### Rule Generator IDA Plugin
 
-The capa explorer IDA plugin now helps you quickly build new capa rules using features extracted directly from your IDA database. Without leaving the plugin interface you can use the features extracted by capa explorer to develop and test new rules and save your work directly to your capa rules directory. To get started select the new `Rule Generator` tab, navigate to a function in the IDA `Disassembly` view, and click `Analyze`. For more information check out the capa explorer [readme](https://github.com/fireeye/capa/blob/master/capa/ida/plugin/README.md).
+The capa explorer IDA plugin now helps you quickly build new capa rules using features extracted directly from your IDA database. Without leaving the plugin interface you can use the features extracted by capa explorer to develop and test new rules and save your work directly to your capa rules directory. To get started select the new `Rule Generator` tab, navigate to a function in the IDA `Disassembly` view, and click `Analyze`. For more information check out the capa explorer [readme](https://github.com/mandiant/capa/blob/master/capa/ida/plugin/README.md).
 
 ![](doc/img/rulegen_expanded.png)
 
@@ -498,8 +894,8 @@ If you have workflows that rely on the Python 2 version and need future maintena
 
 ### Raw diffs
 
-  - [capa v1.5.1...v1.6.0](https://github.com/fireeye/capa/compare/v1.5.1...v1.6.0)
-  - [capa-rules v1.5.1...v1.6.0](https://github.com/fireeye/capa-rules/compare/v1.5.1...v1.6.0)
+  - [capa v1.5.1...v1.6.0](https://github.com/mandiant/capa/compare/v1.5.1...v1.6.0)
+  - [capa-rules v1.5.1...v1.6.0](https://github.com/mandiant/capa-rules/compare/v1.5.1...v1.6.0)
 
 
 ## v1.5.1 (2021-02-09)
@@ -512,8 +908,8 @@ This release fixes the version number that we forgot to update for v1.5.0 (there
 
 ### Raw diffs
 
-  - [capa v1.5.0...v1.5.1](https://github.com/fireeye/capa/compare/v1.5.1...v1.6.0)
-  - [capa-rules v1.5.0...v1.5.1](https://github.com/fireeye/capa-rules/compare/v1.5.1...v1.6.0)
+  - [capa v1.5.0...v1.5.1](https://github.com/mandiant/capa/compare/v1.5.1...v1.6.0)
+  - [capa-rules v1.5.0...v1.5.1](https://github.com/mandiant/capa-rules/compare/v1.5.1...v1.6.0)
 
 
 ## v1.5.0 (2021-02-05)
@@ -528,7 +924,7 @@ This release brings support for running capa under Python 3 via [SMDA](https://g
   
 @dzbeck also added [Malware Behavior Catalog](https://github.com/MBCProject/mbc-markdown) (MBC) and ATT&CK mappings for many rules.
 
-Download a standalone binary below and checkout the readme [here on GitHub](https://github.com/fireeye/capa/). Report issues on our [issue tracker](https://github.com/fireeye/capa/issues) and contribute new rules at [capa-rules](https://github.com/fireeye/capa-rules/).
+Download a standalone binary below and checkout the readme [here on GitHub](https://github.com/mandiant/capa/). Report issues on our [issue tracker](https://github.com/mandiant/capa/issues) and contribute new rules at [capa-rules](https://github.com/mandiant/capa-rules/).
 
 
 ### New Features
@@ -611,8 +1007,8 @@ Download a standalone binary below and checkout the readme [here on GitHub](http
   
 ### Raw diffs
 
-  - [capa v1.4.1...v1.5.0](https://github.com/fireeye/capa/compare/v1.4.1...v1.5.0)
-  - [capa-rules v1.4.0...v1.5.0](https://github.com/fireeye/capa-rules/compare/v1.4.0...v1.5.0)
+  - [capa v1.4.1...v1.5.0](https://github.com/mandiant/capa/compare/v1.4.1...v1.5.0)
+  - [capa-rules v1.4.0...v1.5.0](https://github.com/mandiant/capa-rules/compare/v1.4.0...v1.5.0)
 
 ## v1.4.1 (2020-10-23)
 
@@ -624,8 +1020,8 @@ This release fixes an issue building capa on our CI server, which prevented us f
   
 ### Raw diffs
 
-  - [capa v1.4.0...v1.4.1](https://github.com/fireeye/capa/compare/v1.4.0...v1.4.1)
-  - [capa-rules v1.4.0...v1.4.1](https://github.com/fireeye/capa-rules/compare/v1.4.0...v1.4.1)  
+  - [capa v1.4.0...v1.4.1](https://github.com/mandiant/capa/compare/v1.4.0...v1.4.1)
+  - [capa-rules v1.4.0...v1.4.1](https://github.com/mandiant/capa-rules/compare/v1.4.0...v1.4.1)  
 
 ## v1.4.0 (2020-10-23)
 
@@ -636,7 +1032,7 @@ This capa release includes changes to the rule parsing, enhanced feature extract
 
 @dzbeck added [Malware Behavior Catalog](https://github.com/MBCProject/mbc-markdown) (MBC) and ATT&CK mappings for 86 rules.
 
-Download a standalone binary below and checkout the readme [here on GitHub](https://github.com/fireeye/capa/). Report issues on our [issue tracker](https://github.com/fireeye/capa/issues) and contribute new rules at [capa-rules](https://github.com/fireeye/capa-rules/).
+Download a standalone binary below and checkout the readme [here on GitHub](https://github.com/mandiant/capa/). Report issues on our [issue tracker](https://github.com/mandiant/capa/issues) and contribute new rules at [capa-rules](https://github.com/mandiant/capa-rules/).
 
 ### New features
 
@@ -739,8 +1135,8 @@ Download a standalone binary below and checkout the readme [here on GitHub](http
 
 ### Raw diffs
 
-  - [capa v1.3.0...v1.4.0](https://github.com/fireeye/capa/compare/v1.3.0...v1.4.0)
-  - [capa-rules v1.3.0...v1.4.0](https://github.com/fireeye/capa-rules/compare/v1.3.0...v1.4.0)
+  - [capa v1.3.0...v1.4.0](https://github.com/mandiant/capa/compare/v1.3.0...v1.4.0)
+  - [capa-rules v1.3.0...v1.4.0](https://github.com/mandiant/capa-rules/compare/v1.3.0...v1.4.0)
 
 ## v1.3.0 (2020-09-14)
 
@@ -754,7 +1150,7 @@ This release brings newly updated mappings to the [Malware Behavior Catalog vers
   - @weslambert
   - @stevemk14ebr 
   
-Download a standalone binary below and checkout the readme [here on GitHub](https://github.com/fireeye/capa/). Report issues on our [issue tracker](https://github.com/fireeye/capa/issues) and contribute new rules at [capa-rules](https://github.com/fireeye/capa-rules/).
+Download a standalone binary below and checkout the readme [here on GitHub](https://github.com/mandiant/capa/). Report issues on our [issue tracker](https://github.com/mandiant/capa/issues) and contribute new rules at [capa-rules](https://github.com/mandiant/capa-rules/).
 
 ### Key changes to IDA Plugin
 
@@ -764,9 +1160,9 @@ The IDA Pro integration is now distributed as a real plugin, instead of a script
   - updates distributed PyPI/`pip install --upgrade` without touching your `%IDADIR%`
   - generally doing thing the "right way"
 
-How to get this new version? Its easy: download [capa_explorer.py](https://raw.githubusercontent.com/fireeye/capa/master/capa/ida/plugin/capa_explorer.py) to your IDA plugins directory and update your capa installation (incidentally, this is a good opportunity to migrate to `pip install flare-capa` instead of git checkouts). Now you should see the plugin listed in the `Edit > Plugins > FLARE capa explorer` menu in IDA. 
+How to get this new version? Its easy: download [capa_explorer.py](https://raw.githubusercontent.com/mandiant/capa/master/capa/ida/plugin/capa_explorer.py) to your IDA plugins directory and update your capa installation (incidentally, this is a good opportunity to migrate to `pip install flare-capa` instead of git checkouts). Now you should see the plugin listed in the `Edit > Plugins > FLARE capa explorer` menu in IDA. 
 
-Please refer to the plugin [readme](https://github.com/fireeye/capa/blob/master/capa/ida/plugin/README.md) for additional information on installing and using the IDA Pro plugin.
+Please refer to the plugin [readme](https://github.com/mandiant/capa/blob/master/capa/ida/plugin/README.md) for additional information on installing and using the IDA Pro plugin.
 
 Please open an issue in this repository if you notice anything weird.
  
@@ -811,8 +1207,8 @@ Please open an issue in this repository if you notice anything weird.
 
 ### Raw diffs
 
-  - [capa v1.2.0...v1.3.0](https://github.com/fireeye/capa/compare/v1.2.0...v1.3.0)
-  - [capa-rules v1.2.0...v1.3.0](https://github.com/fireeye/capa-rules/compare/v1.2.0...v1.3.0)
+  - [capa v1.2.0...v1.3.0](https://github.com/mandiant/capa/compare/v1.2.0...v1.3.0)
+  - [capa-rules v1.2.0...v1.3.0](https://github.com/mandiant/capa-rules/compare/v1.2.0...v1.3.0)
 
 ## v1.2.0 (2020-08-31)
 
@@ -828,9 +1224,9 @@ We received contributions from ten reverse engineers, including five new ones:
   - @edeca
   - @winniepe 
   
-Download a standalone binary below and checkout the readme [here on GitHub](https://github.com/fireeye/capa/).
-Report issues on our [issue tracker](https://github.com/fireeye/capa/issues)
-and contribute new rules at [capa-rules](https://github.com/fireeye/capa-rules/).
+Download a standalone binary below and checkout the readme [here on GitHub](https://github.com/mandiant/capa/).
+Report issues on our [issue tracker](https://github.com/mandiant/capa/issues)
+and contribute new rules at [capa-rules](https://github.com/mandiant/capa-rules/).
  
 ### New features
 
@@ -909,8 +1305,8 @@ and contribute new rules at [capa-rules](https://github.com/fireeye/capa-rules/)
 
 ### Raw diffs
 
-  - [capa v1.1.0...v1.2.0](https://github.com/fireeye/capa/compare/v1.1.0...v1.2.0)
-  - [capa-rules v1.1.0...v1.2.0](https://github.com/fireeye/capa-rules/compare/v1.1.0...v1.2.0)
+  - [capa v1.1.0...v1.2.0](https://github.com/mandiant/capa/compare/v1.1.0...v1.2.0)
+  - [capa-rules v1.1.0...v1.2.0](https://github.com/mandiant/capa-rules/compare/v1.1.0...v1.2.0)
 
 ## v1.1.0 (2020-08-05)
 
@@ -923,7 +1319,7 @@ We received contributions from eight reverse engineers, including four new ones:
   - @bitsofbinary
   - @threathive
   
-Download a standalone binary below and checkout the readme [here on GitHub](https://github.com/fireeye/capa/). Report issues on our [issue tracker](https://github.com/fireeye/capa/issues) and contribute new rules at [capa-rules](https://github.com/fireeye/capa-rules/).
+Download a standalone binary below and checkout the readme [here on GitHub](https://github.com/mandiant/capa/). Report issues on our [issue tracker](https://github.com/mandiant/capa/issues) and contribute new rules at [capa-rules](https://github.com/mandiant/capa-rules/).
   
 ### New features
 
@@ -989,12 +1385,12 @@ Download a standalone binary below and checkout the readme [here on GitHub](http
   - setup: pin vivisect version @williballenthin
   - setup: bump vivisect dependency version @williballenthin
   - setup: set Python project name to `flare-capa` @williballenthin
-  - ci: run tests and linter via Github Actions @Ana06
+  - ci: run tests and linter via GitHub Actions @Ana06
   - hooks: run style checkers and hide stashed output @Ana06
   - linter: ignore period in rule filename @williballenthin
   - linter: warn on nursery rule with no changes needed @williballenthin
 
 ### Raw diffs
 
-  - [capa v1.0.0...v1.1.0](https://github.com/fireeye/capa/compare/v1.0.0...v1.1.0)
-  - [capa-rules v1.0.0...v1.1.0](https://github.com/fireeye/capa-rules/compare/v1.0.0...v1.1.0)
+  - [capa v1.0.0...v1.1.0](https://github.com/mandiant/capa/compare/v1.0.0...v1.1.0)
+  - [capa-rules v1.0.0...v1.1.0](https://github.com/mandiant/capa-rules/compare/v1.0.0...v1.1.0)

LICENSE.txt

@@ -187,7 +187,7 @@
       same "printed page" as the copyright notice for easier
       identification within third-party archives.
 
-   Copyright (C) 2020 FireEye, Inc.
+   Copyright (C) 2020 Mandiant, Inc.
 
    Licensed under the Apache License, Version 2.0 (the "License");
    you may not use this file except in compliance with the License.

README.md

@@ -1,20 +1,21 @@
-![capa](https://github.com/fireeye/capa/blob/master/.github/logo.png)
+![capa](https://github.com/mandiant/capa/blob/master/.github/logo.png)
 
 [![PyPI - Python Version](https://img.shields.io/pypi/pyversions/flare-capa)](https://pypi.org/project/flare-capa)
-[![Last release](https://img.shields.io/github/v/release/fireeye/capa)](https://github.com/fireeye/capa/releases)
-[![Number of rules](https://img.shields.io/badge/rules-633-blue.svg)](https://github.com/fireeye/capa-rules)
-[![CI status](https://github.com/fireeye/capa/workflows/CI/badge.svg)](https://github.com/fireeye/capa/actions?query=workflow%3ACI+event%3Apush+branch%3Amaster)
-[![Downloads](https://img.shields.io/github/downloads/fireeye/capa/total)](https://github.com/fireeye/capa/releases)
+[![Last release](https://img.shields.io/github/v/release/mandiant/capa)](https://github.com/mandiant/capa/releases)
+[![Number of rules](https://img.shields.io/badge/rules-770-blue.svg)](https://github.com/mandiant/capa-rules)
+[![CI status](https://github.com/mandiant/capa/workflows/CI/badge.svg)](https://github.com/mandiant/capa/actions?query=workflow%3ACI+event%3Apush+branch%3Amaster)
+[![Downloads](https://img.shields.io/github/downloads/mandiant/capa/total)](https://github.com/mandiant/capa/releases)
 [![License](https://img.shields.io/badge/license-Apache--2.0-green.svg)](LICENSE.txt)
 
 capa detects capabilities in executable files.
-You run it against a PE, ELF, or shellcode file and it tells you what it thinks the program can do.
+You run it against a PE, ELF, .NET module, or shellcode file and it tells you what it thinks the program can do.
 For example, it might suggest that the file is a backdoor, is capable of installing services, or relies on HTTP to communicate.
 
 Check out:
-- the overview in our first [capa blog post](https://www.fireeye.com/blog/threat-research/2020/07/capa-automatically-identify-malware-capabilities.html)
-- the major version 2.0 updates described in our [second blog post](https://www.fireeye.com/blog/threat-research/2021/07/capa-2-better-stronger-faster.html)
-- the major version 3.0 (ELF support) described in the [third blog post](https://www.fireeye.com/blog/threat-research/2021/09/elfant-in-the-room-capa-v3.html)
+- the overview in our first [capa blog post](https://www.mandiant.com/resources/capa-automatically-identify-malware-capabilities)
+- the major version 2.0 updates described in our [second blog post](https://www.mandiant.com/resources/capa-2-better-stronger-faster)
+- the major version 3.0 (ELF support) described in the [third blog post](https://www.mandiant.com/resources/elfant-in-the-room-capa-v3)
+- the major version 4.0 (.NET support) described in the [fourth blog post](https://www.mandiant.com/resources/blog/capa-v4-casting-wider-net)
 
 ```
 $ capa.exe suspicious.exe
@@ -66,11 +67,11 @@ $ capa.exe suspicious.exe
 
 # download and usage
 
-Download stable releases of the standalone capa binaries [here](https://github.com/fireeye/capa/releases). You can run the standalone binaries without installation. capa is a command line tool that should be run from the terminal.
+Download stable releases of the standalone capa binaries [here](https://github.com/mandiant/capa/releases). You can run the standalone binaries without installation. capa is a command line tool that should be run from the terminal.
 
-To use capa as a library or integrate with another tool, see [doc/installation.md](https://github.com/fireeye/capa/blob/master/doc/installation.md) for further setup instructions.
+To use capa as a library or integrate with another tool, see [doc/installation.md](https://github.com/mandiant/capa/blob/master/doc/installation.md) for further setup instructions.
 
-For more information about how to use capa, see [doc/usage.md](https://github.com/fireeye/capa/blob/master/doc/usage.md).
+For more information about how to use capa, see [doc/usage.md](https://github.com/mandiant/capa/blob/master/doc/usage.md).
 
 # example
 
@@ -91,27 +92,36 @@ $ capa.exe suspicious.exe -vv
 ...
 execute shell command and capture output
 namespace   c2/shell
-author      matthew.williams@fireeye.com
+author      matthew.williams@mandiant.com
 scope       function
 att&ck      Execution::Command and Scripting Interpreter::Windows Command Shell [T1059.003]
 references  https://docs.microsoft.com/en-us/windows/win32/api/processthreadsapi/ns-processthreadsapi-startupinfoa
-examples    Practical Malware Analysis Lab 14-02.exe_:0x4011C0
-function @ 0x10003A13
+function @ 0x4011C0
   and:
-    match: create a process with modified I/O handles and window @ 0x10003A13
+    match: create a process with modified I/O handles and window @ 0x4011C0
       and:
+        number: 257 = STARTF_USESTDHANDLES | STARTF_USESHOWWINDOW @ 0x4012B8
         or:
-          api: kernel32.CreateProcess @ 0x10003D6D
-        number: 0x101 @ 0x10003B03
-        or:
-          number: 0x44 @ 0x10003ADC
-        optional:
-          api: kernel32.GetStartupInfo @ 0x10003AE4
-    match: create pipe @ 0x10003A13
+          number: 68 = StartupInfo.cb (size) @ 0x401282
+        or: = API functions that accept a pointer to a STARTUPINFO structure
+          api: kernel32.CreateProcess @ 0x401343
+    match: create pipe @ 0x4011C0
       or:
-        api: kernel32.CreatePipe @ 0x10003ACB
+        api: kernel32.CreatePipe @ 0x40126F, 0x401280
+    optional:
+      match: create thread @ 0x40136A, 0x4013BA
+        or:
+          and:
+            os: windows
+            or:
+              api: kernel32.CreateThread @ 0x4013D7
+        or:
+          and:
+            os: windows
+            or:
+              api: kernel32.CreateThread @ 0x401395
     or:
-      string: cmd.exe /c  @ 0x10003AED
+      string: "cmd.exe" @ 0x4012FD
 ...
 ```
 
@@ -127,39 +137,49 @@ rule:
   meta:
     name: hash data with CRC32
     namespace: data-manipulation/checksum/crc32
-    author: moritz.raabe@fireeye.com
+    authors:
+      - moritz.raabe@mandiant.com
     scope: function
+    mbc:
+      - Data::Checksum::CRC32 [C0032.001]
     examples:
       - 2D3EDC218A90F03089CC01715A9F047F:0x403CBD
       - 7D28CB106CB54876B2A5C111724A07CD:0x402350  # RtlComputeCrc32
+      - 7EFF498DE13CC734262F87E6B3EF38AB:0x100084A6
   features:
     - or:
       - and:
         - mnemonic: shr
-        - number: 0xEDB88320
+        - or:
+          - number: 0xEDB88320
+          - bytes: 00 00 00 00 96 30 07 77 2C 61 0E EE BA 51 09 99 19 C4 6D 07 8F F4 6A 70 35 A5 63 E9 A3 95 64 9E = crc32_tab
         - number: 8
         - characteristic: nzxor
+      - and:
+        - number: 0x8320
+        - number: 0xEDB8
+        - characteristic: nzxor
       - api: RtlComputeCrc32
 ```
 
-The [github.com/fireeye/capa-rules](https://github.com/fireeye/capa-rules) repository contains hundreds of standard library rules that are distributed with capa.
+The [github.com/mandiant/capa-rules](https://github.com/mandiant/capa-rules) repository contains hundreds of standard library rules that are distributed with capa.
 Please learn to write rules and contribute new entries as you find interesting techniques in malware.
 
-If you use IDA Pro, then you can use the [capa explorer](https://github.com/fireeye/capa/tree/master/capa/ida/plugin) plugin.
+If you use IDA Pro, then you can use the [capa explorer](https://github.com/mandiant/capa/tree/master/capa/ida/plugin) plugin.
 capa explorer helps you identify interesting areas of a program and build new capa rules using features extracted directly from your IDA Pro database.
 
-![capa + IDA Pro integration](https://github.com/fireeye/capa/blob/master/doc/img/explorer_expanded.png)
+![capa + IDA Pro integration](https://github.com/mandiant/capa/blob/master/doc/img/explorer_expanded.png)
 
 # further information
 ## capa
-- [Installation](https://github.com/fireeye/capa/blob/master/doc/installation.md)
-- [Usage](https://github.com/fireeye/capa/blob/master/doc/usage.md)
-- [Limitations](https://github.com/fireeye/capa/blob/master/doc/limitations.md)
-- [Contributing Guide](https://github.com/fireeye/capa/blob/master/.github/CONTRIBUTING.md)
+- [Installation](https://github.com/mandiant/capa/blob/master/doc/installation.md)
+- [Usage](https://github.com/mandiant/capa/blob/master/doc/usage.md)
+- [Limitations](https://github.com/mandiant/capa/blob/master/doc/limitations.md)
+- [Contributing Guide](https://github.com/mandiant/capa/blob/master/.github/CONTRIBUTING.md)
 
 ## capa rules
-- [capa-rules repository](https://github.com/fireeye/capa-rules)
-- [capa-rules rule format](https://github.com/fireeye/capa-rules/blob/master/doc/format.md)
+- [capa-rules repository](https://github.com/mandiant/capa-rules)
+- [capa-rules rule format](https://github.com/mandiant/capa-rules/blob/master/doc/format.md)
 
 ## capa testfiles
-The [capa-testfiles repository](https://github.com/fireeye/capa-testfiles) contains the data we use to test capa's code and rules
+The [capa-testfiles repository](https://github.com/mandiant/capa-testfiles) contains the data we use to test capa's code and rules

capa/engine.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -8,11 +8,17 @@
 
 import copy
 import collections
-from typing import Set, Dict, List, Tuple, Union, Mapping, Iterable
+from typing import TYPE_CHECKING, Set, Dict, List, Tuple, Union, Mapping, Iterable, Iterator, cast
 
-import capa.rules
+import capa.perf
 import capa.features.common
-from capa.features.common import Feature
+from capa.features.common import Result, Feature
+from capa.features.address import Address
+
+if TYPE_CHECKING:
+    # circular import, otherwise
+    import capa.rules
+
 
 # a collection of features and the locations at which they are found.
 #
@@ -21,7 +27,7 @@ from capa.features.common import Feature
 # to collect the locations of a feature, do: `features[Number(0x10)]`
 #
 # aliased here so that the type can be documented and xref'd.
-FeatureSet = Dict[Feature, Set[int]]
+FeatureSet = Dict[Feature, Set[Address]]
 
 
 class Statement:
@@ -32,7 +38,7 @@ class Statement:
     """
 
     def __init__(self, description=None):
-        super(Statement, self).__init__()
+        super().__init__()
         self.name = self.__class__.__name__
         self.description = description
 
@@ -45,29 +51,33 @@ class Statement:
     def __repr__(self):
         return str(self)
 
-    def evaluate(self, features: FeatureSet) -> "Result":
+    def evaluate(self, features: FeatureSet, short_circuit=True) -> Result:
         """
         classes that inherit `Statement` must implement `evaluate`
 
         args:
-          ctx (defaultdict[Feature, set[VA]])
-
-        returns:
-          Result
+            short_circuit (bool): if true, then statements like and/or/some may short circuit.
         """
         raise NotImplementedError()
 
-    def get_children(self):
+    def get_children(self) -> Iterator[Union["Statement", Feature]]:
         if hasattr(self, "child"):
-            yield self.child
+            # this really confuses mypy because the property may not exist
+            # since its defined in the subclasses.
+            child = self.child  # type: ignore
+            assert isinstance(child, (Statement, Feature))
+            yield child
 
         if hasattr(self, "children"):
             for child in getattr(self, "children"):
+                assert isinstance(child, (Statement, Feature))
                 yield child
 
     def replace_child(self, existing, new):
         if hasattr(self, "child"):
-            if self.child is existing:
+            # this really confuses mypy because the property may not exist
+            # since its defined in the subclasses.
+            if self.child is existing:  # type: ignore
                 self.child = new
 
         if hasattr(self, "children"):
@@ -77,113 +87,143 @@ class Statement:
                     children[i] = new
 
 
-class Result:
-    """
-    represents the results of an evaluation of statements against features.
-
-    instances of this class should behave like a bool,
-    e.g. `assert Result(True, ...) == True`
-
-    instances track additional metadata about evaluation results.
-    they contain references to the statement node (e.g. an And statement),
-     as well as the children Result instances.
-
-    we need this so that we can render the tree of expressions and their results.
-    """
-
-    def __init__(self, success: bool, statement: Union[Statement, Feature], children: List["Result"], locations=None):
-        """
-        args:
-          success (bool)
-          statement (capa.engine.Statement or capa.features.Feature)
-          children (list[Result])
-          locations (iterable[VA])
-        """
-        super(Result, self).__init__()
-        self.success = success
-        self.statement = statement
-        self.children = children
-        self.locations = locations if locations is not None else ()
-
-    def __eq__(self, other):
-        if isinstance(other, bool):
-            return self.success == other
-        return False
-
-    def __bool__(self):
-        return self.success
-
-    def __nonzero__(self):
-        return self.success
-
-
 class And(Statement):
-    """match if all of the children evaluate to True."""
+    """
+    match if all of the children evaluate to True.
+
+    the order of evaluation is dictated by the property
+    `And.children` (type: List[Statement|Feature]).
+    a query optimizer may safely manipulate the order of these children.
+    """
 
     def __init__(self, children, description=None):
-        super(And, self).__init__(description=description)
+        super().__init__(description=description)
         self.children = children
 
-    def evaluate(self, ctx):
-        results = [child.evaluate(ctx) for child in self.children]
-        success = all(results)
-        return Result(success, self, results)
+    def evaluate(self, ctx, short_circuit=True):
+        capa.perf.counters["evaluate.feature"] += 1
+        capa.perf.counters["evaluate.feature.and"] += 1
+
+        if short_circuit:
+            results = []
+            for child in self.children:
+                result = child.evaluate(ctx, short_circuit=short_circuit)
+                results.append(result)
+                if not result:
+                    # short circuit
+                    return Result(False, self, results)
+
+            return Result(True, self, results)
+        else:
+            results = [child.evaluate(ctx, short_circuit=short_circuit) for child in self.children]
+            success = all(results)
+            return Result(success, self, results)
 
 
 class Or(Statement):
-    """match if any of the children evaluate to True."""
+    """
+    match if any of the children evaluate to True.
+
+    the order of evaluation is dictated by the property
+    `Or.children` (type: List[Statement|Feature]).
+    a query optimizer may safely manipulate the order of these children.
+    """
 
     def __init__(self, children, description=None):
-        super(Or, self).__init__(description=description)
+        super().__init__(description=description)
         self.children = children
 
-    def evaluate(self, ctx):
-        results = [child.evaluate(ctx) for child in self.children]
-        success = any(results)
-        return Result(success, self, results)
+    def evaluate(self, ctx, short_circuit=True):
+        capa.perf.counters["evaluate.feature"] += 1
+        capa.perf.counters["evaluate.feature.or"] += 1
+
+        if short_circuit:
+            results = []
+            for child in self.children:
+                result = child.evaluate(ctx, short_circuit=short_circuit)
+                results.append(result)
+                if result:
+                    # short circuit as soon as we hit one match
+                    return Result(True, self, results)
+
+            return Result(False, self, results)
+        else:
+            results = [child.evaluate(ctx, short_circuit=short_circuit) for child in self.children]
+            success = any(results)
+            return Result(success, self, results)
 
 
 class Not(Statement):
     """match only if the child evaluates to False."""
 
     def __init__(self, child, description=None):
-        super(Not, self).__init__(description=description)
+        super().__init__(description=description)
         self.child = child
 
-    def evaluate(self, ctx):
-        results = [self.child.evaluate(ctx)]
+    def evaluate(self, ctx, short_circuit=True):
+        capa.perf.counters["evaluate.feature"] += 1
+        capa.perf.counters["evaluate.feature.not"] += 1
+
+        results = [self.child.evaluate(ctx, short_circuit=short_circuit)]
         success = not results[0]
         return Result(success, self, results)
 
 
 class Some(Statement):
-    """match if at least N of the children evaluate to True."""
+    """
+    match if at least N of the children evaluate to True.
+
+    the order of evaluation is dictated by the property
+    `Some.children` (type: List[Statement|Feature]).
+    a query optimizer may safely manipulate the order of these children.
+    """
 
     def __init__(self, count, children, description=None):
-        super(Some, self).__init__(description=description)
+        super().__init__(description=description)
         self.count = count
         self.children = children
 
-    def evaluate(self, ctx):
-        results = [child.evaluate(ctx) for child in self.children]
-        # note that here we cast the child result as a bool
-        # because we've overridden `__bool__` above.
-        #
-        # we can't use `if child is True` because the instance is not True.
-        success = sum([1 for child in results if bool(child) is True]) >= self.count
-        return Result(success, self, results)
+    def evaluate(self, ctx, short_circuit=True):
+        capa.perf.counters["evaluate.feature"] += 1
+        capa.perf.counters["evaluate.feature.some"] += 1
+
+        if short_circuit:
+            results = []
+            satisfied_children_count = 0
+            for child in self.children:
+                result = child.evaluate(ctx, short_circuit=short_circuit)
+                results.append(result)
+                if result:
+                    satisfied_children_count += 1
+
+                if satisfied_children_count >= self.count:
+                    # short circuit as soon as we hit the threshold
+                    return Result(True, self, results)
+
+            return Result(False, self, results)
+        else:
+            results = [child.evaluate(ctx, short_circuit=short_circuit) for child in self.children]
+            # note that here we cast the child result as a bool
+            # because we've overridden `__bool__` above.
+            #
+            # we can't use `if child is True` because the instance is not True.
+            success = sum([1 for child in results if bool(child) is True]) >= self.count
+            return Result(success, self, results)
 
 
 class Range(Statement):
     """match if the child is contained in the ctx set with a count in the given range."""
 
     def __init__(self, child, min=None, max=None, description=None):
-        super(Range, self).__init__(description=description)
+        super().__init__(description=description)
         self.child = child
         self.min = min if min is not None else 0
         self.max = max if max is not None else (1 << 64 - 1)
 
-    def evaluate(self, ctx):
+    def evaluate(self, ctx, **kwargs):
+        capa.perf.counters["evaluate.feature"] += 1
+        capa.perf.counters["evaluate.feature.range"] += 1
+
         count = len(ctx.get(self.child, []))
         if self.min == 0 and count == 0:
             return Result(True, self, [])
@@ -203,12 +243,12 @@ class Subscope(Statement):
     the engine should preprocess rules to extract subscope statements into their own rules.
     """
 
-    def __init__(self, scope, child):
-        super(Subscope, self).__init__()
+    def __init__(self, scope, child, description=None):
+        super().__init__(description=description)
         self.scope = scope
         self.child = child
 
-    def evaluate(self, ctx):
+    def evaluate(self, ctx, **kwargs):
         raise ValueError("cannot evaluate a subscope directly!")
 
 
@@ -225,10 +265,10 @@ class Subscope(Statement):
 #         inspect(match_details)
 #
 # aliased here so that the type can be documented and xref'd.
-MatchResults = Mapping[str, List[Tuple[int, Result]]]
+MatchResults = Mapping[str, List[Tuple[Address, Result]]]
 
 
-def index_rule_matches(features: FeatureSet, rule: "capa.rules.Rule", locations: Iterable[int]):
+def index_rule_matches(features: FeatureSet, rule: "capa.rules.Rule", locations: Iterable[Address]):
     """
     record into the given featureset that the given rule matched at the given locations.
 
@@ -245,17 +285,22 @@ def index_rule_matches(features: FeatureSet, rule: "capa.rules.Rule", locations:
             namespace, _, _ = namespace.rpartition("/")
 
 
-def match(rules: List["capa.rules.Rule"], features: FeatureSet, va: int) -> Tuple[FeatureSet, MatchResults]:
+def match(rules: List["capa.rules.Rule"], features: FeatureSet, addr: Address) -> Tuple[FeatureSet, MatchResults]:
     """
-    Args:
-      rules (List[capa.rules.Rule]): these must already be ordered topologically by dependency.
-      features (Mapping[capa.features.Feature, int]):
-      va (int): location of the features
+    match the given rules against the given features,
+    returning an updated set of features and the matches.
 
-    Returns:
-      Tuple[FeatureSet, MatchResults]: two-tuple with entries:
-        - set of features used for matching (which may be a superset of the given `features` argument, due to rule match features), and
-        - mapping from rule name to [(location of match, result object)]
+    the updated features are just like the input,
+    but extended to include the match features (e.g. names of rules that matched).
+    the given feature set is not modified; an updated copy is returned.
+
+    the given list of rules must be ordered topologically by dependency,
+    or else `match` statements will not be handled correctly.
+
+    this routine should be fairly optimized, but is not guaranteed to be the fastest matcher possible.
+    it has a particularly convenient signature: (rules, features) -> matches
+    other strategies can be imagined that match differently; implement these elsewhere.
+    specifically, this routine does "top down" matching of the given rules against the feature set.
     """
     results = collections.defaultdict(list)  # type: MatchResults
 
@@ -266,12 +311,22 @@ def match(rules: List["capa.rules.Rule"], features: FeatureSet, va: int) -> Tupl
     features = collections.defaultdict(set, copy.copy(features))
 
     for rule in rules:
-        res = rule.evaluate(features)
+        res = rule.evaluate(features, short_circuit=True)
         if res:
-            results[rule.name].append((va, res))
+            # we first matched the rule with short circuiting enabled.
+            # this is much faster than without short circuiting.
+            # however, we want to collect all results thoroughly,
+            # so once we've found a match quickly,
+            # go back and capture results without short circuiting.
+            res = rule.evaluate(features, short_circuit=False)
+
+            # sanity check
+            assert bool(res) is True
+
+            results[rule.name].append((addr, res))
             # we need to update the current `features`
             # because subsequent iterations of this loop may use newly added features,
             # such as rule or namespace matches.
-            index_rule_matches(features, rule, [va])
+            index_rule_matches(features, rule, [addr])
 
     return (features, results)

capa/exceptions.py

@@ -0,0 +1,14 @@
+class UnsupportedRuntimeError(RuntimeError):
+    pass
+
+
+class UnsupportedFormatError(ValueError):
+    pass
+
+
+class UnsupportedArchError(ValueError):
+    pass
+
+
+class UnsupportedOSError(ValueError):
+    pass

capa/features/address.py

@@ -0,0 +1,114 @@
+import abc
+
+
+class Address(abc.ABC):
+    @abc.abstractmethod
+    def __eq__(self, other):
+        ...
+
+    @abc.abstractmethod
+    def __lt__(self, other):
+        # implement < so that addresses can be sorted from low to high
+        ...
+
+    @abc.abstractmethod
+    def __hash__(self):
+        # implement hash so that addresses can be used in sets and dicts
+        ...
+
+    @abc.abstractmethod
+    def __repr__(self):
+        # implement repr to help during debugging
+        ...
+
+
+class AbsoluteVirtualAddress(int, Address):
+    """an absolute memory address"""
+
+    def __new__(cls, v):
+        assert v >= 0
+        return int.__new__(cls, v)
+
+    def __repr__(self):
+        return f"absolute(0x{self:x})"
+
+    def __hash__(self):
+        return int.__hash__(self)
+
+
+class RelativeVirtualAddress(int, Address):
+    """a memory address relative to a base address"""
+
+    def __repr__(self):
+        return f"relative(0x{self:x})"
+
+    def __hash__(self):
+        return int.__hash__(self)
+
+
+class FileOffsetAddress(int, Address):
+    """an address relative to the start of a file"""
+
+    def __new__(cls, v):
+        assert v >= 0
+        return int.__new__(cls, v)
+
+    def __repr__(self):
+        return f"file(0x{self:x})"
+
+    def __hash__(self):
+        return int.__hash__(self)
+
+
+class DNTokenAddress(int, Address):
+    """a .NET token"""
+
+    def __new__(cls, token: int):
+        return int.__new__(cls, token)
+
+    def __repr__(self):
+        return f"token(0x{self:x})"
+
+    def __hash__(self):
+        return int.__hash__(self)
+
+
+class DNTokenOffsetAddress(Address):
+    """an offset into an object specified by a .NET token"""
+
+    def __init__(self, token: int, offset: int):
+        assert offset >= 0
+        self.token = token
+        self.offset = offset
+
+    def __eq__(self, other):
+        return (self.token, self.offset) == (other.token, other.offset)
+
+    def __lt__(self, other):
+        return (self.token, self.offset) < (other.token, other.offset)
+
+    def __hash__(self):
+        return hash((self.token, self.offset))
+
+    def __repr__(self):
+        return f"token(0x{self.token:x})+(0x{self.offset:x})"
+
+    def __index__(self):
+        return self.token + self.offset
+
+
+class _NoAddress(Address):
+    def __eq__(self, other):
+        return True
+
+    def __lt__(self, other):
+        return False
+
+    def __hash__(self):
+        return hash(0)
+
+    def __repr__(self):
+        return "no address"
+
+
+NO_ADDRESS = _NoAddress()

capa/features/basicblock.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -10,18 +10,11 @@ from capa.features.common import Feature
 
 
 class BasicBlock(Feature):
-    def __init__(self):
-        super(BasicBlock, self).__init__(None)
+    def __init__(self, description=None):
+        super().__init__(0, description=description)
 
     def __str__(self):
         return "basic block"
 
     def get_value_str(self):
         return ""
-
-    def freeze_serialize(self):
-        return (self.__class__.__name__, [])
-
-    @classmethod
-    def freeze_deserialize(cls, args):
-        return cls()

capa/features/common.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -7,14 +7,21 @@
 # See the License for the specific language governing permissions and limitations under the License.
 
 import re
+import abc
 import codecs
+import typing
 import logging
 import collections
-from typing import Set, Dict, Union
+from typing import TYPE_CHECKING, Set, Dict, List, Union, Optional
 
-import capa.engine
+if TYPE_CHECKING:
+    # circular import, otherwise
+    import capa.engine
+
+import capa.perf
 import capa.features
 import capa.features.extractors.elf
+from capa.features.address import Address
 
 logger = logging.getLogger(__name__)
 MAX_BYTES_FEATURE_SIZE = 0x100
@@ -23,6 +30,14 @@ MAX_BYTES_FEATURE_SIZE = 0x100
 THUNK_CHAIN_DEPTH_DELTA = 5
 
 
+class FeatureAccess:
+    READ = "read"
+    WRITE = "write"
+
+
+VALID_FEATURE_ACCESS = (FeatureAccess.READ, FeatureAccess.WRITE)
+
+
 def bytes_to_str(b: bytes) -> str:
     return str(codecs.encode(b, "hex").decode("utf-8"))
 
@@ -46,105 +61,153 @@ def escape_string(s: str) -> str:
     return s
 
 
-class Feature:
-    def __init__(self, value: Union[str, int, bytes], bitness=None, description=None):
+class Result:
+    """
+    represents the results of an evaluation of statements against features.
+
+    instances of this class should behave like a bool,
+    e.g. `assert Result(True, ...) == True`
+
+    instances track additional metadata about evaluation results.
+    they contain references to the statement node (e.g. an And statement),
+     as well as the children Result instances.
+
+    we need this so that we can render the tree of expressions and their results.
+    """
+
+    def __init__(
+        self,
+        success: bool,
+        statement: Union["capa.engine.Statement", "Feature"],
+        children: List["Result"],
+        locations: Optional[Set[Address]] = None,
+    ):
+        super().__init__()
+        self.success = success
+        self.statement = statement
+        self.children = children
+        self.locations = locations if locations is not None else set()
+
+    def __eq__(self, other):
+        if isinstance(other, bool):
+            return self.success == other
+        return False
+
+    def __bool__(self):
+        return self.success
+
+    def __nonzero__(self):
+        return self.success
+
+
+class Feature(abc.ABC):
+    def __init__(
+        self,
+        value: Union[str, int, float, bytes],
+        description: Optional[str] = None,
+    ):
         """
         Args:
           value (any): the value of the feature, such as the number or string.
-          bitness (str): one of the VALID_BITNESS values, or None.
-            When None, then the feature applies to any bitness.
-            Modifies the feature name from `feature` to `feature/bitness`, like `offset/x32`.
           description (str): a human-readable description that explains the feature value.
         """
-        super(Feature, self).__init__()
-
-        if bitness is not None:
-            if bitness not in VALID_BITNESS:
-                raise ValueError("bitness '%s' must be one of %s" % (bitness, VALID_BITNESS))
-            self.name = self.__class__.__name__.lower() + "/" + bitness
-        else:
-            self.name = self.__class__.__name__.lower()
+        super().__init__()
 
+        self.name = self.__class__.__name__.lower()
         self.value = value
-        self.bitness = bitness
         self.description = description
 
     def __hash__(self):
-        return hash((self.name, self.value, self.bitness))
+        return hash((self.name, self.value))
 
     def __eq__(self, other):
-        return self.name == other.name and self.value == other.value and self.bitness == other.bitness
+        return self.name == other.name and self.value == other.value
+
+    def __lt__(self, other):
+        # TODO: this is a huge hack!
+        import capa.features.freeze.features
+
+        return (
+            capa.features.freeze.features.feature_from_capa(self).json()
+            < capa.features.freeze.features.feature_from_capa(other).json()
+        )
+
+    def get_name_str(self) -> str:
+        """
+        render the name of this feature, for use by `__str__` and friends.
+        subclasses should override to customize the rendering.
+        """
+        return self.name
 
     def get_value_str(self) -> str:
         """
         render the value of this feature, for use by `__str__` and friends.
         subclasses should override to customize the rendering.
-
-        Returns: any
         """
         return str(self.value)
 
     def __str__(self):
         if self.value is not None:
             if self.description:
-                return "%s(%s = %s)" % (self.name, self.get_value_str(), self.description)
+                return "%s(%s = %s)" % (self.get_name_str(), self.get_value_str(), self.description)
             else:
-                return "%s(%s)" % (self.name, self.get_value_str())
+                return "%s(%s)" % (self.get_name_str(), self.get_value_str())
         else:
-            return "%s" % self.name
+            return "%s" % self.get_name_str()
 
     def __repr__(self):
         return str(self)
 
-    def evaluate(self, ctx: Dict["Feature", Set[int]]) -> "capa.engine.Result":
-        return capa.engine.Result(self in ctx, self, [], locations=ctx.get(self, []))
-
-    def freeze_serialize(self):
-        if self.bitness is not None:
-            return (self.__class__.__name__, [self.value, {"bitness": self.bitness}])
-        else:
-            return (self.__class__.__name__, [self.value])
-
-    @classmethod
-    def freeze_deserialize(cls, args):
-        # as you can see below in code,
-        # if the last argument is a dictionary,
-        # consider it to be kwargs passed to the feature constructor.
-        if len(args) == 1:
-            return cls(*args)
-        elif isinstance(args[-1], dict):
-            kwargs = args[-1]
-            args = args[:-1]
-            return cls(*args, **kwargs)
+    def evaluate(self, ctx: Dict["Feature", Set[Address]], **kwargs) -> Result:
+        capa.perf.counters["evaluate.feature"] += 1
+        capa.perf.counters["evaluate.feature." + self.name] += 1
+        return Result(self in ctx, self, [], locations=ctx.get(self, set()))
 
 
 class MatchedRule(Feature):
     def __init__(self, value: str, description=None):
-        super(MatchedRule, self).__init__(value, description=description)
+        super().__init__(value, description=description)
         self.name = "match"
 
 
 class Characteristic(Feature):
     def __init__(self, value: str, description=None):
-
-        super(Characteristic, self).__init__(value, description=description)
+        super().__init__(value, description=description)
 
 
 class String(Feature):
     def __init__(self, value: str, description=None):
-        super(String, self).__init__(value, description=description)
+        super().__init__(value, description=description)
+
+    def get_value_str(self) -> str:
+        assert isinstance(self.value, str)
+        return escape_string(self.value)
+
+
+class Class(Feature):
+    def __init__(self, value: str, description=None):
+        super().__init__(value, description=description)
+
+
+class Namespace(Feature):
+    def __init__(self, value: str, description=None):
+        super().__init__(value, description=description)
 
 
 class Substring(String):
     def __init__(self, value: str, description=None):
-        super(Substring, self).__init__(value, description=description)
+        super().__init__(value, description=description)
         self.value = value
 
-    def evaluate(self, ctx):
+    def evaluate(self, ctx, short_circuit=True):
+        capa.perf.counters["evaluate.feature"] += 1
+        capa.perf.counters["evaluate.feature.substring"] += 1
+
         # mapping from string value to list of locations.
         # will unique the locations later on.
-        matches = collections.defaultdict(list)
+        matches: typing.DefaultDict[str, Set[Address]] = collections.defaultdict(set)
 
+        assert isinstance(self.value, str)
         for feature, locations in ctx.items():
             if not isinstance(feature, (String,)):
                 continue
@@ -154,28 +217,32 @@ class Substring(String):
                 raise ValueError("unexpected feature value type")
 
             if self.value in feature.value:
-                matches[feature.value].extend(locations)
+                matches[feature.value].update(locations)
+                if short_circuit:
+                    # we found one matching string, thats sufficient to match.
+                    # don't collect other matching strings in this mode.
+                    break
 
         if matches:
-            # finalize: defaultdict -> dict
-            # which makes json serialization easier
-            matches = dict(matches)
-
             # collect all locations
             locations = set()
-            for s in matches.keys():
-                matches[s] = list(set(matches[s]))
-                locations.update(matches[s])
+            for locs in matches.values():
+                locations.update(locs)
 
             # unlike other features, we cannot return put a reference to `self` directly in a `Result`.
             # this is because `self` may match on many strings, so we can't stuff the matched value into it.
             # instead, return a new instance that has a reference to both the substring and the matched values.
-            return capa.engine.Result(True, _MatchedSubstring(self, matches), [], locations=locations)
+            return Result(True, _MatchedSubstring(self, dict(matches)), [], locations=locations)
         else:
-            return capa.engine.Result(False, _MatchedSubstring(self, None), [])
+            return Result(False, _MatchedSubstring(self, {}), [])
+
+    def get_value_str(self) -> str:
+        assert isinstance(self.value, str)
+        return escape_string(self.value)
 
     def __str__(self):
-        return "substring(%s)" % self.value
+        assert isinstance(self.value, str)
+        return "substring(%s)" % escape_string(self.value)
 
 
 class _MatchedSubstring(Substring):
@@ -186,13 +253,13 @@ class _MatchedSubstring(Substring):
     note: this type should only ever be constructed by `Substring.evaluate()`. it is not part of the public API.
     """
 
-    def __init__(self, substring: Substring, matches):
+    def __init__(self, substring: Substring, matches: Dict[str, Set[Address]]):
         """
         args:
-          substring (Substring): the substring feature that matches.
-          match (Dict[string, List[int]]|None): mapping from matching string to its locations.
+          substring: the substring feature that matches.
+          match: mapping from matching string to its locations.
         """
-        super(_MatchedSubstring, self).__init__(str(substring.value), description=substring.description)
+        super().__init__(str(substring.value), description=substring.description)
         # we want this to collide with the name of `Substring` above,
         # so that it works nicely with the renderers.
         self.name = "substring"
@@ -200,6 +267,7 @@ class _MatchedSubstring(Substring):
         self.matches = matches
 
     def __str__(self):
+        assert isinstance(self.value, str)
         return 'substring("%s", matches = %s)' % (
             self.value,
             ", ".join(map(lambda s: '"' + s + '"', (self.matches or {}).keys())),
@@ -208,7 +276,7 @@ class _MatchedSubstring(Substring):
 
 class Regex(String):
     def __init__(self, value: str, description=None):
-        super(Regex, self).__init__(value, description=description)
+        super().__init__(value, description=description)
         self.value = value
 
         pat = self.value[len("/") : -len("/")]
@@ -218,17 +286,20 @@ class Regex(String):
             flags |= re.IGNORECASE
         try:
             self.re = re.compile(pat, flags)
-        except re.error:
+        except re.error as exc:
             if value.endswith("/i"):
                 value = value[: -len("i")]
             raise ValueError(
                 "invalid regular expression: %s it should use Python syntax, try it at https://pythex.org" % value
-            )
+            ) from exc
+
+    def evaluate(self, ctx, short_circuit=True):
+        capa.perf.counters["evaluate.feature"] += 1
+        capa.perf.counters["evaluate.feature.regex"] += 1
 
-    def evaluate(self, ctx):
         # mapping from string value to list of locations.
         # will unique the locations later on.
-        matches = collections.defaultdict(list)
+        matches: typing.DefaultDict[str, Set[Address]] = collections.defaultdict(set)
 
         for feature, locations in ctx.items():
             if not isinstance(feature, (String,)):
@@ -243,28 +314,28 @@ class Regex(String):
             # using this mode cleans is more convenient for rule authors,
             # so that they don't have to prefix/suffix their terms like: /.*foo.*/.
             if self.re.search(feature.value):
-                matches[feature.value].extend(locations)
+                matches[feature.value].update(locations)
+                if short_circuit:
+                    # we found one matching string, thats sufficient to match.
+                    # don't collect other matching strings in this mode.
+                    break
 
         if matches:
-            # finalize: defaultdict -> dict
-            # which makes json serialization easier
-            matches = dict(matches)
-
             # collect all locations
             locations = set()
-            for s in matches.keys():
-                matches[s] = list(set(matches[s]))
-                locations.update(matches[s])
+            for locs in matches.values():
+                locations.update(locs)
 
             # unlike other features, we cannot return put a reference to `self` directly in a `Result`.
             # this is because `self` may match on many strings, so we can't stuff the matched value into it.
             # instead, return a new instance that has a reference to both the regex and the matched values.
             # see #262.
-            return capa.engine.Result(True, _MatchedRegex(self, matches), [], locations=locations)
+            return Result(True, _MatchedRegex(self, dict(matches)), [], locations=locations)
         else:
-            return capa.engine.Result(False, _MatchedRegex(self, None), [])
+            return Result(False, _MatchedRegex(self, {}), [])
 
     def __str__(self):
+        assert isinstance(self.value, str)
         return "regex(string =~ %s)" % self.value
 
 
@@ -276,13 +347,13 @@ class _MatchedRegex(Regex):
     note: this type should only ever be constructed by `Regex.evaluate()`. it is not part of the public API.
     """
 
-    def __init__(self, regex: Regex, matches):
+    def __init__(self, regex: Regex, matches: Dict[str, Set[Address]]):
         """
         args:
-          regex (Regex): the regex feature that matches.
-          match (Dict[string, List[int]]|None): mapping from matching string to its locations.
+          regex: the regex feature that matches.
+          matches: mapping from matching string to its locations.
         """
-        super(_MatchedRegex, self).__init__(str(regex.value), description=regex.description)
+        super().__init__(str(regex.value), description=regex.description)
         # we want this to collide with the name of `Regex` above,
         # so that it works nicely with the renderers.
         self.name = "regex"
@@ -290,6 +361,7 @@ class _MatchedRegex(Regex):
         self.matches = matches
 
     def __str__(self):
+        assert isinstance(self.value, str)
         return "regex(string =~ %s, matches = %s)" % (
             self.value,
             ", ".join(map(lambda s: '"' + s + '"', (self.matches or {}).keys())),
@@ -305,70 +377,73 @@ class StringFactory:
 
 class Bytes(Feature):
     def __init__(self, value: bytes, description=None):
-        super(Bytes, self).__init__(value, description=description)
+        super().__init__(value, description=description)
         self.value = value
 
-    def evaluate(self, ctx):
+    def evaluate(self, ctx, **kwargs):
+        capa.perf.counters["evaluate.feature"] += 1
+        capa.perf.counters["evaluate.feature.bytes"] += 1
+
+        assert isinstance(self.value, bytes)
         for feature, locations in ctx.items():
             if not isinstance(feature, (Bytes,)):
                 continue
 
+            assert isinstance(feature.value, bytes)
             if feature.value.startswith(self.value):
-                return capa.engine.Result(True, self, [], locations=locations)
+                return Result(True, self, [], locations=locations)
 
-        return capa.engine.Result(False, self, [])
+        return Result(False, self, [])
 
     def get_value_str(self):
+        assert isinstance(self.value, bytes)
         return hex_string(bytes_to_str(self.value))
 
-    def freeze_serialize(self):
-        return (self.__class__.__name__, [bytes_to_str(self.value).upper()])
-
-    @classmethod
-    def freeze_deserialize(cls, args):
-        return cls(*[codecs.decode(x, "hex") for x in args])
-
-
-# identifiers for supported bitness names that tweak a feature
-# for example, offset/x32
-BITNESS_X32 = "x32"
-BITNESS_X64 = "x64"
-VALID_BITNESS = (BITNESS_X32, BITNESS_X64)
-
 
 # other candidates here: https://docs.microsoft.com/en-us/windows/win32/debug/pe-format#machine-types
 ARCH_I386 = "i386"
 ARCH_AMD64 = "amd64"
-VALID_ARCH = (ARCH_I386, ARCH_AMD64)
+# dotnet
+ARCH_ANY = "any"
+VALID_ARCH = (ARCH_I386, ARCH_AMD64, ARCH_ANY)
 
 
 class Arch(Feature):
     def __init__(self, value: str, description=None):
-        super(Arch, self).__init__(value, description=description)
+        super().__init__(value, description=description)
         self.name = "arch"
 
 
 OS_WINDOWS = "windows"
 OS_LINUX = "linux"
 OS_MACOS = "macos"
+# dotnet
+OS_ANY = "any"
 VALID_OS = {os.value for os in capa.features.extractors.elf.OS}
-VALID_OS.update({OS_WINDOWS, OS_LINUX, OS_MACOS})
+VALID_OS.update({OS_WINDOWS, OS_LINUX, OS_MACOS, OS_ANY})
 
 
 class OS(Feature):
     def __init__(self, value: str, description=None):
-        super(OS, self).__init__(value, description=description)
+        super().__init__(value, description=description)
         self.name = "os"
 
 
 FORMAT_PE = "pe"
 FORMAT_ELF = "elf"
-VALID_FORMAT = (FORMAT_PE, FORMAT_ELF)
+FORMAT_DOTNET = "dotnet"
+VALID_FORMAT = (FORMAT_PE, FORMAT_ELF, FORMAT_DOTNET)
+# internal only, not to be used in rules
+FORMAT_AUTO = "auto"
+FORMAT_SC32 = "sc32"
+FORMAT_SC64 = "sc64"
+FORMAT_FREEZE = "freeze"
+FORMAT_UNKNOWN = "unknown"
 
 
 class Format(Feature):
     def __init__(self, value: str, description=None):
-        super(Format, self).__init__(value, description=description)
+        super().__init__(value, description=description)
         self.name = "format"
 
 

capa/features/extractors/base_extractor.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -7,23 +7,60 @@
 # See the License for the specific language governing permissions and limitations under the License.
 
 import abc
-from typing import Tuple, Iterator, SupportsInt
+import dataclasses
+from typing import Any, Dict, Tuple, Union, Iterator
+from dataclasses import dataclass
 
+import capa.features.address
 from capa.features.common import Feature
+from capa.features.address import Address, AbsoluteVirtualAddress
 
 # feature extractors may reference functions, BBs, insns by opaque handle values.
-# the only requirement of these handles are that they support `__int__`,
-# so that they can be rendered as addresses.
+# you can use the `.address` property to get and render the address of the feature.
 #
 # these handles are only consumed by routines on
 # the feature extractor from which they were created.
-#
-# int(FunctionHandle) -> function start address
-# int(BBHandle) -> BasicBlock start address
-# int(InsnHandle) -> instruction address
-FunctionHandle = SupportsInt
-BBHandle = SupportsInt
-InsnHandle = SupportsInt
+
+
+@dataclass
+class FunctionHandle:
+    """reference to a function recognized by a feature extractor.
+
+    Attributes:
+        address: the address of the function.
+        inner: extractor-specific data.
+        ctx: a context object for the extractor.
+    """
+
+    address: Address
+    inner: Any
+    ctx: Dict[str, Any] = dataclasses.field(default_factory=dict)
+
+
+@dataclass
+class BBHandle:
+    """reference to a basic block recognized by a feature extractor.
+
+    Attributes:
+        address: the address of the basic block start address.
+        inner: extractor-specific data.
+    """
+
+    address: Address
+    inner: Any
+
+
+@dataclass
+class InsnHandle:
+    """reference to a instruction recognized by a feature extractor.
+
+    Attributes:
+        address: the address of the instruction address.
+        inner: extractor-specific data.
+    """
+
+    address: Address
+    inner: Any
 
 
 class FeatureExtractor:
@@ -50,17 +87,21 @@ class FeatureExtractor:
         #  for example, the Vivisect feature extract might require the vw and/or path.
         # this base class doesn't know what to do with that info, though.
         #
-        super(FeatureExtractor, self).__init__()
+        super().__init__()
 
     @abc.abstractmethod
-    def get_base_address(self) -> int:
+    def get_base_address(self) -> Union[AbsoluteVirtualAddress, capa.features.address._NoAddress]:
         """
         fetch the preferred load address at which the sample was analyzed.
+
+        when the base address is `NO_ADDRESS`, then the loader has no concept of a preferred load address.
+        such as: shellcode, .NET modules, etc.
+        in these scenarios, RelativeVirtualAddresses aren't used.
         """
         raise NotImplementedError()
 
     @abc.abstractmethod
-    def extract_global_features(self) -> Iterator[Tuple[Feature, int]]:
+    def extract_global_features(self) -> Iterator[Tuple[Feature, Address]]:
         """
         extract features found at every scope ("global").
 
@@ -71,12 +112,12 @@ class FeatureExtractor:
                 print('0x%x: %s', va, feature)
 
         yields:
-          Tuple[Feature, int]: feature and its location
+          Tuple[Feature, Address]: feature and its location
         """
         raise NotImplementedError()
 
     @abc.abstractmethod
-    def extract_file_features(self) -> Iterator[Tuple[Feature, int]]:
+    def extract_file_features(self) -> Iterator[Tuple[Feature, Address]]:
         """
         extract file-scope features.
 
@@ -87,7 +128,7 @@ class FeatureExtractor:
                 print('0x%x: %s', va, feature)
 
         yields:
-          Tuple[Feature, int]: feature and its location
+          Tuple[Feature, Address]: feature and its location
         """
         raise NotImplementedError()
 
@@ -99,32 +140,33 @@ class FeatureExtractor:
         """
         raise NotImplementedError()
 
-    def is_library_function(self, va: int) -> bool:
+    def is_library_function(self, addr: Address) -> bool:
         """
         is the given address a library function?
         the backend may implement its own function matching algorithm, or none at all.
-        we accept a VA here, rather than function object, to handle addresses identified in instructions.
+        we accept an address here, rather than function object,
+         to handle addresses identified in instructions.
 
         this information is used to:
           - filter out matches in library functions (by default), and
           - recognize when to fetch symbol names for called (non-API) functions
 
         args:
-          va (int): the virtual address of a function.
+          addr (Address): the address of a function.
 
         returns:
           bool: True if the given address is the start of a library function.
         """
         return False
 
-    def get_function_name(self, va: int) -> str:
+    def get_function_name(self, addr: Address) -> str:
         """
         fetch any recognized name for the given address.
         this is only guaranteed to return a value when the given function is a recognized library function.
         we accept a VA here, rather than function object, to handle addresses identified in instructions.
 
         args:
-          va (int): the virtual address of a function.
+          addr (Address): the address of a function.
 
         returns:
           str: the function name
@@ -132,10 +174,10 @@ class FeatureExtractor:
         raises:
           KeyError: when the given function does not have a name.
         """
-        raise KeyError(va)
+        raise KeyError(addr)
 
     @abc.abstractmethod
-    def extract_function_features(self, f: FunctionHandle) -> Iterator[Tuple[Feature, int]]:
+    def extract_function_features(self, f: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
         """
         extract function-scope features.
         the arguments are opaque values previously provided by `.get_functions()`, etc.
@@ -144,14 +186,14 @@ class FeatureExtractor:
 
             extractor = VivisectFeatureExtractor(vw, path)
             for function in extractor.get_functions():
-                for feature, va in extractor.extract_function_features(function):
-                    print('0x%x: %s', va, feature)
+                for feature, address in extractor.extract_function_features(function):
+                    print('0x%x: %s', address, feature)
 
         args:
           f [FunctionHandle]: an opaque value previously fetched from `.get_functions()`.
 
         yields:
-          Tuple[Feature, int]: feature and its location
+          Tuple[Feature, Address]: feature and its location
         """
         raise NotImplementedError()
 
@@ -164,7 +206,7 @@ class FeatureExtractor:
         raise NotImplementedError()
 
     @abc.abstractmethod
-    def extract_basic_block_features(self, f: FunctionHandle, bb: BBHandle) -> Iterator[Tuple[Feature, int]]:
+    def extract_basic_block_features(self, f: FunctionHandle, bb: BBHandle) -> Iterator[Tuple[Feature, Address]]:
         """
         extract basic block-scope features.
         the arguments are opaque values previously provided by `.get_functions()`, etc.
@@ -174,15 +216,15 @@ class FeatureExtractor:
             extractor = VivisectFeatureExtractor(vw, path)
             for function in extractor.get_functions():
                 for bb in extractor.get_basic_blocks(function):
-                    for feature, va in extractor.extract_basic_block_features(function, bb):
-                        print('0x%x: %s', va, feature)
+                    for feature, address in extractor.extract_basic_block_features(function, bb):
+                        print('0x%x: %s', address, feature)
 
         args:
           f [FunctionHandle]: an opaque value previously fetched from `.get_functions()`.
           bb [BBHandle]: an opaque value previously fetched from `.get_basic_blocks()`.
 
         yields:
-          Tuple[Feature, int]: feature and its location
+          Tuple[Feature, Address]: feature and its location
         """
         raise NotImplementedError()
 
@@ -195,7 +237,9 @@ class FeatureExtractor:
         raise NotImplementedError()
 
     @abc.abstractmethod
-    def extract_insn_features(self, f: FunctionHandle, bb: BBHandle, insn: InsnHandle) -> Iterator[Tuple[Feature, int]]:
+    def extract_insn_features(
+        self, f: FunctionHandle, bb: BBHandle, insn: InsnHandle
+    ) -> Iterator[Tuple[Feature, Address]]:
         """
         extract instruction-scope features.
         the arguments are opaque values previously provided by `.get_functions()`, etc.
@@ -206,8 +250,8 @@ class FeatureExtractor:
             for function in extractor.get_functions():
                 for bb in extractor.get_basic_blocks(function):
                     for insn in extractor.get_instructions(function, bb):
-                        for feature, va in extractor.extract_insn_features(function, bb, insn):
-                            print('0x%x: %s', va, feature)
+                        for feature, address in extractor.extract_insn_features(function, bb, insn):
+                            print('0x%x: %s', address, feature)
 
         args:
           f [FunctionHandle]: an opaque value previously fetched from `.get_functions()`.
@@ -215,123 +259,6 @@ class FeatureExtractor:
           insn [InsnHandle]: an opaque value previously fetched from `.get_instructions()`.
 
         yields:
-          Tuple[Feature, int]: feature and its location
+          Tuple[Feature, Address]: feature and its location
         """
         raise NotImplementedError()
-
-
-class NullFeatureExtractor(FeatureExtractor):
-    """
-    An extractor that extracts some user-provided features.
-    The structure of the single parameter is demonstrated in the example below.
-
-    This is useful for testing, as we can provide expected values and see if matching works.
-    Also, this is how we represent features deserialized from a freeze file.
-
-    example::
-
-        extractor = NullFeatureExtractor({
-            'base address: 0x401000,
-            'global features': [
-                (0x0, capa.features.Arch('i386')),
-                (0x0, capa.features.OS('linux')),
-            ],
-            'file features': [
-                (0x402345, capa.features.Characteristic('embedded pe')),
-            ],
-            'functions': {
-                0x401000: {
-                    'features': [
-                        (0x401000, capa.features.Characteristic('nzxor')),
-                    ],
-                    'basic blocks': {
-                        0x401000: {
-                            'features': [
-                                (0x401000, capa.features.Characteristic('tight-loop')),
-                            ],
-                            'instructions': {
-                                0x401000: {
-                                    'features': [
-                                        (0x401000, capa.features.Characteristic('nzxor')),
-                                    ],
-                                },
-                                0x401002: ...
-                            }
-                        },
-                        0x401005: ...
-                    }
-                },
-                0x40200: ...
-            }
-        )
-    """
-
-    def __init__(self, features):
-        super(NullFeatureExtractor, self).__init__()
-        self.features = features
-
-    def get_base_address(self):
-        return self.features["base address"]
-
-    def extract_global_features(self):
-        for p in self.features.get("global features", []):
-            va, feature = p
-            yield feature, va
-
-    def extract_file_features(self):
-        for p in self.features.get("file features", []):
-            va, feature = p
-            yield feature, va
-
-    def get_functions(self):
-        for va in sorted(self.features["functions"].keys()):
-            yield va
-
-    def extract_function_features(self, f):
-        for p in self.features.get("functions", {}).get(f, {}).get("features", []):  # noqa: E127 line over-indented
-            va, feature = p
-            yield feature, va
-
-    def get_basic_blocks(self, f):
-        for va in sorted(
-            self.features.get("functions", {})  # noqa: E127 line over-indented
-            .get(f, {})
-            .get("basic blocks", {})
-            .keys()
-        ):
-            yield va
-
-    def extract_basic_block_features(self, f, bb):
-        for p in (
-            self.features.get("functions", {})  # noqa: E127 line over-indented
-            .get(f, {})
-            .get("basic blocks", {})
-            .get(bb, {})
-            .get("features", [])
-        ):
-            va, feature = p
-            yield feature, va
-
-    def get_instructions(self, f, bb):
-        for va in sorted(
-            self.features.get("functions", {})  # noqa: E127 line over-indented
-            .get(f, {})
-            .get("basic blocks", {})
-            .get(bb, {})
-            .get("instructions", {})
-            .keys()
-        ):
-            yield va
-
-    def extract_insn_features(self, f, bb, insn):
-        for p in (
-            self.features.get("functions", {})  # noqa: E127 line over-indented
-            .get(f, {})
-            .get("basic blocks", {})
-            .get(bb, {})
-            .get("instructions", {})
-            .get(insn, {})
-            .get("features", [])
-        ):
-            va, feature = p
-            yield feature, va

capa/features/extractors/common.py

@@ -2,33 +2,39 @@ import io
 import logging
 import binascii
 import contextlib
+from typing import Tuple, Iterator
 
 import pefile
 
 import capa.features
 import capa.features.extractors.elf
 import capa.features.extractors.pefile
-from capa.features.common import OS, FORMAT_PE, FORMAT_ELF, OS_WINDOWS, Arch, Format, String
+import capa.features.extractors.strings
+from capa.features.common import OS, FORMAT_PE, FORMAT_ELF, OS_WINDOWS, FORMAT_FREEZE, Arch, Format, String, Feature
+from capa.features.freeze import is_freeze
+from capa.features.address import NO_ADDRESS, Address, FileOffsetAddress
 
 logger = logging.getLogger(__name__)
 
 
-def extract_file_strings(buf, **kwargs):
+def extract_file_strings(buf, **kwargs) -> Iterator[Tuple[String, Address]]:
     """
     extract ASCII and UTF-16 LE strings from file
     """
     for s in capa.features.extractors.strings.extract_ascii_strings(buf):
-        yield String(s.s), s.offset
+        yield String(s.s), FileOffsetAddress(s.offset)
 
     for s in capa.features.extractors.strings.extract_unicode_strings(buf):
-        yield String(s.s), s.offset
+        yield String(s.s), FileOffsetAddress(s.offset)
 
 
-def extract_format(buf):
+def extract_format(buf) -> Iterator[Tuple[Feature, Address]]:
     if buf.startswith(b"MZ"):
-        yield Format(FORMAT_PE), 0x0
+        yield Format(FORMAT_PE), NO_ADDRESS
     elif buf.startswith(b"\x7fELF"):
-        yield Format(FORMAT_ELF), 0x0
+        yield Format(FORMAT_ELF), NO_ADDRESS
+    elif is_freeze(buf):
+        yield Format(FORMAT_FREEZE), NO_ADDRESS
     else:
         # we likely end up here:
         #  1. handling a file format (e.g. macho)
@@ -38,7 +44,7 @@ def extract_format(buf):
         return
 
 
-def extract_arch(buf):
+def extract_arch(buf) -> Iterator[Tuple[Feature, Address]]:
     if buf.startswith(b"MZ"):
         yield from capa.features.extractors.pefile.extract_file_arch(pe=pefile.PE(data=buf))
 
@@ -50,7 +56,7 @@ def extract_arch(buf):
             logger.debug("unsupported arch: %s", arch)
             return
 
-        yield Arch(arch), 0x0
+        yield Arch(arch), NO_ADDRESS
 
     else:
         # we likely end up here:
@@ -58,7 +64,7 @@ def extract_arch(buf):
         #  2. handling a new file format (e.g. macho)
         #
         # for (1) we can't do much - its shellcode and all bets are off.
-        # we could maybe accept a futher CLI argument to specify the arch,
+        # we could maybe accept a further CLI argument to specify the arch,
         # but i think this would be rarely used.
         # rules that rely on arch conditions will fail to match on shellcode.
         #
@@ -67,9 +73,9 @@ def extract_arch(buf):
         return
 
 
-def extract_os(buf):
+def extract_os(buf) -> Iterator[Tuple[Feature, Address]]:
     if buf.startswith(b"MZ"):
-        yield OS(OS_WINDOWS), 0x0
+        yield OS(OS_WINDOWS), NO_ADDRESS
     elif buf.startswith(b"\x7fELF"):
         with contextlib.closing(io.BytesIO(buf)) as f:
             os = capa.features.extractors.elf.detect_elf_os(f)
@@ -78,7 +84,7 @@ def extract_os(buf):
             logger.debug("unsupported os: %s", os)
             return
 
-        yield OS(os), 0x0
+        yield OS(os), NO_ADDRESS
 
     else:
         # we likely end up here:
@@ -86,7 +92,7 @@ def extract_os(buf):
         #  2. handling a new file format (e.g. macho)
         #
         # for (1) we can't do much - its shellcode and all bets are off.
-        # we could maybe accept a futher CLI argument to specify the OS,
+        # we could maybe accept a further CLI argument to specify the OS,
         # but i think this would be rarely used.
         # rules that rely on OS conditions will fail to match on shellcode.
         #

capa/features/extractors/dnfile/extractor.py

@@ -0,0 +1,154 @@
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
+# Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at: [package root]/LICENSE.txt
+# Unless required by applicable law or agreed to in writing, software distributed under the License
+#  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and limitations under the License.
+
+from __future__ import annotations
+
+from typing import Dict, List, Tuple, Union, Iterator, Optional
+
+import dnfile
+from dncil.cil.opcode import OpCodes
+
+import capa.features.extractors
+import capa.features.extractors.dotnetfile
+import capa.features.extractors.dnfile.file
+import capa.features.extractors.dnfile.insn
+import capa.features.extractors.dnfile.function
+from capa.features.common import Feature
+from capa.features.address import NO_ADDRESS, Address, DNTokenAddress, DNTokenOffsetAddress
+from capa.features.extractors.dnfile.types import DnType, DnUnmanagedMethod
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle, FeatureExtractor
+from capa.features.extractors.dnfile.helpers import (
+    get_dotnet_types,
+    get_dotnet_fields,
+    get_dotnet_managed_imports,
+    get_dotnet_managed_methods,
+    get_dotnet_unmanaged_imports,
+    get_dotnet_managed_method_bodies,
+)
+
+
+class DnFileFeatureExtractorCache:
+    def __init__(self, pe: dnfile.dnPE):
+        self.imports: Dict[int, Union[DnType, DnUnmanagedMethod]] = {}
+        self.native_imports: Dict[int, Union[DnType, DnUnmanagedMethod]] = {}
+        self.methods: Dict[int, Union[DnType, DnUnmanagedMethod]] = {}
+        self.fields: Dict[int, Union[DnType, DnUnmanagedMethod]] = {}
+        self.types: Dict[int, Union[DnType, DnUnmanagedMethod]] = {}
+
+        for import_ in get_dotnet_managed_imports(pe):
+            self.imports[import_.token] = import_
+        for native_import in get_dotnet_unmanaged_imports(pe):
+            self.native_imports[native_import.token] = native_import
+        for method in get_dotnet_managed_methods(pe):
+            self.methods[method.token] = method
+        for field in get_dotnet_fields(pe):
+            self.fields[field.token] = field
+        for type_ in get_dotnet_types(pe):
+            self.types[type_.token] = type_
+
+    def get_import(self, token: int) -> Optional[Union[DnType, DnUnmanagedMethod]]:
+        return self.imports.get(token, None)
+
+    def get_native_import(self, token: int) -> Optional[Union[DnType, DnUnmanagedMethod]]:
+        return self.native_imports.get(token, None)
+
+    def get_method(self, token: int) -> Optional[Union[DnType, DnUnmanagedMethod]]:
+        return self.methods.get(token, None)
+
+    def get_field(self, token: int) -> Optional[Union[DnType, DnUnmanagedMethod]]:
+        return self.fields.get(token, None)
+
+    def get_type(self, token: int) -> Optional[Union[DnType, DnUnmanagedMethod]]:
+        return self.types.get(token, None)
+
+
+class DnfileFeatureExtractor(FeatureExtractor):
+    def __init__(self, path: str):
+        super().__init__()
+        self.pe: dnfile.dnPE = dnfile.dnPE(path)
+
+        # pre-compute .NET token lookup tables; each .NET method has access to this cache for feature extraction
+        # most relevant at instruction scope
+        self.token_cache: DnFileFeatureExtractorCache = DnFileFeatureExtractorCache(self.pe)
+
+        # pre-compute these because we'll yield them at *every* scope.
+        self.global_features: List[Tuple[Feature, Address]] = []
+        self.global_features.extend(capa.features.extractors.dotnetfile.extract_file_format())
+        self.global_features.extend(capa.features.extractors.dotnetfile.extract_file_os(pe=self.pe))
+        self.global_features.extend(capa.features.extractors.dotnetfile.extract_file_arch(pe=self.pe))
+
+    def get_base_address(self):
+        return NO_ADDRESS
+
+    def extract_global_features(self):
+        yield from self.global_features
+
+    def extract_file_features(self):
+        yield from capa.features.extractors.dnfile.file.extract_features(self.pe)
+
+    def get_functions(self) -> Iterator[FunctionHandle]:
+        # create a method lookup table
+        methods: Dict[Address, FunctionHandle] = {}
+        for token, method in get_dotnet_managed_method_bodies(self.pe):
+            fh: FunctionHandle = FunctionHandle(
+                address=DNTokenAddress(token),
+                inner=method,
+                ctx={"pe": self.pe, "calls_from": set(), "calls_to": set(), "cache": self.token_cache},
+            )
+
+            # method tokens should be unique
+            assert fh.address not in methods.keys()
+            methods[fh.address] = fh
+
+        # calculate unique calls to/from each method
+        for fh in methods.values():
+            for insn in fh.inner.instructions:
+                if insn.opcode not in (
+                    OpCodes.Call,
+                    OpCodes.Callvirt,
+                    OpCodes.Jmp,
+                    OpCodes.Newobj,
+                ):
+                    continue
+
+                address: DNTokenAddress = DNTokenAddress(insn.operand.value)
+
+                # record call to destination method; note: we only consider MethodDef methods for destinations
+                dest: Optional[FunctionHandle] = methods.get(address, None)
+                if dest is not None:
+                    dest.ctx["calls_to"].add(fh.address)
+
+                # record call from source method; note: we record all unique calls from a MethodDef method, not just
+                # those calls to other MethodDef methods e.g. calls to imported MemberRef methods
+                fh.ctx["calls_from"].add(address)
+
+        yield from methods.values()
+
+    def extract_function_features(self, fh) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.dnfile.function.extract_features(fh)
+
+    def get_basic_blocks(self, f) -> Iterator[BBHandle]:
+        # each dotnet method is considered 1 basic block
+        yield BBHandle(
+            address=f.address,
+            inner=f.inner,
+        )
+
+    def extract_basic_block_features(self, fh, bbh):
+        # we don't support basic block features
+        yield from []
+
+    def get_instructions(self, fh, bbh):
+        for insn in bbh.inner.instructions:
+            yield InsnHandle(
+                address=DNTokenOffsetAddress(bbh.address, insn.offset - (fh.inner.offset + fh.inner.header_size)),
+                inner=insn,
+            )
+
+    def extract_insn_features(self, fh, bbh, ih) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.dnfile.insn.extract_features(fh, bbh, ih)

capa/features/extractors/dnfile/file.py

@@ -0,0 +1,63 @@
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
+# Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at: [package root]/LICENSE.txt
+# Unless required by applicable law or agreed to in writing, software distributed under the License
+#  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and limitations under the License.
+
+from __future__ import annotations
+
+from typing import Tuple, Iterator
+
+import dnfile
+
+import capa.features.extractors.dotnetfile
+from capa.features.file import Import, FunctionName
+from capa.features.common import Class, Format, String, Feature, Namespace, Characteristic
+from capa.features.address import Address
+
+
+def extract_file_import_names(pe: dnfile.dnPE) -> Iterator[Tuple[Import, Address]]:
+    yield from capa.features.extractors.dotnetfile.extract_file_import_names(pe=pe)
+
+
+def extract_file_format(pe: dnfile.dnPE) -> Iterator[Tuple[Format, Address]]:
+    yield from capa.features.extractors.dotnetfile.extract_file_format(pe=pe)
+
+
+def extract_file_function_names(pe: dnfile.dnPE) -> Iterator[Tuple[FunctionName, Address]]:
+    yield from capa.features.extractors.dotnetfile.extract_file_function_names(pe=pe)
+
+
+def extract_file_strings(pe: dnfile.dnPE) -> Iterator[Tuple[String, Address]]:
+    yield from capa.features.extractors.dotnetfile.extract_file_strings(pe=pe)
+
+
+def extract_file_mixed_mode_characteristic_features(pe: dnfile.dnPE) -> Iterator[Tuple[Characteristic, Address]]:
+    yield from capa.features.extractors.dotnetfile.extract_file_mixed_mode_characteristic_features(pe=pe)
+
+
+def extract_file_namespace_features(pe: dnfile.dnPE) -> Iterator[Tuple[Namespace, Address]]:
+    yield from capa.features.extractors.dotnetfile.extract_file_namespace_features(pe=pe)
+
+
+def extract_file_class_features(pe: dnfile.dnPE) -> Iterator[Tuple[Class, Address]]:
+    yield from capa.features.extractors.dotnetfile.extract_file_class_features(pe=pe)
+
+
+def extract_features(pe: dnfile.dnPE) -> Iterator[Tuple[Feature, Address]]:
+    for file_handler in FILE_HANDLERS:
+        for feature, address in file_handler(pe):
+            yield feature, address
+
+
+FILE_HANDLERS = (
+    extract_file_import_names,
+    extract_file_function_names,
+    extract_file_strings,
+    extract_file_format,
+    extract_file_mixed_mode_characteristic_features,
+    extract_file_namespace_features,
+    extract_file_class_features,
+)

capa/features/extractors/dnfile/function.py

@@ -0,0 +1,50 @@
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
+# Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at: [package root]/LICENSE.txt
+# Unless required by applicable law or agreed to in writing, software distributed under the License
+#  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and limitations under the License.
+
+from __future__ import annotations
+
+import logging
+from typing import Tuple, Iterator
+
+from capa.features.common import Feature, Characteristic
+from capa.features.address import Address
+from capa.features.extractors.base_extractor import FunctionHandle
+
+logger = logging.getLogger(__name__)
+
+
+def extract_function_calls_to(fh: FunctionHandle) -> Iterator[Tuple[Characteristic, Address]]:
+    """extract callers to a function"""
+    for dest in fh.ctx["calls_to"]:
+        yield Characteristic("calls to"), dest
+
+
+def extract_function_calls_from(fh: FunctionHandle) -> Iterator[Tuple[Characteristic, Address]]:
+    """extract callers from a function"""
+    for src in fh.ctx["calls_from"]:
+        yield Characteristic("calls from"), src
+
+
+def extract_recursive_call(fh: FunctionHandle) -> Iterator[Tuple[Characteristic, Address]]:
+    """extract recursive function call"""
+    if fh.address in fh.ctx["calls_to"]:
+        yield Characteristic("recursive call"), fh.address
+
+
+def extract_function_loop(fh: FunctionHandle) -> Iterator[Tuple[Characteristic, Address]]:
+    """extract loop indicators from a function"""
+    raise NotImplementedError()
+
+
+def extract_features(fh: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
+    for func_handler in FUNCTION_HANDLERS:
+        for feature, addr in func_handler(fh):
+            yield feature, addr
+
+
+FUNCTION_HANDLERS = (extract_function_calls_to, extract_function_calls_from, extract_recursive_call)

capa/features/extractors/dnfile/helpers.py

@@ -0,0 +1,335 @@
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
+# Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at: [package root]/LICENSE.txt
+# Unless required by applicable law or agreed to in writing, software distributed under the License
+#  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and limitations under the License.
+
+from __future__ import annotations
+
+import logging
+from typing import Dict, Tuple, Union, Iterator, Optional
+
+import dnfile
+from dncil.cil.body import CilMethodBody
+from dncil.cil.error import MethodBodyFormatError
+from dncil.clr.token import Token, StringToken, InvalidToken
+from dncil.cil.body.reader import CilMethodBodyReaderBase
+
+from capa.features.common import FeatureAccess
+from capa.features.extractors.dnfile.types import DnType, DnUnmanagedMethod
+
+logger = logging.getLogger(__name__)
+
+
+class DnfileMethodBodyReader(CilMethodBodyReaderBase):
+    def __init__(self, pe: dnfile.dnPE, row: dnfile.mdtable.MethodDefRow):
+        self.pe: dnfile.dnPE = pe
+        self.offset: int = self.pe.get_offset_from_rva(row.Rva)
+
+    def read(self, n: int) -> bytes:
+        data: bytes = self.pe.get_data(self.pe.get_rva_from_offset(self.offset), n)
+        self.offset += n
+        return data
+
+    def tell(self) -> int:
+        return self.offset
+
+    def seek(self, offset: int) -> int:
+        self.offset = offset
+        return self.offset
+
+
+def resolve_dotnet_token(pe: dnfile.dnPE, token: Token) -> Union[dnfile.base.MDTableRow, InvalidToken, str]:
+    """map generic token to string or table row"""
+    assert pe.net is not None
+    assert pe.net.mdtables is not None
+
+    if isinstance(token, StringToken):
+        user_string: Optional[str] = read_dotnet_user_string(pe, token)
+        if user_string is None:
+            return InvalidToken(token.value)
+        return user_string
+
+    table: Optional[dnfile.base.ClrMetaDataTable] = pe.net.mdtables.tables.get(token.table, None)
+    if table is None:
+        # table index is not valid
+        return InvalidToken(token.value)
+
+    try:
+        return table.rows[token.rid - 1]
+    except IndexError:
+        # table index is valid but row index is not valid
+        return InvalidToken(token.value)
+
+
+def read_dotnet_method_body(pe: dnfile.dnPE, row: dnfile.mdtable.MethodDefRow) -> Optional[CilMethodBody]:
+    """read dotnet method body"""
+    try:
+        return CilMethodBody(DnfileMethodBodyReader(pe, row))
+    except MethodBodyFormatError as e:
+        logger.debug("failed to parse managed method body @ 0x%08x (%s)", row.Rva, e)
+        return None
+
+
+def read_dotnet_user_string(pe: dnfile.dnPE, token: StringToken) -> Optional[str]:
+    """read user string from #US stream"""
+    assert pe.net is not None
+
+    if pe.net.user_strings is None:
+        # stream may not exist (seen in obfuscated .NET)
+        logger.debug("#US stream does not exist for stream index 0x%08x", token.rid)
+        return None
+
+    try:
+        user_string: Optional[dnfile.stream.UserString] = pe.net.user_strings.get_us(token.rid)
+    except UnicodeDecodeError as e:
+        logger.debug("failed to decode #US stream index 0x%08x (%s)", token.rid, e)
+        return None
+
+    if user_string is None:
+        return None
+
+    return user_string.value
+
+
+def get_dotnet_managed_imports(pe: dnfile.dnPE) -> Iterator[DnType]:
+    """get managed imports from MemberRef table
+
+    see https://www.ntcore.com/files/dotnetformat.htm
+
+    10 - MemberRef Table
+        Each row represents an imported method
+            Class (index into the TypeRef, ModuleRef, MethodDef, TypeSpec or TypeDef tables)
+            Name (index into String heap)
+    01 - TypeRef Table
+        Each row represents an imported class, its namespace and the assembly which contains it
+            TypeName (index into String heap)
+            TypeNamespace (index into String heap)
+    """
+    for rid, member_ref in iter_dotnet_table(pe, dnfile.mdtable.MemberRef.number):
+        assert isinstance(member_ref, dnfile.mdtable.MemberRefRow)
+
+        if not isinstance(member_ref.Class.row, dnfile.mdtable.TypeRefRow):
+            # only process class imports from TypeRef table
+            continue
+
+        token: int = calculate_dotnet_token_value(dnfile.mdtable.MemberRef.number, rid)
+        access: Optional[str]
+
+        # assume .NET imports starting with get_/set_ are used to access a property
+        if member_ref.Name.startswith("get_"):
+            access = FeatureAccess.READ
+        elif member_ref.Name.startswith("set_"):
+            access = FeatureAccess.WRITE
+        else:
+            access = None
+
+        member_ref_name: str = member_ref.Name
+        if member_ref_name.startswith(("get_", "set_")):
+            # remove get_/set_ from MemberRef name
+            member_ref_name = member_ref_name[4:]
+
+        yield DnType(
+            token,
+            member_ref.Class.row.TypeName,
+            namespace=member_ref.Class.row.TypeNamespace,
+            member=member_ref_name,
+            access=access,
+        )
+
+
+def get_dotnet_methoddef_property_accessors(pe: dnfile.dnPE) -> Iterator[Tuple[int, str]]:
+    """get MethodDef methods used to access properties
+
+    see https://www.ntcore.com/files/dotnetformat.htm
+
+    24 - MethodSemantics Table
+        Links Events and Properties to specific methods. For example one Event can be associated to more methods. A property uses this table to associate get/set methods.
+            Semantics (a 2-byte bitmask of type MethodSemanticsAttributes)
+            Method (index into the MethodDef table)
+            Association (index into the Event or Property table; more precisely, a HasSemantics coded index)
+    """
+    for rid, method_semantics in iter_dotnet_table(pe, dnfile.mdtable.MethodSemantics.number):
+        assert isinstance(method_semantics, dnfile.mdtable.MethodSemanticsRow)
+
+        if method_semantics.Association.row is None:
+            logger.debug("MethodSemantics[0x%X] Association row is None", rid)
+            continue
+
+        if isinstance(method_semantics.Association.row, dnfile.mdtable.EventRow):
+            # ignore events
+            logger.debug("MethodSemantics[0x%X] ignoring Event", rid)
+            continue
+
+        if method_semantics.Method.table is None:
+            logger.debug("MethodSemantics[0x%X] Method table is None", rid)
+            continue
+
+        token: int = calculate_dotnet_token_value(
+            method_semantics.Method.table.number, method_semantics.Method.row_index
+        )
+
+        if method_semantics.Semantics.msSetter:
+            yield token, FeatureAccess.WRITE
+        elif method_semantics.Semantics.msGetter:
+            yield token, FeatureAccess.READ
+
+
+def get_dotnet_managed_methods(pe: dnfile.dnPE) -> Iterator[DnType]:
+    """get managed method names from TypeDef table
+
+    see https://www.ntcore.com/files/dotnetformat.htm
+
+    02 - TypeDef Table
+        Each row represents a class in the current assembly.
+            TypeName (index into String heap)
+            TypeNamespace (index into String heap)
+            MethodList (index into MethodDef table; it marks the first of a contiguous run of Methods owned by this Type)
+    """
+    accessor_map: Dict[int, str] = {}
+    for methoddef, methoddef_access in get_dotnet_methoddef_property_accessors(pe):
+        accessor_map[methoddef] = methoddef_access
+
+    for rid, typedef in iter_dotnet_table(pe, dnfile.mdtable.TypeDef.number):
+        assert isinstance(typedef, dnfile.mdtable.TypeDefRow)
+
+        for idx, method in enumerate(typedef.MethodList):
+            if method.table is None:
+                logger.debug("TypeDef[0x%X] MethodList[0x%X] table is None", rid, idx)
+                continue
+            if method.row is None:
+                logger.debug("TypeDef[0x%X] MethodList[0x%X] row is None", rid, idx)
+                continue
+
+            token: int = calculate_dotnet_token_value(method.table.number, method.row_index)
+            access: Optional[str] = accessor_map.get(token, None)
+
+            method_name: str = method.row.Name
+            if method_name.startswith(("get_", "set_")):
+                # remove get_/set_
+                method_name = method_name[4:]
+
+            yield DnType(token, typedef.TypeName, namespace=typedef.TypeNamespace, member=method_name, access=access)
+
+
+def get_dotnet_fields(pe: dnfile.dnPE) -> Iterator[DnType]:
+    """get fields from TypeDef table
+
+    see https://www.ntcore.com/files/dotnetformat.htm
+
+    02 - TypeDef Table
+        Each row represents a class in the current assembly.
+            TypeName (index into String heap)
+            TypeNamespace (index into String heap)
+            FieldList (index into Field table; it marks the first of a contiguous run of Fields owned by this Type)
+    """
+    for rid, typedef in iter_dotnet_table(pe, dnfile.mdtable.TypeDef.number):
+        assert isinstance(typedef, dnfile.mdtable.TypeDefRow)
+
+        for idx, field in enumerate(typedef.FieldList):
+            if field.table is None:
+                logger.debug("TypeDef[0x%X] FieldList[0x%X] table is None", rid, idx)
+                continue
+            if field.row is None:
+                logger.debug("TypeDef[0x%X] FieldList[0x%X] row is None", rid, idx)
+                continue
+            token: int = calculate_dotnet_token_value(field.table.number, field.row_index)
+            yield DnType(token, typedef.TypeName, namespace=typedef.TypeNamespace, member=field.row.Name)
+
+
+def get_dotnet_managed_method_bodies(pe: dnfile.dnPE) -> Iterator[Tuple[int, CilMethodBody]]:
+    """get managed methods from MethodDef table"""
+    for rid, method_def in iter_dotnet_table(pe, dnfile.mdtable.MethodDef.number):
+        assert isinstance(method_def, dnfile.mdtable.MethodDefRow)
+
+        if not method_def.ImplFlags.miIL or any((method_def.Flags.mdAbstract, method_def.Flags.mdPinvokeImpl)):
+            # skip methods that do not have a method body
+            continue
+
+        body: Optional[CilMethodBody] = read_dotnet_method_body(pe, method_def)
+        if body is None:
+            logger.debug("MethodDef[0x%X] method body is None", rid)
+            continue
+
+        token: int = calculate_dotnet_token_value(dnfile.mdtable.MethodDef.number, rid)
+        yield token, body
+
+
+def get_dotnet_unmanaged_imports(pe: dnfile.dnPE) -> Iterator[DnUnmanagedMethod]:
+    """get unmanaged imports from ImplMap table
+
+    see https://www.ntcore.com/files/dotnetformat.htm
+
+    28 - ImplMap Table
+        ImplMap table holds information about unmanaged methods that can be reached from managed code, using PInvoke dispatch
+            MemberForwarded (index into the Field or MethodDef table; more precisely, a MemberForwarded coded index)
+            ImportName (index into the String heap)
+            ImportScope (index into the ModuleRef table)
+    """
+    for rid, impl_map in iter_dotnet_table(pe, dnfile.mdtable.ImplMap.number):
+        assert isinstance(impl_map, dnfile.mdtable.ImplMapRow)
+
+        module: str
+        if impl_map.ImportScope.row is None:
+            logger.debug("ImplMap[0x%X] ImportScope row is None", rid)
+            module = ""
+        else:
+            module = impl_map.ImportScope.row.Name
+        method: str = impl_map.ImportName
+
+        member_forward_table: int
+        if impl_map.MemberForwarded.table is None:
+            logger.debug("ImplMap[0x%X] MemberForwarded table is None", rid)
+            continue
+        else:
+            member_forward_table = impl_map.MemberForwarded.table.number
+        member_forward_row: int = impl_map.MemberForwarded.row_index
+
+        # ECMA says "Each row of the ImplMap table associates a row in the MethodDef table (MemberForwarded) with the
+        # name of a routine (ImportName) in some unmanaged DLL (ImportScope)"; so we calculate and map the MemberForwarded
+        # MethodDef table token to help us later record native import method calls made from CIL
+        token: int = calculate_dotnet_token_value(member_forward_table, member_forward_row)
+
+        # like Kernel32.dll
+        if module and "." in module:
+            module = module.split(".")[0]
+
+        # like kernel32.CreateFileA
+        yield DnUnmanagedMethod(token, module, method)
+
+
+def get_dotnet_types(pe: dnfile.dnPE) -> Iterator[DnType]:
+    """get .NET types from TypeDef and TypeRef tables"""
+    for rid, typedef in iter_dotnet_table(pe, dnfile.mdtable.TypeDef.number):
+        assert isinstance(typedef, dnfile.mdtable.TypeDefRow)
+
+        typedef_token: int = calculate_dotnet_token_value(dnfile.mdtable.TypeDef.number, rid)
+        yield DnType(typedef_token, typedef.TypeName, namespace=typedef.TypeNamespace)
+
+    for rid, typeref in iter_dotnet_table(pe, dnfile.mdtable.TypeRef.number):
+        assert isinstance(typeref, dnfile.mdtable.TypeRefRow)
+
+        typeref_token: int = calculate_dotnet_token_value(dnfile.mdtable.TypeRef.number, rid)
+        yield DnType(typeref_token, typeref.TypeName, namespace=typeref.TypeNamespace)
+
+
+def calculate_dotnet_token_value(table: int, rid: int) -> int:
+    return ((table & 0xFF) << Token.TABLE_SHIFT) | (rid & Token.RID_MASK)
+
+
+def is_dotnet_mixed_mode(pe: dnfile.dnPE) -> bool:
+    assert pe.net is not None
+    assert pe.net.Flags is not None
+
+    return not bool(pe.net.Flags.CLR_ILONLY)
+
+
+def iter_dotnet_table(pe: dnfile.dnPE, table_index: int) -> Iterator[Tuple[int, dnfile.base.MDTableRow]]:
+    assert pe.net is not None
+    assert pe.net.mdtables is not None
+
+    for rid, row in enumerate(pe.net.mdtables.tables.get(table_index, [])):
+        # .NET tables are 1-indexed
+        yield rid + 1, row

capa/features/extractors/dnfile/insn.py

@@ -0,0 +1,227 @@
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
+# Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at: [package root]/LICENSE.txt
+# Unless required by applicable law or agreed to in writing, software distributed under the License
+#  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and limitations under the License.
+
+from __future__ import annotations
+
+import logging
+from typing import TYPE_CHECKING, Any, Dict, Tuple, Union, Iterator, Optional
+
+if TYPE_CHECKING:
+    from capa.features.extractors.dnfile.extractor import DnFileFeatureExtractorCache
+
+import dnfile
+from dncil.clr.token import Token, StringToken, InvalidToken
+from dncil.cil.opcode import OpCodes
+
+import capa.features.extractors.helpers
+from capa.features.insn import API, Number, Property
+from capa.features.common import Class, String, Feature, Namespace, FeatureAccess, Characteristic
+from capa.features.address import Address
+from capa.features.extractors.dnfile.types import DnType, DnUnmanagedMethod
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle
+from capa.features.extractors.dnfile.helpers import (
+    resolve_dotnet_token,
+    read_dotnet_user_string,
+    calculate_dotnet_token_value,
+)
+
+logger = logging.getLogger(__name__)
+
+
+def get_callee(
+    pe: dnfile.dnPE, cache: DnFileFeatureExtractorCache, token: Token
+) -> Optional[Union[DnType, DnUnmanagedMethod]]:
+    """map .NET token to un/managed (generic) method"""
+    token_: int
+    if token.table == dnfile.mdtable.MethodSpec.number:
+        # map MethodSpec to MethodDef or MemberRef
+        row: Union[dnfile.base.MDTableRow, InvalidToken, str] = resolve_dotnet_token(pe, token)
+        assert isinstance(row, dnfile.mdtable.MethodSpecRow)
+
+        if row.Method.table is None:
+            logger.debug("MethodSpec[0x%X] Method table is None", token.rid)
+            return None
+
+        token_ = calculate_dotnet_token_value(row.Method.table.number, row.Method.row_index)
+    else:
+        token_ = token.value
+
+    callee: Optional[Union[DnType, DnUnmanagedMethod]] = cache.get_import(token_)
+    if callee is None:
+        # we must check unmanaged imports before managed methods because we map forwarded managed methods
+        # to their unmanaged imports; we prefer a forwarded managed method be mapped to its unmanaged import for analysis
+        callee = cache.get_native_import(token_)
+        if callee is None:
+            callee = cache.get_method(token_)
+    return callee
+
+
+def extract_insn_api_features(fh: FunctionHandle, bh, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """parse instruction API features"""
+    if ih.inner.opcode not in (
+        OpCodes.Call,
+        OpCodes.Callvirt,
+        OpCodes.Jmp,
+        OpCodes.Newobj,
+    ):
+        return
+
+    callee: Optional[Union[DnType, DnUnmanagedMethod]] = get_callee(fh.ctx["pe"], fh.ctx["cache"], ih.inner.operand)
+    if isinstance(callee, DnType):
+        # ignore methods used to access properties
+        if callee.access is None:
+            # like System.IO.File::Delete
+            yield API(str(callee)), ih.address
+    elif isinstance(callee, DnUnmanagedMethod):
+        # like kernel32.CreateFileA
+        for name in capa.features.extractors.helpers.generate_symbols(callee.module, callee.method):
+            yield API(name), ih.address
+
+
+def extract_insn_property_features(fh: FunctionHandle, bh, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """parse instruction property features"""
+    name: Optional[str] = None
+    access: Optional[str] = None
+
+    if ih.inner.opcode in (OpCodes.Call, OpCodes.Callvirt, OpCodes.Jmp):
+        # property access via MethodDef or MemberRef
+        callee: Optional[Union[DnType, DnUnmanagedMethod]] = get_callee(fh.ctx["pe"], fh.ctx["cache"], ih.inner.operand)
+        if isinstance(callee, DnType):
+            if callee.access is not None:
+                name = str(callee)
+                access = callee.access
+
+    elif ih.inner.opcode in (OpCodes.Ldfld, OpCodes.Ldflda, OpCodes.Ldsfld, OpCodes.Ldsflda):
+        # property read via Field
+        read_field: Optional[Union[DnType, DnUnmanagedMethod]] = fh.ctx["cache"].get_field(ih.inner.operand.value)
+        if read_field is not None:
+            name = str(read_field)
+            access = FeatureAccess.READ
+
+    elif ih.inner.opcode in (OpCodes.Stfld, OpCodes.Stsfld):
+        # property write via Field
+        write_field: Optional[Union[DnType, DnUnmanagedMethod]] = fh.ctx["cache"].get_field(ih.inner.operand.value)
+        if write_field is not None:
+            name = str(write_field)
+            access = FeatureAccess.WRITE
+
+    if name is not None:
+        if access is not None:
+            yield Property(name, access=access), ih.address
+        yield Property(name), ih.address
+
+
+def extract_insn_namespace_class_features(
+    fh: FunctionHandle, bh, ih: InsnHandle
+) -> Iterator[Tuple[Union[Namespace, Class], Address]]:
+    """parse instruction namespace and class features"""
+    type_: Optional[Union[DnType, DnUnmanagedMethod]] = None
+
+    if ih.inner.opcode in (
+        OpCodes.Call,
+        OpCodes.Callvirt,
+        OpCodes.Jmp,
+        OpCodes.Ldvirtftn,
+        OpCodes.Ldftn,
+        OpCodes.Newobj,
+    ):
+        # method call - includes managed methods (MethodDef, TypeRef) and properties (MethodSemantics, TypeRef)
+        type_ = get_callee(fh.ctx["pe"], fh.ctx["cache"], ih.inner.operand)
+
+    elif ih.inner.opcode in (
+        OpCodes.Ldfld,
+        OpCodes.Ldflda,
+        OpCodes.Ldsfld,
+        OpCodes.Ldsflda,
+        OpCodes.Stfld,
+        OpCodes.Stsfld,
+    ):
+        # field access
+        type_ = fh.ctx["cache"].get_field(ih.inner.operand.value)
+
+    # ECMA 335 VI.C.4.10
+    elif ih.inner.opcode in (
+        OpCodes.Initobj,
+        OpCodes.Box,
+        OpCodes.Castclass,
+        OpCodes.Cpobj,
+        OpCodes.Isinst,
+        OpCodes.Ldelem,
+        OpCodes.Ldelema,
+        OpCodes.Ldobj,
+        OpCodes.Mkrefany,
+        OpCodes.Newarr,
+        OpCodes.Refanyval,
+        OpCodes.Sizeof,
+        OpCodes.Stobj,
+        OpCodes.Unbox,
+        OpCodes.Constrained,
+        OpCodes.Stelem,
+        OpCodes.Unbox_Any,
+    ):
+        # type access
+        type_ = fh.ctx["cache"].get_type(ih.inner.operand.value)
+
+    if isinstance(type_, DnType):
+        yield Class(DnType.format_name(type_.class_, namespace=type_.namespace)), ih.address
+        if type_.namespace:
+            yield Namespace(type_.namespace), ih.address
+
+
+def extract_insn_number_features(fh, bh, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """parse instruction number features"""
+    if ih.inner.is_ldc():
+        yield Number(ih.inner.get_ldc()), ih.address
+
+
+def extract_insn_string_features(fh: FunctionHandle, bh, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """parse instruction string features"""
+    if not ih.inner.is_ldstr():
+        return
+
+    if not isinstance(ih.inner.operand, StringToken):
+        return
+
+    user_string: Optional[str] = read_dotnet_user_string(fh.ctx["pe"], ih.inner.operand)
+    if user_string is None:
+        return
+
+    if len(user_string) >= 4:
+        yield String(user_string), ih.address
+
+
+def extract_unmanaged_call_characteristic_features(
+    fh: FunctionHandle, bb: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Characteristic, Address]]:
+    if ih.inner.opcode not in (OpCodes.Call, OpCodes.Callvirt, OpCodes.Jmp):
+        return
+
+    row: Union[str, InvalidToken, dnfile.base.MDTableRow] = resolve_dotnet_token(fh.ctx["pe"], ih.inner.operand)
+    if not isinstance(row, dnfile.mdtable.MethodDefRow):
+        return
+
+    if any((row.Flags.mdPinvokeImpl, row.ImplFlags.miUnmanaged, row.ImplFlags.miNative)):
+        yield Characteristic("unmanaged call"), ih.address
+
+
+def extract_features(fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """extract instruction features"""
+    for inst_handler in INSTRUCTION_HANDLERS:
+        for feature, addr in inst_handler(fh, bbh, ih):
+            assert isinstance(addr, Address)
+            yield feature, addr
+
+
+INSTRUCTION_HANDLERS = (
+    extract_insn_api_features,
+    extract_insn_property_features,
+    extract_insn_number_features,
+    extract_insn_string_features,
+    extract_insn_namespace_class_features,
+    extract_unmanaged_call_characteristic_features,
+)

capa/features/extractors/dnfile/types.py

@@ -0,0 +1,75 @@
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
+# Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at: [package root]/LICENSE.txt
+# Unless required by applicable law or agreed to in writing, software distributed under the License
+#  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and limitations under the License.
+
+from enum import Enum
+from typing import Union, Optional
+
+
+class DnType(object):
+    def __init__(self, token: int, class_: str, namespace: str = "", member: str = "", access: Optional[str] = None):
+        self.token: int = token
+        self.access: Optional[str] = access
+        self.namespace: str = namespace
+        self.class_: str = class_
+
+        if member == ".ctor":
+            member = "ctor"
+        if member == ".cctor":
+            member = "cctor"
+
+        self.member: str = member
+
+    def __hash__(self):
+        return hash((self.token, self.access, self.namespace, self.class_, self.member))
+
+    def __eq__(self, other):
+        return (
+            self.token == other.token
+            and self.access == other.access
+            and self.namespace == other.namespace
+            and self.class_ == other.class_
+            and self.member == other.member
+        )
+
+    def __str__(self):
+        return DnType.format_name(self.class_, namespace=self.namespace, member=self.member)
+
+    def __repr__(self):
+        return str(self)
+
+    @staticmethod
+    def format_name(class_: str, namespace: str = "", member: str = ""):
+        # like File::OpenRead
+        name: str = f"{class_}::{member}" if member else class_
+        if namespace:
+            # like System.IO.File::OpenRead
+            name = f"{namespace}.{name}"
+        return name
+
+
+class DnUnmanagedMethod:
+    def __init__(self, token: int, module: str, method: str):
+        self.token: int = token
+        self.module: str = module
+        self.method: str = method
+
+    def __hash__(self):
+        return hash((self.token, self.module, self.method))
+
+    def __eq__(self, other):
+        return self.token == other.token and self.module == other.module and self.method == other.method
+
+    def __str__(self):
+        return DnUnmanagedMethod.format_name(self.module, self.method)
+
+    def __repr__(self):
+        return str(self)
+
+    @staticmethod
+    def format_name(module, method):
+        return f"{module}.{method}"

capa/features/extractors/dnfile_.py

@@ -0,0 +1,150 @@
+import logging
+from typing import Tuple, Iterator
+
+import dnfile
+import pefile
+
+from capa.features.common import (
+    OS,
+    OS_ANY,
+    ARCH_ANY,
+    ARCH_I386,
+    FORMAT_PE,
+    ARCH_AMD64,
+    FORMAT_DOTNET,
+    Arch,
+    Format,
+    Feature,
+)
+from capa.features.address import NO_ADDRESS, Address, AbsoluteVirtualAddress
+from capa.features.extractors.base_extractor import FeatureExtractor
+
+logger = logging.getLogger(__name__)
+
+
+def extract_file_format(**kwargs) -> Iterator[Tuple[Feature, Address]]:
+    yield Format(FORMAT_PE), NO_ADDRESS
+    yield Format(FORMAT_DOTNET), NO_ADDRESS
+
+
+def extract_file_os(**kwargs) -> Iterator[Tuple[Feature, Address]]:
+    yield OS(OS_ANY), NO_ADDRESS
+
+
+def extract_file_arch(pe: dnfile.dnPE, **kwargs) -> Iterator[Tuple[Feature, Address]]:
+    # to distinguish in more detail, see https://stackoverflow.com/a/23614024/10548020
+    # .NET 4.5 added option: any CPU, 32-bit preferred
+    assert pe.net is not None
+    assert pe.net.Flags is not None
+
+    if pe.net.Flags.CLR_32BITREQUIRED and pe.PE_TYPE == pefile.OPTIONAL_HEADER_MAGIC_PE:
+        yield Arch(ARCH_I386), NO_ADDRESS
+    elif not pe.net.Flags.CLR_32BITREQUIRED and pe.PE_TYPE == pefile.OPTIONAL_HEADER_MAGIC_PE_PLUS:
+        yield Arch(ARCH_AMD64), NO_ADDRESS
+    else:
+        yield Arch(ARCH_ANY), NO_ADDRESS
+
+
+def extract_file_features(pe: dnfile.dnPE) -> Iterator[Tuple[Feature, Address]]:
+    for file_handler in FILE_HANDLERS:
+        for feature, address in file_handler(pe=pe):  # type: ignore
+            yield feature, address
+
+
+FILE_HANDLERS = (
+    # extract_file_export_names,
+    # extract_file_import_names,
+    # extract_file_section_names,
+    # extract_file_strings,
+    # extract_file_function_names,
+    extract_file_format,
+)
+
+
+def extract_global_features(pe: dnfile.dnPE) -> Iterator[Tuple[Feature, Address]]:
+    for handler in GLOBAL_HANDLERS:
+        for feature, addr in handler(pe=pe):  # type: ignore
+            yield feature, addr
+
+
+GLOBAL_HANDLERS = (
+    extract_file_os,
+    extract_file_arch,
+)
+
+
+class DnfileFeatureExtractor(FeatureExtractor):
+    def __init__(self, path: str):
+        super().__init__()
+        self.path: str = path
+        self.pe: dnfile.dnPE = dnfile.dnPE(path)
+
+    def get_base_address(self) -> AbsoluteVirtualAddress:
+        return AbsoluteVirtualAddress(0x0)
+
+    def get_entry_point(self) -> int:
+        # self.pe.net.Flags.CLT_NATIVE_ENTRYPOINT
+        #  True: native EP: Token
+        #  False: managed EP: RVA
+        assert self.pe.net is not None
+        assert self.pe.net.struct is not None
+
+        return self.pe.net.struct.EntryPointTokenOrRva
+
+    def extract_global_features(self):
+        yield from extract_global_features(self.pe)
+
+    def extract_file_features(self):
+        yield from extract_file_features(self.pe)
+
+    def is_dotnet_file(self) -> bool:
+        return bool(self.pe.net)
+
+    def is_mixed_mode(self) -> bool:
+        assert self.pe is not None
+        assert self.pe.net is not None
+        assert self.pe.net.Flags is not None
+
+        return not bool(self.pe.net.Flags.CLR_ILONLY)
+
+    def get_runtime_version(self) -> Tuple[int, int]:
+        assert self.pe is not None
+        assert self.pe.net is not None
+        assert self.pe.net.struct is not None
+
+        return self.pe.net.struct.MajorRuntimeVersion, self.pe.net.struct.MinorRuntimeVersion
+
+    def get_meta_version_string(self) -> str:
+        assert self.pe.net is not None
+        assert self.pe.net.metadata is not None
+        assert self.pe.net.metadata.struct is not None
+        assert self.pe.net.metadata.struct.Version is not None
+
+        vbuf = self.pe.net.metadata.struct.Version
+        assert isinstance(vbuf, bytes)
+
+        return vbuf.rstrip(b"\x00").decode("utf-8")
+
+    def get_functions(self):
+        raise NotImplementedError("DnfileFeatureExtractor can only be used to extract file features")
+
+    def extract_function_features(self, f):
+        raise NotImplementedError("DnfileFeatureExtractor can only be used to extract file features")
+
+    def get_basic_blocks(self, f):
+        raise NotImplementedError("DnfileFeatureExtractor can only be used to extract file features")
+
+    def extract_basic_block_features(self, f, bb):
+        raise NotImplementedError("DnfileFeatureExtractor can only be used to extract file features")
+
+    def get_instructions(self, f, bb):
+        raise NotImplementedError("DnfileFeatureExtractor can only be used to extract file features")
+
+    def extract_insn_features(self, f, bb, insn):
+        raise NotImplementedError("DnfileFeatureExtractor can only be used to extract file features")
+
+    def is_library_function(self, va):
+        raise NotImplementedError("DnfileFeatureExtractor can only be used to extract file features")
+
+    def get_function_name(self, va):
+        raise NotImplementedError("DnfileFeatureExtractor can only be used to extract file features")

capa/features/extractors/dotnetfile.py

@@ -0,0 +1,231 @@
+import logging
+from typing import Tuple, Iterator, cast
+
+import dnfile
+import pefile
+
+import capa.features.extractors.helpers
+from capa.features.file import Import, FunctionName
+from capa.features.common import (
+    OS,
+    OS_ANY,
+    ARCH_ANY,
+    ARCH_I386,
+    FORMAT_PE,
+    ARCH_AMD64,
+    FORMAT_DOTNET,
+    Arch,
+    Class,
+    Format,
+    String,
+    Feature,
+    Namespace,
+    Characteristic,
+)
+from capa.features.address import NO_ADDRESS, Address, DNTokenAddress
+from capa.features.extractors.base_extractor import FeatureExtractor
+from capa.features.extractors.dnfile.helpers import (
+    DnType,
+    iter_dotnet_table,
+    is_dotnet_mixed_mode,
+    get_dotnet_managed_imports,
+    get_dotnet_managed_methods,
+    calculate_dotnet_token_value,
+    get_dotnet_unmanaged_imports,
+)
+
+logger = logging.getLogger(__name__)
+
+
+def extract_file_format(**kwargs) -> Iterator[Tuple[Format, Address]]:
+    yield Format(FORMAT_PE), NO_ADDRESS
+    yield Format(FORMAT_DOTNET), NO_ADDRESS
+
+
+def extract_file_import_names(pe: dnfile.dnPE, **kwargs) -> Iterator[Tuple[Import, Address]]:
+    for method in get_dotnet_managed_imports(pe):
+        # like System.IO.File::OpenRead
+        yield Import(str(method)), DNTokenAddress(method.token)
+
+    for imp in get_dotnet_unmanaged_imports(pe):
+        # like kernel32.CreateFileA
+        for name in capa.features.extractors.helpers.generate_symbols(imp.module, imp.method):
+            yield Import(name), DNTokenAddress(imp.token)
+
+
+def extract_file_function_names(pe: dnfile.dnPE, **kwargs) -> Iterator[Tuple[FunctionName, Address]]:
+    for method in get_dotnet_managed_methods(pe):
+        yield FunctionName(str(method)), DNTokenAddress(method.token)
+
+
+def extract_file_namespace_features(pe: dnfile.dnPE, **kwargs) -> Iterator[Tuple[Namespace, Address]]:
+    """emit namespace features from TypeRef and TypeDef tables"""
+
+    # namespaces may be referenced multiple times, so we need to filter
+    namespaces = set()
+
+    for _, typedef in iter_dotnet_table(pe, dnfile.mdtable.TypeDef.number):
+        # emit internal .NET namespaces
+        assert isinstance(typedef, dnfile.mdtable.TypeDefRow)
+        namespaces.add(typedef.TypeNamespace)
+
+    for _, typeref in iter_dotnet_table(pe, dnfile.mdtable.TypeRef.number):
+        # emit external .NET namespaces
+        assert isinstance(typeref, dnfile.mdtable.TypeRefRow)
+        namespaces.add(typeref.TypeNamespace)
+
+    # namespaces may be empty, discard
+    namespaces.discard("")
+
+    for namespace in namespaces:
+        # namespace do not have an associated token, so we yield 0x0
+        yield Namespace(namespace), NO_ADDRESS
+
+
+def extract_file_class_features(pe: dnfile.dnPE, **kwargs) -> Iterator[Tuple[Class, Address]]:
+    """emit class features from TypeRef and TypeDef tables"""
+    for rid, typedef in iter_dotnet_table(pe, dnfile.mdtable.TypeDef.number):
+        # emit internal .NET classes
+        assert isinstance(typedef, dnfile.mdtable.TypeDefRow)
+
+        token = calculate_dotnet_token_value(dnfile.mdtable.TypeDef.number, rid)
+        yield Class(DnType.format_name(typedef.TypeName, namespace=typedef.TypeNamespace)), DNTokenAddress(token)
+
+    for rid, typeref in iter_dotnet_table(pe, dnfile.mdtable.TypeRef.number):
+        # emit external .NET classes
+        assert isinstance(typeref, dnfile.mdtable.TypeRefRow)
+
+        token = calculate_dotnet_token_value(dnfile.mdtable.TypeRef.number, rid)
+        yield Class(DnType.format_name(typeref.TypeName, namespace=typeref.TypeNamespace)), DNTokenAddress(token)
+
+
+def extract_file_os(**kwargs) -> Iterator[Tuple[OS, Address]]:
+    yield OS(OS_ANY), NO_ADDRESS
+
+
+def extract_file_arch(pe: dnfile.dnPE, **kwargs) -> Iterator[Tuple[Arch, Address]]:
+    # to distinguish in more detail, see https://stackoverflow.com/a/23614024/10548020
+    # .NET 4.5 added option: any CPU, 32-bit preferred
+    assert pe.net is not None
+    assert pe.net.Flags is not None
+
+    if pe.net.Flags.CLR_32BITREQUIRED and pe.PE_TYPE == pefile.OPTIONAL_HEADER_MAGIC_PE:
+        yield Arch(ARCH_I386), NO_ADDRESS
+    elif not pe.net.Flags.CLR_32BITREQUIRED and pe.PE_TYPE == pefile.OPTIONAL_HEADER_MAGIC_PE_PLUS:
+        yield Arch(ARCH_AMD64), NO_ADDRESS
+    else:
+        yield Arch(ARCH_ANY), NO_ADDRESS
+
+
+def extract_file_strings(pe: dnfile.dnPE, **kwargs) -> Iterator[Tuple[String, Address]]:
+    yield from capa.features.extractors.common.extract_file_strings(pe.__data__)
+
+
+def extract_file_mixed_mode_characteristic_features(
+    pe: dnfile.dnPE, **kwargs
+) -> Iterator[Tuple[Characteristic, Address]]:
+    if is_dotnet_mixed_mode(pe):
+        yield Characteristic("mixed mode"), NO_ADDRESS
+
+
+def extract_file_features(pe: dnfile.dnPE) -> Iterator[Tuple[Feature, Address]]:
+    for file_handler in FILE_HANDLERS:
+        for feature, addr in file_handler(pe=pe):  # type: ignore
+            yield feature, addr
+
+
+FILE_HANDLERS = (
+    extract_file_import_names,
+    extract_file_function_names,
+    extract_file_strings,
+    extract_file_format,
+    extract_file_mixed_mode_characteristic_features,
+    extract_file_namespace_features,
+    extract_file_class_features,
+)
+
+
+def extract_global_features(pe: dnfile.dnPE) -> Iterator[Tuple[Feature, Address]]:
+    for handler in GLOBAL_HANDLERS:
+        for feature, va in handler(pe=pe):  # type: ignore
+            yield feature, va
+
+
+GLOBAL_HANDLERS = (
+    extract_file_os,
+    extract_file_arch,
+)
+
+
+class DotnetFileFeatureExtractor(FeatureExtractor):
+    def __init__(self, path: str):
+        super().__init__()
+        self.path: str = path
+        self.pe: dnfile.dnPE = dnfile.dnPE(path)
+
+    def get_base_address(self):
+        return NO_ADDRESS
+
+    def get_entry_point(self) -> int:
+        # self.pe.net.Flags.CLT_NATIVE_ENTRYPOINT
+        #  True: native EP: Token
+        #  False: managed EP: RVA
+        assert self.pe.net is not None
+        assert self.pe.net.struct is not None
+
+        return self.pe.net.struct.EntryPointTokenOrRva
+
+    def extract_global_features(self):
+        yield from extract_global_features(self.pe)
+
+    def extract_file_features(self):
+        yield from extract_file_features(self.pe)
+
+    def is_dotnet_file(self) -> bool:
+        return bool(self.pe.net)
+
+    def is_mixed_mode(self) -> bool:
+        return is_dotnet_mixed_mode(self.pe)
+
+    def get_runtime_version(self) -> Tuple[int, int]:
+        assert self.pe.net is not None
+        assert self.pe.net.struct is not None
+        assert self.pe.net.struct.MajorRuntimeVersion is not None
+        assert self.pe.net.struct.MinorRuntimeVersion is not None
+
+        return self.pe.net.struct.MajorRuntimeVersion, self.pe.net.struct.MinorRuntimeVersion
+
+    def get_meta_version_string(self) -> str:
+        assert self.pe.net is not None
+        assert self.pe.net.metadata is not None
+        assert self.pe.net.metadata.struct is not None
+        assert self.pe.net.metadata.struct.Version is not None
+
+        vbuf = self.pe.net.metadata.struct.Version
+        assert isinstance(vbuf, bytes)
+
+        return vbuf.rstrip(b"\x00").decode("utf-8")
+
+    def get_functions(self):
+        raise NotImplementedError("DotnetFileFeatureExtractor can only be used to extract file features")
+
+    def extract_function_features(self, f):
+        raise NotImplementedError("DotnetFileFeatureExtractor can only be used to extract file features")
+
+    def get_basic_blocks(self, f):
+        raise NotImplementedError("DotnetFileFeatureExtractor can only be used to extract file features")
+
+    def extract_basic_block_features(self, f, bb):
+        raise NotImplementedError("DotnetFileFeatureExtractor can only be used to extract file features")
+
+    def get_instructions(self, f, bb):
+        raise NotImplementedError("DotnetFileFeatureExtractor can only be used to extract file features")
+
+    def extract_insn_features(self, f, bb, insn):
+        raise NotImplementedError("DotnetFileFeatureExtractor can only be used to extract file features")
+
+    def is_library_function(self, va):
+        raise NotImplementedError("DotnetFileFeatureExtractor can only be used to extract file features")
+
+    def get_function_name(self, va):
+        raise NotImplementedError("DotnetFileFeatureExtractor can only be used to extract file features")

capa/features/extractors/elf.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -7,8 +7,11 @@
 # See the License for the specific language governing permissions and limitations under the License.
 import struct
 import logging
+import itertools
+import collections
 from enum import Enum
-from typing import BinaryIO
+from typing import Set, Dict, List, Tuple, BinaryIO, Iterator, Optional
+from dataclasses import dataclass
 
 logger = logging.getLogger(__name__)
 
@@ -21,6 +24,12 @@ def align(v, alignment):
         return v + (alignment - remainder)
 
 
+def read_cstr(buf, offset):
+    s = buf[offset:]
+    s, _, _ = s.partition(b"\x00")
+    return s.decode("utf-8")
+
+
 class CorruptElfFile(ValueError):
     pass
 
@@ -47,47 +56,107 @@ class OS(str, Enum):
     NACL = "nacl"
 
 
-def detect_elf_os(f: BinaryIO) -> str:
-    f.seek(0x0)
-    file_header = f.read(0x40)
+# via readelf: https://github.com/bminor/binutils-gdb/blob/c0e94211e1ac05049a4ce7c192c9d14d1764eb3e/binutils/readelf.c#L19635-L19658
+# and here: https://github.com/bminor/binutils-gdb/blob/34c54daa337da9fadf87d2706d6a590ae1f88f4d/include/elf/common.h#L933-L939
+GNU_ABI_TAG = {
+    0: OS.LINUX,
+    1: OS.HURD,
+    2: OS.SOLARIS,
+    3: OS.FREEBSD,
+    4: OS.NETBSD,
+    5: OS.SYLLABLE,
+    6: OS.NACL,
+}
 
-    # we'll set this to the detected OS
-    # prefer the first heuristics,
-    # but rather than short circuiting,
-    # we'll still parse out the remainder, for debugging.
-    ret = None
 
-    if not file_header.startswith(b"\x7fELF"):
-        raise CorruptElfFile("missing magic header")
+@dataclass
+class Phdr:
+    type: int
+    offset: int
+    vaddr: int
+    paddr: int
+    filesz: int
+    buf: bytes
 
-    ei_class, ei_data = struct.unpack_from("BB", file_header, 4)
-    logger.debug("ei_class: 0x%02x ei_data: 0x%02x", ei_class, ei_data)
-    if ei_class == 1:
-        bitness = 32
-    elif ei_class == 2:
-        bitness = 64
-    else:
-        raise CorruptElfFile("invalid ei_class: 0x%02x" % ei_class)
 
-    if ei_data == 1:
-        endian = "<"
-    elif ei_data == 2:
-        endian = ">"
-    else:
-        raise CorruptElfFile("not an ELF file: invalid ei_data: 0x%02x" % ei_data)
+@dataclass
+class Shdr:
+    name: int
+    type: int
+    flags: int
+    addr: int
+    offset: int
+    size: int
+    link: int
+    buf: bytes
 
-    if bitness == 32:
-        (e_phoff,) = struct.unpack_from(endian + "I", file_header, 0x1C)
-        e_phentsize, e_phnum = struct.unpack_from(endian + "HH", file_header, 0x2A)
-    elif bitness == 64:
-        (e_phoff,) = struct.unpack_from(endian + "Q", file_header, 0x20)
-        e_phentsize, e_phnum = struct.unpack_from(endian + "HH", file_header, 0x36)
-    else:
-        raise NotImplementedError()
 
-    logger.debug("e_phoff: 0x%02x e_phentsize: 0x%02x e_phnum: %d", e_phoff, e_phentsize, e_phnum)
+class ELF:
+    def __init__(self, f: BinaryIO):
+        self.f = f
+
+        # these will all be initialized in `_parse()`
+        self.bitness: int
+        self.endian: str
+        self.e_phentsize: int
+        self.e_phnum: int
+        self.e_shentsize: int
+        self.e_shnum: int
+        self.phbuf: bytes
+        self.shbuf: bytes
+
+        self._parse()
+
+    def _parse(self):
+        self.f.seek(0x0)
+        self.file_header = self.f.read(0x40)
+
+        if not self.file_header.startswith(b"\x7fELF"):
+            raise CorruptElfFile("missing magic header")
+
+        ei_class, ei_data = struct.unpack_from("BB", self.file_header, 4)
+        logger.debug("ei_class: 0x%02x ei_data: 0x%02x", ei_class, ei_data)
+        if ei_class == 1:
+            self.bitness = 32
+        elif ei_class == 2:
+            self.bitness = 64
+        else:
+            raise CorruptElfFile("invalid ei_class: 0x%02x" % ei_class)
+
+        if ei_data == 1:
+            self.endian = "<"
+        elif ei_data == 2:
+            self.endian = ">"
+        else:
+            raise CorruptElfFile("not an ELF file: invalid ei_data: 0x%02x" % ei_data)
+
+        if self.bitness == 32:
+            e_phoff, e_shoff = struct.unpack_from(self.endian + "II", self.file_header, 0x1C)
+            self.e_phentsize, self.e_phnum = struct.unpack_from(self.endian + "HH", self.file_header, 0x2A)
+            self.e_shentsize, self.e_shnum = struct.unpack_from(self.endian + "HH", self.file_header, 0x2E)
+        elif self.bitness == 64:
+            e_phoff, e_shoff = struct.unpack_from(self.endian + "QQ", self.file_header, 0x20)
+            self.e_phentsize, self.e_phnum = struct.unpack_from(self.endian + "HH", self.file_header, 0x36)
+            self.e_shentsize, self.e_shnum = struct.unpack_from(self.endian + "HH", self.file_header, 0x3A)
+        else:
+            raise NotImplementedError()
+
+        logger.debug("e_phoff: 0x%02x e_phentsize: 0x%02x e_phnum: %d", e_phoff, self.e_phentsize, self.e_phnum)
+
+        self.f.seek(e_phoff)
+        program_header_size = self.e_phnum * self.e_phentsize
+        self.phbuf = self.f.read(program_header_size)
+        if len(self.phbuf) != program_header_size:
+            logger.warning("failed to read program headers")
+            self.e_phnum = 0
+
+        self.f.seek(e_shoff)
+        section_header_size = self.e_shnum * self.e_shentsize
+        self.shbuf = self.f.read(section_header_size)
+        if len(self.shbuf) != section_header_size:
+            logger.warning("failed to read section headers")
+            self.e_shnum = 0
 
-    (ei_osabi,) = struct.unpack_from(endian + "B", file_header, 7)
     OSABI = {
         # via pyelftools: https://github.com/eliben/pyelftools/blob/0664de05ed2db3d39041e2d51d19622a8ef4fb0f/elftools/elf/enums.py#L35-L58
         # some candidates are commented out because the are not useful values,
@@ -115,162 +184,598 @@ def detect_elf_os(f: BinaryIO) -> str:
         # 97: "ARM",          # not an OS
         # 255: "STANDALONE",  # not an OS
     }
-    logger.debug("ei_osabi: 0x%02x (%s)", ei_osabi, OSABI.get(ei_osabi, "unknown"))
 
-    # os_osabi == 0 is commonly set even when the OS is not SYSV.
-    # other values are unused or unknown.
-    if ei_osabi in OSABI and ei_osabi != 0x0:
-        # subsequent strategies may overwrite this value
-        ret = OSABI[ei_osabi]
+    @property
+    def ei_osabi(self) -> Optional[OS]:
+        (ei_osabi,) = struct.unpack_from(self.endian + "B", self.file_header, 7)
+        return ELF.OSABI.get(ei_osabi)
 
-    f.seek(e_phoff)
-    program_header_size = e_phnum * e_phentsize
-    program_headers = f.read(program_header_size)
-    if len(program_headers) != program_header_size:
-        logger.warning("failed to read program headers")
-        e_phnum = 0
+    MACHINE = {
+        # via https://refspecs.linuxfoundation.org/elf/gabi4+/ch4.eheader.html
+        1: "M32",
+        2: "SPARC",
+        3: "i386",
+        4: "68K",
+        5: "88K",
+        6: "486",
+        7: "860",
+        8: "MIPS",
+        9: "S370",
+        10: "MIPS_RS3_LE",
+        11: "RS6000",
+        15: "PA_RISC",
+        16: "nCUBE",
+        17: "VPP500",
+        18: "SPARC32PLUS",
+        19: "960",
+        20: "PPC",
+        21: "PPC64",
+        22: "S390",
+        23: "SPU",
+        36: "V800",
+        37: "FR20",
+        38: "RH32",
+        39: "RCE",
+        40: "ARM",
+        41: "ALPHA",
+        42: "SH",
+        43: "SPARCV9",
+        44: "TRICORE",
+        45: "ARC",
+        46: "H8_300",
+        47: "H8_300H",
+        48: "H8S",
+        49: "H8_500",
+        50: "IA_64",
+        51: "MIPS_X",
+        52: "COLDFIRE",
+        53: "68HC12",
+        54: "MMA",
+        55: "PCP",
+        56: "NCPU",
+        57: "NDR1",
+        58: "STARCORE",
+        59: "ME16",
+        60: "ST100",
+        61: "TINYJ",
+        62: "amd64",
+        63: "PDSP",
+        64: "PDP10",
+        65: "PDP11",
+        66: "FX66",
+        67: "ST9PLUS",
+        68: "ST7",
+        69: "68HC16",
+        70: "68HC11",
+        71: "68HC08",
+        72: "68HC05",
+        73: "SVX",
+        74: "ST19",
+        75: "VAX",
+        76: "CRIS",
+        77: "JAVELIN",
+        78: "FIREPATH",
+        79: "ZSP",
+        80: "MMIX",
+        81: "HUANY",
+        82: "PRISM",
+        83: "AVR",
+        84: "FR30",
+        85: "D10V",
+        86: "D30V",
+        87: "V850",
+        88: "M32R",
+        89: "MN10300",
+        90: "MN10200",
+        91: "PJ",
+        92: "OPENRISC",
+        93: "ARC_A5",
+        94: "XTENSA",
+        95: "VIDEOCORE",
+        96: "TMM_GPP",
+        97: "NS32K",
+        98: "TPC",
+        99: "SNP1K",
+        100: "ST200",
+    }
 
-    # search for PT_NOTE sections that specify an OS
-    # for example, on Linux there is a GNU section with minimum kernel version
-    for i in range(e_phnum):
-        offset = i * e_phentsize
-        phent = program_headers[offset : offset + e_phentsize]
+    @property
+    def e_machine(self) -> Optional[str]:
+        (e_machine,) = struct.unpack_from(self.endian + "H", self.file_header, 0x12)
+        return ELF.MACHINE.get(e_machine)
 
-        PT_NOTE = 0x4
+    def parse_program_header(self, i) -> Phdr:
+        phent_offset = i * self.e_phentsize
+        phent = self.phbuf[phent_offset : phent_offset + self.e_phentsize]
 
-        (p_type,) = struct.unpack_from(endian + "I", phent, 0x0)
-        logger.debug("p_type: 0x%04x", p_type)
-        if p_type != PT_NOTE:
-            continue
+        (p_type,) = struct.unpack_from(self.endian + "I", phent, 0x0)
+        logger.debug("ph:p_type: 0x%04x", p_type)
 
-        if bitness == 32:
-            p_offset, _, _, p_filesz = struct.unpack_from(endian + "IIII", phent, 0x4)
-        elif bitness == 64:
-            p_offset, _, _, p_filesz = struct.unpack_from(endian + "QQQQ", phent, 0x8)
+        if self.bitness == 32:
+            p_offset, p_vaddr, p_paddr, p_filesz = struct.unpack_from(self.endian + "IIII", phent, 0x4)
+        elif self.bitness == 64:
+            p_offset, p_vaddr, p_paddr, p_filesz = struct.unpack_from(self.endian + "QQQQ", phent, 0x8)
         else:
             raise NotImplementedError()
 
-        logger.debug("p_offset: 0x%02x p_filesz: 0x%04x", p_offset, p_filesz)
+        logger.debug("ph:p_offset: 0x%02x p_filesz: 0x%04x", p_offset, p_filesz)
 
-        f.seek(p_offset)
-        note = f.read(p_filesz)
-        if len(note) != p_filesz:
-            logger.warning("failed to read note content")
-            continue
+        self.f.seek(p_offset)
+        buf = self.f.read(p_filesz)
+        if len(buf) != p_filesz:
+            raise ValueError("failed to read program header content")
 
-        namesz, descsz, type_ = struct.unpack_from(endian + "III", note, 0x0)
-        name_offset = 0xC
-        desc_offset = name_offset + align(namesz, 0x4)
+        return Phdr(p_type, p_offset, p_vaddr, p_paddr, p_filesz, buf)
 
-        logger.debug("namesz: 0x%02x descsz: 0x%02x type: 0x%04x", namesz, descsz, type_)
-
-        name = note[name_offset : name_offset + namesz].partition(b"\x00")[0].decode("ascii")
-        logger.debug("name: %s", name)
-
-        if type_ != 1:
-            continue
-
-        if name == "GNU":
-            if descsz < 16:
+    @property
+    def program_headers(self):
+        for i in range(self.e_phnum):
+            try:
+                yield self.parse_program_header(i)
+            except ValueError:
                 continue
 
-            desc = note[desc_offset : desc_offset + descsz]
-            abi_tag, kmajor, kminor, kpatch = struct.unpack_from(endian + "IIII", desc, 0x0)
-            # via readelf: https://github.com/bminor/binutils-gdb/blob/c0e94211e1ac05049a4ce7c192c9d14d1764eb3e/binutils/readelf.c#L19635-L19658
-            # and here: https://github.com/bminor/binutils-gdb/blob/34c54daa337da9fadf87d2706d6a590ae1f88f4d/include/elf/common.h#L933-L939
-            GNU_ABI_TAG = {
-                0: OS.LINUX,
-                1: OS.HURD,
-                2: OS.SOLARIS,
-                3: OS.FREEBSD,
-                4: OS.NETBSD,
-                5: OS.SYLLABLE,
-                6: OS.NACL,
-            }
-            logger.debug("GNU_ABI_TAG: 0x%02x", abi_tag)
+    def parse_section_header(self, i) -> Shdr:
+        shent_offset = i * self.e_shentsize
+        shent = self.shbuf[shent_offset : shent_offset + self.e_shentsize]
 
-            if abi_tag in GNU_ABI_TAG:
-                # update only if not set
-                # so we can get the debugging output of subsequent strategies
-                ret = GNU_ABI_TAG[abi_tag] if not ret else ret
-                logger.debug("abi tag: %s earliest compatible kernel: %d.%d.%d", ret, kmajor, kminor, kpatch)
-        elif name == "OpenBSD":
-            logger.debug("note owner: %s", "OPENBSD")
-            ret = OS.OPENBSD if not ret else ret
-        elif name == "NetBSD":
-            logger.debug("note owner: %s", "NETBSD")
-            ret = OS.NETBSD if not ret else ret
-        elif name == "FreeBSD":
-            logger.debug("note owner: %s", "FREEBSD")
-            ret = OS.FREEBSD if not ret else ret
-
-    # search for recognizable dynamic linkers (interpreters)
-    # for example, on linux, we see file paths like: /lib64/ld-linux-x86-64.so.2
-    for i in range(e_phnum):
-        offset = i * e_phentsize
-        phent = program_headers[offset : offset + e_phentsize]
-
-        PT_INTERP = 0x3
-
-        (p_type,) = struct.unpack_from(endian + "I", phent, 0x0)
-        if p_type != PT_INTERP:
-            continue
-
-        if bitness == 32:
-            p_offset, _, _, p_filesz = struct.unpack_from(endian + "IIII", phent, 0x4)
-        elif bitness == 64:
-            p_offset, _, _, p_filesz = struct.unpack_from(endian + "QQQQ", phent, 0x8)
+        if self.bitness == 32:
+            sh_name, sh_type, sh_flags, sh_addr, sh_offset, sh_size, sh_link = struct.unpack_from(
+                self.endian + "IIIIIII", shent, 0x0
+            )
+        elif self.bitness == 64:
+            sh_name, sh_type, sh_flags, sh_addr, sh_offset, sh_size, sh_link = struct.unpack_from(
+                self.endian + "IIQQQQI", shent, 0x0
+            )
         else:
             raise NotImplementedError()
 
-        f.seek(p_offset)
-        interp = f.read(p_filesz)
-        if len(interp) != p_filesz:
-            logger.warning("failed to read interp content")
+        logger.debug("sh:sh_offset: 0x%02x sh_size: 0x%04x", sh_offset, sh_size)
+
+        self.f.seek(sh_offset)
+        buf = self.f.read(sh_size)
+        if len(buf) != sh_size:
+            raise ValueError("failed to read section header content")
+
+        return Shdr(sh_name, sh_type, sh_flags, sh_addr, sh_offset, sh_size, sh_link, buf)
+
+    @property
+    def section_headers(self):
+        for i in range(self.e_shnum):
+            try:
+                yield self.parse_section_header(i)
+            except ValueError:
+                continue
+
+    @property
+    def linker(self):
+        PT_INTERP = 0x3
+        for phdr in self.program_headers:
+            if phdr.type != PT_INTERP:
+                continue
+
+            return read_cstr(phdr.buf, 0)
+
+    @property
+    def versions_needed(self) -> Dict[str, Set[str]]:
+        # symbol version requirements are stored in the .gnu.version_r section,
+        # which has type SHT_GNU_verneed (0x6ffffffe).
+        #
+        # this contains a linked list of ElfXX_Verneed structs,
+        # each referencing a linked list of ElfXX_Vernaux structs.
+        # strings are stored in the section referenced by the sh_link field of the section header.
+        # each Verneed struct contains a reference to the name of the library,
+        # each Vernaux struct contains a reference to the name of a symbol.
+        SHT_GNU_VERNEED = 0x6FFFFFFE
+        for shdr in self.section_headers:
+            if shdr.type != SHT_GNU_VERNEED:
+                continue
+
+            # the linked section contains strings referenced by the verneed structures.
+            linked_shdr = self.parse_section_header(shdr.link)
+
+            versions_needed = collections.defaultdict(set)
+
+            # read verneed structures from the start of the section
+            # until the vn_next link is 0x0.
+            # each entry describes a shared object that is required by this binary.
+            vn_offset = 0x0
+            while True:
+                # ElfXX_Verneed layout is the same on 32 and 64 bit
+                vn_version, vn_cnt, vn_file, vn_aux, vn_next = struct.unpack_from(
+                    self.endian + "HHIII", shdr.buf, vn_offset
+                )
+                if vn_version != 1:
+                    # unexpected format, don't try to keep parsing
+                    break
+
+                # shared object names, like: "libdl.so.2"
+                so_name = read_cstr(linked_shdr.buf, vn_file)
+
+                # read vernaux structures linked from the verneed structure.
+                # there should be vn_cnt of these.
+                # each entry describes an ABI name required by the shared object.
+                vna_offset = vn_offset + vn_aux
+                for i in range(vn_cnt):
+                    # ElfXX_Vernaux layout is the same on 32 and 64 bit
+                    _, _, _, vna_name, vna_next = struct.unpack_from(self.endian + "IHHII", shdr.buf, vna_offset)
+
+                    # ABI names, like: "GLIBC_2.2.5"
+                    abi = read_cstr(linked_shdr.buf, vna_name)
+                    versions_needed[so_name].add(abi)
+
+                    vna_offset += vna_next
+
+                vn_offset += vn_next
+                if vn_next == 0:
+                    break
+
+            return dict(versions_needed)
+
+        return {}
+
+    @property
+    def dynamic_entries(self) -> Iterator[Tuple[int, int]]:
+        """
+        read the entries from the dynamic section,
+        yielding the tag and value for each entry.
+        """
+        DT_NULL = 0x0
+        PT_DYNAMIC = 0x2
+        for phdr in self.program_headers:
+            if phdr.type != PT_DYNAMIC:
+                continue
+
+            offset = 0x0
+            while True:
+                if self.bitness == 32:
+                    d_tag, d_val = struct.unpack_from(self.endian + "II", phdr.buf, offset)
+                    offset += 8
+                elif self.bitness == 64:
+                    d_tag, d_val = struct.unpack_from(self.endian + "QQ", phdr.buf, offset)
+                    offset += 16
+                else:
+                    raise NotImplementedError()
+
+                if d_tag == DT_NULL:
+                    break
+
+                yield d_tag, d_val
+
+    @property
+    def strtab(self) -> Optional[bytes]:
+        """
+        fetch the bytes of the string table
+        referenced by the dynamic section.
+        """
+        DT_STRTAB = 0x5
+        DT_STRSZ = 0xA
+
+        strtab_addr = None
+        strtab_size = None
+
+        for d_tag, d_val in self.dynamic_entries:
+            if d_tag == DT_STRTAB:
+                strtab_addr = d_val
+
+        for d_tag, d_val in self.dynamic_entries:
+            if d_tag == DT_STRSZ:
+                strtab_size = d_val
+
+        if strtab_addr is None:
+            return None
+
+        if strtab_size is None:
+            return None
+
+        strtab_offset = None
+        for shdr in self.section_headers:
+            if shdr.addr <= strtab_addr < shdr.addr + shdr.size:
+                strtab_offset = shdr.offset + (strtab_addr - shdr.addr)
+
+        if strtab_offset is None:
+            return None
+
+        self.f.seek(strtab_offset)
+        strtab_buf = self.f.read(strtab_size)
+
+        if len(strtab_buf) != strtab_size:
+            return None
+
+        return strtab_buf
+
+    @property
+    def needed(self) -> Iterator[str]:
+        """
+        read the names of DT_NEEDED entries from the dynamic section,
+        which correspond to dependencies on other shared objects,
+        like: `libpthread.so.0`
+        """
+        DT_NEEDED = 0x1
+        strtab = self.strtab
+        if not strtab:
+            return
+
+        for d_tag, d_val in self.dynamic_entries:
+            if d_tag != DT_NEEDED:
+                continue
+
+            yield read_cstr(strtab, d_val)
+
+
+@dataclass
+class ABITag:
+    os: OS
+    kmajor: int
+    kminor: int
+    kpatch: int
+
+
+class PHNote:
+    def __init__(self, endian: str, buf: bytes):
+        self.endian = endian
+        self.buf = buf
+
+        # these will be initialized in `_parse()`
+        self.type_: int
+        self.descsz: int
+        self.name: str
+
+        self._parse()
+
+    def _parse(self):
+        namesz, self.descsz, self.type_ = struct.unpack_from(self.endian + "III", self.buf, 0x0)
+        name_offset = 0xC
+        self.desc_offset = name_offset + align(namesz, 0x4)
+
+        logger.debug("ph:namesz: 0x%02x descsz: 0x%02x type: 0x%04x", namesz, self.descsz, self.type_)
+
+        self.name = self.buf[name_offset : name_offset + namesz].partition(b"\x00")[0].decode("ascii")
+        logger.debug("name: %s", self.name)
+
+    @property
+    def abi_tag(self) -> Optional[ABITag]:
+        if self.type_ != 1:
+            # > The type field shall be 1.
+            # Linux Standard Base Specification 1.2
+            # ref: https://refspecs.linuxfoundation.org/LSB_1.2.0/gLSB/noteabitag.html
+            return None
+
+        if self.name != "GNU":
+            return None
+
+        if self.descsz < 16:
+            return None
+
+        desc = self.buf[self.desc_offset : self.desc_offset + self.descsz]
+        abi_tag, kmajor, kminor, kpatch = struct.unpack_from(self.endian + "IIII", desc, 0x0)
+        logger.debug("GNU_ABI_TAG: 0x%02x", abi_tag)
+
+        os = GNU_ABI_TAG.get(abi_tag)
+        if not os:
+            return None
+
+        logger.debug("abi tag: %s earliest compatible kernel: %d.%d.%d", os, kmajor, kminor, kpatch)
+
+        return ABITag(os, kmajor, kminor, kpatch)
+
+
+class SHNote:
+    def __init__(self, endian: str, buf: bytes):
+        self.endian = endian
+        self.buf = buf
+
+        # these will be initialized in `_parse()`
+        self.type_: int
+        self.descsz: int
+        self.name: str
+
+        self._parse()
+
+    def _parse(self):
+        namesz, self.descsz, self.type_ = struct.unpack_from(self.endian + "III", self.buf, 0x0)
+        name_offset = 0xC
+        self.desc_offset = name_offset + align(namesz, 0x4)
+
+        logger.debug("sh:namesz: 0x%02x descsz: 0x%02x type: 0x%04x", namesz, self.descsz, self.type_)
+
+        name_buf = self.buf[name_offset : name_offset + namesz]
+        self.name = read_cstr(name_buf, 0x0)
+        logger.debug("sh:name: %s", self.name)
+
+    @property
+    def abi_tag(self) -> Optional[ABITag]:
+        if self.name != "GNU":
+            return None
+
+        if self.descsz < 16:
+            return None
+
+        desc = self.buf[self.desc_offset : self.desc_offset + self.descsz]
+        abi_tag, kmajor, kminor, kpatch = struct.unpack_from(self.endian + "IIII", desc, 0x0)
+        logger.debug("GNU_ABI_TAG: 0x%02x", abi_tag)
+
+        os = GNU_ABI_TAG.get(abi_tag)
+        if not os:
+            return None
+
+        logger.debug("abi tag: %s earliest compatible kernel: %d.%d.%d", os, kmajor, kminor, kpatch)
+        return ABITag(os, kmajor, kminor, kpatch)
+
+
+def guess_os_from_osabi(elf) -> Optional[OS]:
+    return elf.ei_osabi
+
+
+def guess_os_from_ph_notes(elf) -> Optional[OS]:
+    # search for PT_NOTE sections that specify an OS
+    # for example, on Linux there is a GNU section with minimum kernel version
+    PT_NOTE = 0x4
+    for phdr in elf.program_headers:
+        if phdr.type != PT_NOTE:
             continue
 
-        linker = interp.partition(b"\x00")[0].decode("ascii")
-        logger.debug("linker: %s", linker)
-        if "ld-linux" in linker:
-            # update only if not set
-            # so we can get the debugging output of subsequent strategies
-            ret = OS.LINUX if ret is None else ret
+        note = PHNote(elf.endian, phdr.buf)
+
+        if note.type_ != 1:
+            # > The type field shall be 1.
+            # Linux Standard Base Specification 1.2
+            # ref: https://refspecs.linuxfoundation.org/LSB_1.2.0/gLSB/noteabitag.html
+            continue
+
+        if note.name == "Linux":
+            logger.debug("note owner: %s", "LINUX")
+            return OS.LINUX
+        elif note.name == "OpenBSD":
+            logger.debug("note owner: %s", "OPENBSD")
+            return OS.OPENBSD
+        elif note.name == "NetBSD":
+            logger.debug("note owner: %s", "NETBSD")
+            return OS.NETBSD
+        elif note.name == "FreeBSD":
+            logger.debug("note owner: %s", "FREEBSD")
+            return OS.FREEBSD
+        elif note.name == "GNU":
+            abi_tag = note.abi_tag
+            if abi_tag:
+                return abi_tag.os
+            else:
+                # cannot make a guess about the OS, but probably linux or hurd
+                pass
+
+    return None
+
+
+def guess_os_from_sh_notes(elf) -> Optional[OS]:
+    # search for notes stored in sections that aren't visible in program headers.
+    # e.g. .note.Linux in Linux kernel modules.
+    SHT_NOTE = 0x7
+    for shdr in elf.section_headers:
+        if shdr.type != SHT_NOTE:
+            continue
+
+        note = SHNote(elf.endian, shdr.buf)
+
+        if note.name == "Linux":
+            logger.debug("note owner: %s", "LINUX")
+            return OS.LINUX
+        elif note.name == "OpenBSD":
+            logger.debug("note owner: %s", "OPENBSD")
+            return OS.OPENBSD
+        elif note.name == "NetBSD":
+            logger.debug("note owner: %s", "NETBSD")
+            return OS.NETBSD
+        elif note.name == "FreeBSD":
+            logger.debug("note owner: %s", "FREEBSD")
+            return OS.FREEBSD
+        elif note.name == "GNU":
+            abi_tag = note.abi_tag
+            if abi_tag:
+                return abi_tag.os
+            else:
+                # cannot make a guess about the OS, but probably linux or hurd
+                pass
+
+    return None
+
+
+def guess_os_from_linker(elf) -> Optional[OS]:
+    # search for recognizable dynamic linkers (interpreters)
+    # for example, on linux, we see file paths like: /lib64/ld-linux-x86-64.so.2
+    linker = elf.linker
+    if linker and "ld-linux" in elf.linker:
+        return OS.LINUX
+
+    return None
+
+
+def guess_os_from_abi_versions_needed(elf) -> Optional[OS]:
+    # then lets look for GLIBC symbol versioning requirements.
+    # this will let us guess about linux/hurd in some cases.
+
+    versions_needed = elf.versions_needed
+    if any(map(lambda abi: abi.startswith("GLIBC"), itertools.chain(*versions_needed.values()))):
+        # there are any GLIBC versions needed
+
+        if elf.e_machine != "i386":
+            # GLIBC runs on Linux and Hurd.
+            # for Hurd, its *only* on i386.
+            # so if we're not on i386, then we're on Linux.
+            return OS.LINUX
+
+        else:
+            # we're on i386, so we could be on either Linux or Hurd.
+            linker = elf.linker
+
+            if linker and "ld-linux" in linker:
+                return OS.LINUX
+
+            elif linker and "/ld.so" in linker:
+                return OS.HURD
+
+            else:
+                # we don't have any good guesses based on versions needed
+                pass
+
+    return None
+
+
+def guess_os_from_needed_dependencies(elf) -> Optional[OS]:
+    for needed in elf.needed:
+        if needed.startswith("libmachuser.so"):
+            return OS.HURD
+        if needed.startswith("libhurduser.so"):
+            return OS.HURD
+
+    return None
+
+
+def detect_elf_os(f) -> str:
+    """
+    f: type Union[BinaryIO, IDAIO]
+    """
+    elf = ELF(f)
+
+    osabi_guess = guess_os_from_osabi(elf)
+    logger.debug("guess: osabi: %s", osabi_guess)
+
+    ph_notes_guess = guess_os_from_ph_notes(elf)
+    logger.debug("guess: ph notes: %s", ph_notes_guess)
+
+    sh_notes_guess = guess_os_from_sh_notes(elf)
+    logger.debug("guess: sh notes: %s", sh_notes_guess)
+
+    linker_guess = guess_os_from_linker(elf)
+    logger.debug("guess: linker: %s", linker_guess)
+
+    abi_versions_needed_guess = guess_os_from_abi_versions_needed(elf)
+    logger.debug("guess: ABI versions needed: %s", abi_versions_needed_guess)
+
+    needed_dependencies_guess = guess_os_from_needed_dependencies(elf)
+    logger.debug("guess: needed dependencies: %s", needed_dependencies_guess)
+
+    ret = None
+
+    if osabi_guess:
+        ret = osabi_guess
+
+    elif ph_notes_guess:
+        ret = ph_notes_guess
+
+    elif sh_notes_guess:
+        ret = sh_notes_guess
+
+    elif linker_guess:
+        ret = linker_guess
+
+    elif abi_versions_needed_guess:
+        ret = abi_versions_needed_guess
+
+    elif needed_dependencies_guess:
+        ret = needed_dependencies_guess
 
     return ret.value if ret is not None else "unknown"
 
 
-class Arch(str, Enum):
-    I386 = "i386"
-    AMD64 = "amd64"
-
-
 def detect_elf_arch(f: BinaryIO) -> str:
-    f.seek(0x0)
-    file_header = f.read(0x40)
-
-    if not file_header.startswith(b"\x7fELF"):
-        raise CorruptElfFile("missing magic header")
-
-    (ei_data,) = struct.unpack_from("B", file_header, 5)
-    logger.debug("ei_data: 0x%02x", ei_data)
-
-    if ei_data == 1:
-        endian = "<"
-    elif ei_data == 2:
-        endian = ">"
-    else:
-        raise CorruptElfFile("not an ELF file: invalid ei_data: 0x%02x" % ei_data)
-
-    (ei_machine,) = struct.unpack_from(endian + "H", file_header, 0x12)
-    logger.debug("ei_machine: 0x%02x", ei_machine)
-
-    EM_386 = 0x3
-    EM_X86_64 = 0x3E
-    if ei_machine == EM_386:
-        return Arch.I386
-    elif ei_machine == EM_X86_64:
-        return Arch.AMD64
-    else:
-        # not really unknown, but unsupport at the moment:
-        # https://github.com/eliben/pyelftools/blob/ab444d982d1849191e910299a985989857466620/elftools/elf/enums.py#L73
-        return "unknown"
+    return ELF(f).e_machine or "unknown"

capa/features/extractors/elffile.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -7,7 +7,6 @@
 # See the License for the specific language governing permissions and limitations under the License.
 import io
 import logging
-import contextlib
 from typing import Tuple, Iterator
 
 from elftools.elf.elffile import ELFFile, SymbolTableSection
@@ -15,7 +14,7 @@ from elftools.elf.elffile import ELFFile, SymbolTableSection
 import capa.features.extractors.common
 from capa.features.file import Import, Section
 from capa.features.common import OS, FORMAT_ELF, Arch, Format, Feature
-from capa.features.extractors.elf import Arch as ElfArch
+from capa.features.address import NO_ADDRESS, FileOffsetAddress, AbsoluteVirtualAddress
 from capa.features.extractors.base_extractor import FeatureExtractor
 
 logger = logging.getLogger(__name__)
@@ -25,29 +24,29 @@ def extract_file_import_names(elf, **kwargs):
     # see https://github.com/eliben/pyelftools/blob/0664de05ed2db3d39041e2d51d19622a8ef4fb0f/scripts/readelf.py#L372
     symbol_tables = [(idx, s) for idx, s in enumerate(elf.iter_sections()) if isinstance(s, SymbolTableSection)]
 
-    for section_index, section in symbol_tables:
+    for _, section in symbol_tables:
         if not isinstance(section, SymbolTableSection):
             continue
 
         if section["sh_entsize"] == 0:
-            logger.debug("Symbol table '%s' has a sh_entsize of zero!" % (section.name))
+            logger.debug("Symbol table '%s' has a sh_entsize of zero!", section.name)
             continue
 
-        logger.debug("Symbol table '%s' contains %s entries:" % (section.name, section.num_symbols()))
+        logger.debug("Symbol table '%s' contains %s entries:", section.name, section.num_symbols())
 
-        for nsym, symbol in enumerate(section.iter_symbols()):
+        for _, symbol in enumerate(section.iter_symbols()):
             if symbol.name and symbol.entry.st_info.type == "STT_FUNC":
                 # TODO symbol address
                 # TODO symbol version info?
-                yield Import(symbol.name), 0x0
+                yield Import(symbol.name), FileOffsetAddress(0x0)
 
 
 def extract_file_section_names(elf, **kwargs):
     for section in elf.iter_sections():
         if section.name:
-            yield Section(section.name), section.header.sh_addr
+            yield Section(section.name), AbsoluteVirtualAddress(section.header.sh_addr)
         elif section.is_null():
-            yield Section("NULL"), section.header.sh_addr
+            yield Section("NULL"), AbsoluteVirtualAddress(section.header.sh_addr)
 
 
 def extract_file_strings(buf, **kwargs):
@@ -58,31 +57,31 @@ def extract_file_os(elf, buf, **kwargs):
     # our current approach does not always get an OS value, e.g. for packed samples
     # for file limitation purposes, we're more lax here
     try:
-        os = next(capa.features.extractors.common.extract_os(buf))
-        yield os
+        os_tuple = next(capa.features.extractors.common.extract_os(buf))
+        yield os_tuple
     except StopIteration:
-        yield OS("unknown"), 0x0
+        yield OS("unknown"), NO_ADDRESS
 
 
 def extract_file_format(**kwargs):
-    yield Format(FORMAT_ELF), 0x0
+    yield Format(FORMAT_ELF), NO_ADDRESS
 
 
 def extract_file_arch(elf, **kwargs):
     # TODO merge with capa.features.extractors.elf.detect_elf_arch()
     arch = elf.get_machine_arch()
     if arch == "x86":
-        yield Arch(ElfArch.I386), 0x0
+        yield Arch("i386"), NO_ADDRESS
     elif arch == "x64":
-        yield Arch(ElfArch.AMD64), 0x0
+        yield Arch("amd64"), NO_ADDRESS
     else:
         logger.warning("unsupported architecture: %s", arch)
 
 
 def extract_file_features(elf: ELFFile, buf: bytes) -> Iterator[Tuple[Feature, int]]:
     for file_handler in FILE_HANDLERS:
-        for feature, va in file_handler(elf=elf, buf=buf):  # type: ignore
-            yield feature, va
+        for feature, addr in file_handler(elf=elf, buf=buf):  # type: ignore
+            yield feature, addr
 
 
 FILE_HANDLERS = (
@@ -97,8 +96,8 @@ FILE_HANDLERS = (
 
 def extract_global_features(elf: ELFFile, buf: bytes) -> Iterator[Tuple[Feature, int]]:
     for global_handler in GLOBAL_HANDLERS:
-        for feature, va in global_handler(elf=elf, buf=buf):  # type: ignore
-            yield feature, va
+        for feature, addr in global_handler(elf=elf, buf=buf):  # type: ignore
+            yield feature, addr
 
 
 GLOBAL_HANDLERS = (
@@ -109,7 +108,7 @@ GLOBAL_HANDLERS = (
 
 class ElfFeatureExtractor(FeatureExtractor):
     def __init__(self, path: str):
-        super(ElfFeatureExtractor, self).__init__()
+        super().__init__()
         self.path = path
         with open(self.path, "rb") as f:
             self.elf = ELFFile(io.BytesIO(f.read()))
@@ -118,21 +117,21 @@ class ElfFeatureExtractor(FeatureExtractor):
         # virtual address of the first segment with type LOAD
         for segment in self.elf.iter_segments():
             if segment.header.p_type == "PT_LOAD":
-                return segment.header.p_vaddr
+                return AbsoluteVirtualAddress(segment.header.p_vaddr)
 
     def extract_global_features(self):
         with open(self.path, "rb") as f:
             buf = f.read()
 
-        for feature, va in extract_global_features(self.elf, buf):
-            yield feature, va
+        for feature, addr in extract_global_features(self.elf, buf):
+            yield feature, addr
 
     def extract_file_features(self):
         with open(self.path, "rb") as f:
             buf = f.read()
 
-        for feature, va in extract_file_features(self.elf, buf):
-            yield feature, va
+        for feature, addr in extract_file_features(self.elf, buf):
+            yield feature, addr
 
     def get_functions(self):
         raise NotImplementedError("ElfFeatureExtractor can only be used to extract file features")
@@ -152,8 +151,8 @@ class ElfFeatureExtractor(FeatureExtractor):
     def extract_insn_features(self, f, bb, insn):
         raise NotImplementedError("ElfFeatureExtractor can only be used to extract file features")
 
-    def is_library_function(self, va):
+    def is_library_function(self, addr):
         raise NotImplementedError("ElfFeatureExtractor can only be used to extract file features")
 
-    def get_function_name(self, va):
+    def get_function_name(self, addr):
         raise NotImplementedError("ElfFeatureExtractor can only be used to extract file features")

capa/features/extractors/helpers.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -29,8 +29,7 @@ def is_aw_function(symbol: str) -> bool:
     if symbol[-1] not in ("A", "W"):
         return False
 
-    # second to last character should be lowercase letter
-    return "a" <= symbol[-2] <= "z" or "0" <= symbol[-2] <= "9"
+    return True
 
 
 def is_ordinal(symbol: str) -> bool:
@@ -52,6 +51,9 @@ def generate_symbols(dll: str, symbol: str) -> Iterator[str]:
       - CreateFileA
       - CreateFile
     """
+    # normalize dll name
+    dll = dll.lower()
+
     # kernel32.CreateFileA
     yield "%s.%s" % (dll, symbol)
 
@@ -110,7 +112,6 @@ def carve_pe(pbytes: bytes, offset: int = 0) -> Iterator[Tuple[int, int]]:
     todo = [(off, mzx, pex, key) for (off, mzx, pex, key) in todo if off != -1]
 
     while len(todo):
-
         off, mzx, pex, key = todo.pop()
 
         # The MZ header has one field we will check

capa/features/extractors/ida/basicblock.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -8,22 +8,21 @@
 
 import string
 import struct
+from typing import Tuple, Iterator
 
 import idaapi
 
 import capa.features.extractors.ida.helpers
-from capa.features.common import Characteristic
+from capa.features.common import Feature, Characteristic
+from capa.features.address import Address
 from capa.features.basicblock import BasicBlock
 from capa.features.extractors.ida import helpers
 from capa.features.extractors.helpers import MIN_STACKSTRING_LEN
+from capa.features.extractors.base_extractor import BBHandle, FunctionHandle
 
 
-def get_printable_len(op):
-    """Return string length if all operand bytes are ascii or utf16-le printable
-
-    args:
-        op (IDA op_t)
-    """
+def get_printable_len(op: idaapi.op_t) -> int:
+    """Return string length if all operand bytes are ascii or utf16-le printable"""
     op_val = capa.features.extractors.ida.helpers.mask_op_val(op)
 
     if op.dtype == idaapi.dt_byte:
@@ -37,12 +36,12 @@ def get_printable_len(op):
     else:
         raise ValueError("Unhandled operand data type 0x%x." % op.dtype)
 
-    def is_printable_ascii(chars):
-        return all(c < 127 and chr(c) in string.printable for c in chars)
+    def is_printable_ascii(chars_: bytes):
+        return all(c < 127 and chr(c) in string.printable for c in chars_)
 
-    def is_printable_utf16le(chars):
-        if all(c == 0x00 for c in chars[1::2]):
-            return is_printable_ascii(chars[::2])
+    def is_printable_utf16le(chars_: bytes):
+        if all(c == 0x00 for c in chars_[1::2]):
+            return is_printable_ascii(chars_[::2])
 
     if is_printable_ascii(chars):
         return idaapi.get_dtype_size(op.dtype)
@@ -53,12 +52,8 @@ def get_printable_len(op):
     return 0
 
 
-def is_mov_imm_to_stack(insn):
-    """verify instruction moves immediate onto stack
-
-    args:
-        insn (IDA insn_t)
-    """
+def is_mov_imm_to_stack(insn: idaapi.insn_t) -> bool:
+    """verify instruction moves immediate onto stack"""
     if insn.Op2.type != idaapi.o_imm:
         return False
 
@@ -71,14 +66,10 @@ def is_mov_imm_to_stack(insn):
     return True
 
 
-def bb_contains_stackstring(f, bb):
+def bb_contains_stackstring(f: idaapi.func_t, bb: idaapi.BasicBlock) -> bool:
     """check basic block for stackstring indicators
 
     true if basic block contains enough moves of constant bytes to the stack
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
     """
     count = 0
     for insn in capa.features.extractors.ida.helpers.get_instructions_in_range(bb.start_ea, bb.end_ea):
@@ -89,39 +80,24 @@ def bb_contains_stackstring(f, bb):
     return False
 
 
-def extract_bb_stackstring(f, bb):
-    """extract stackstring indicators from basic block
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-    """
-    if bb_contains_stackstring(f, bb):
-        yield Characteristic("stack string"), bb.start_ea
+def extract_bb_stackstring(fh: FunctionHandle, bbh: BBHandle) -> Iterator[Tuple[Feature, Address]]:
+    """extract stackstring indicators from basic block"""
+    if bb_contains_stackstring(fh.inner, bbh.inner):
+        yield Characteristic("stack string"), bbh.address
 
 
-def extract_bb_tight_loop(f, bb):
-    """extract tight loop indicators from a basic block
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-    """
-    if capa.features.extractors.ida.helpers.is_basic_block_tight_loop(bb):
-        yield Characteristic("tight loop"), bb.start_ea
+def extract_bb_tight_loop(fh: FunctionHandle, bbh: BBHandle) -> Iterator[Tuple[Feature, Address]]:
+    """extract tight loop indicators from a basic block"""
+    if capa.features.extractors.ida.helpers.is_basic_block_tight_loop(bbh.inner):
+        yield Characteristic("tight loop"), bbh.address
 
 
-def extract_features(f, bb):
-    """extract basic block features
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-    """
+def extract_features(fh: FunctionHandle, bbh: BBHandle) -> Iterator[Tuple[Feature, Address]]:
+    """extract basic block features"""
     for bb_handler in BASIC_BLOCK_HANDLERS:
-        for (feature, ea) in bb_handler(f, bb):
-            yield feature, ea
-    yield BasicBlock(), bb.start_ea
+        for feature, addr in bb_handler(fh, bbh):
+            yield feature, addr
+    yield BasicBlock(), bbh.address
 
 
 BASIC_BLOCK_HANDLERS = (
@@ -132,9 +108,10 @@ BASIC_BLOCK_HANDLERS = (
 
 def main():
     features = []
-    for f in helpers.get_functions(skip_thunks=True, skip_libs=True):
+    for fhandle in helpers.get_functions(skip_thunks=True, skip_libs=True):
+        f: idaapi.func_t = fhandle.inner
         for bb in idaapi.FlowChart(f, flags=idaapi.FC_PREDS):
-            features.extend(list(extract_features(f, bb)))
+            features.extend(list(extract_features(fhandle, bb)))
 
     import pprint
 

capa/features/extractors/ida/extractor.py

@@ -1,10 +1,12 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
+from typing import List, Tuple, Iterator
+
 import idaapi
 
 import capa.ida.helpers
@@ -14,57 +16,21 @@ import capa.features.extractors.ida.insn
 import capa.features.extractors.ida.global_
 import capa.features.extractors.ida.function
 import capa.features.extractors.ida.basicblock
-from capa.features.extractors.base_extractor import FeatureExtractor
-
-
-class FunctionHandle:
-    """this acts like an idaapi.func_t but with __int__()"""
-
-    def __init__(self, inner):
-        self._inner = inner
-
-    def __int__(self):
-        return self.start_ea
-
-    def __getattr__(self, name):
-        return getattr(self._inner, name)
-
-
-class BasicBlockHandle:
-    """this acts like an idaapi.BasicBlock but with __int__()"""
-
-    def __init__(self, inner):
-        self._inner = inner
-
-    def __int__(self):
-        return self.start_ea
-
-    def __getattr__(self, name):
-        return getattr(self._inner, name)
-
-
-class InstructionHandle:
-    """this acts like an idaapi.insn_t but with __int__()"""
-
-    def __init__(self, inner):
-        self._inner = inner
-
-    def __int__(self):
-        return self.ea
-
-    def __getattr__(self, name):
-        return getattr(self._inner, name)
+from capa.features.common import Feature
+from capa.features.address import Address, AbsoluteVirtualAddress
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle, FeatureExtractor
 
 
 class IdaFeatureExtractor(FeatureExtractor):
     def __init__(self):
-        super(IdaFeatureExtractor, self).__init__()
-        self.global_features = []
+        super().__init__()
+        self.global_features: List[Tuple[Feature, Address]] = []
+        self.global_features.extend(capa.features.extractors.ida.file.extract_file_format())
         self.global_features.extend(capa.features.extractors.ida.global_.extract_os())
         self.global_features.extend(capa.features.extractors.ida.global_.extract_arch())
 
     def get_base_address(self):
-        return idaapi.get_imagebase()
+        return AbsoluteVirtualAddress(idaapi.get_imagebase())
 
     def extract_global_features(self):
         yield from self.global_features
@@ -72,41 +38,34 @@ class IdaFeatureExtractor(FeatureExtractor):
     def extract_file_features(self):
         yield from capa.features.extractors.ida.file.extract_features()
 
-    def get_functions(self):
+    def get_functions(self) -> Iterator[FunctionHandle]:
         import capa.features.extractors.ida.helpers as ida_helpers
 
-        # data structure shared across functions yielded here.
-        # useful for caching analysis relevant across a single workspace.
-        ctx = {}
-
         # ignore library functions and thunk functions as identified by IDA
-        for f in ida_helpers.get_functions(skip_thunks=True, skip_libs=True):
-            setattr(f, "ctx", ctx)
-            yield FunctionHandle(f)
+        yield from ida_helpers.get_functions(skip_thunks=True, skip_libs=True)
 
     @staticmethod
-    def get_function(ea):
+    def get_function(ea: int) -> FunctionHandle:
         f = idaapi.get_func(ea)
-        setattr(f, "ctx", {})
-        return FunctionHandle(f)
+        return FunctionHandle(address=AbsoluteVirtualAddress(f.start_ea), inner=f)
 
-    def extract_function_features(self, f):
-        yield from capa.features.extractors.ida.function.extract_features(f)
+    def extract_function_features(self, fh: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.ida.function.extract_features(fh)
 
-    def get_basic_blocks(self, f):
+    def get_basic_blocks(self, fh: FunctionHandle) -> Iterator[BBHandle]:
         import capa.features.extractors.ida.helpers as ida_helpers
 
-        for bb in ida_helpers.get_function_blocks(f):
-            yield BasicBlockHandle(bb)
+        for bb in ida_helpers.get_function_blocks(fh.inner):
+            yield BBHandle(address=AbsoluteVirtualAddress(bb.start_ea), inner=bb)
 
-    def extract_basic_block_features(self, f, bb):
-        yield from capa.features.extractors.ida.basicblock.extract_features(f, bb)
+    def extract_basic_block_features(self, fh: FunctionHandle, bbh: BBHandle) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.ida.basicblock.extract_features(fh, bbh)
 
-    def get_instructions(self, f, bb):
+    def get_instructions(self, fh: FunctionHandle, bbh: BBHandle) -> Iterator[InsnHandle]:
         import capa.features.extractors.ida.helpers as ida_helpers
 
-        for insn in ida_helpers.get_instructions_in_range(bb.start_ea, bb.end_ea):
-            yield InstructionHandle(insn)
+        for insn in ida_helpers.get_instructions_in_range(bbh.inner.start_ea, bbh.inner.end_ea):
+            yield InsnHandle(address=AbsoluteVirtualAddress(insn.ea), inner=insn)
 
-    def extract_insn_features(self, f, bb, insn):
-        yield from capa.features.extractors.ida.insn.extract_features(f, bb, insn)
+    def extract_insn_features(self, fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle):
+        yield from capa.features.extractors.ida.insn.extract_features(fh, bbh, ih)

capa/features/extractors/ida/file.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -7,27 +7,26 @@
 # See the License for the specific language governing permissions and limitations under the License.
 
 import struct
+from typing import Tuple, Iterator
 
 import idc
 import idaapi
 import idautils
-import ida_loader
 
+import capa.features.extractors.common
 import capa.features.extractors.helpers
 import capa.features.extractors.strings
 import capa.features.extractors.ida.helpers
 from capa.features.file import Export, Import, Section, FunctionName
-from capa.features.common import OS, FORMAT_PE, FORMAT_ELF, OS_WINDOWS, Format, String, Characteristic
+from capa.features.common import FORMAT_PE, FORMAT_ELF, Format, String, Feature, Characteristic
+from capa.features.address import NO_ADDRESS, Address, FileOffsetAddress, AbsoluteVirtualAddress
 
 
-def check_segment_for_pe(seg):
+def check_segment_for_pe(seg: idaapi.segment_t) -> Iterator[Tuple[int, int]]:
     """check segment for embedded PE
 
     adapted for IDA from:
     https://github.com/vivisect/vivisect/blob/7be4037b1cecc4551b397f840405a1fc606f9b53/PE/carve.py#L19
-
-    args:
-        seg (IDA segment_t)
     """
     seg_max = seg.end_ea
     mz_xor = [
@@ -40,7 +39,7 @@ def check_segment_for_pe(seg):
     ]
 
     todo = []
-    for (mzx, pex, i) in mz_xor:
+    for mzx, pex, i in mz_xor:
         for off in capa.features.extractors.ida.helpers.find_byte_sequence(seg.start_ea, seg.end_ea, mzx):
             todo.append((off, mzx, pex, i))
 
@@ -60,13 +59,13 @@ def check_segment_for_pe(seg):
             continue
 
         if idc.get_bytes(peoff, 2) == pex:
-            yield (off, i)
+            yield off, i
 
         for nextres in capa.features.extractors.ida.helpers.find_byte_sequence(off + 1, seg.end_ea, mzx):
             todo.append((nextres, mzx, pex, i))
 
 
-def extract_file_embedded_pe():
+def extract_file_embedded_pe() -> Iterator[Tuple[Feature, Address]]:
     """extract embedded PE features
 
     IDA must load resource sections for this to be complete
@@ -74,17 +73,17 @@ def extract_file_embedded_pe():
         - Check 'Load resource sections' when opening binary in IDA manually
     """
     for seg in capa.features.extractors.ida.helpers.get_segments(skip_header_segments=True):
-        for (ea, _) in check_segment_for_pe(seg):
-            yield Characteristic("embedded pe"), ea
+        for ea, _ in check_segment_for_pe(seg):
+            yield Characteristic("embedded pe"), FileOffsetAddress(ea)
 
 
-def extract_file_export_names():
+def extract_file_export_names() -> Iterator[Tuple[Feature, Address]]:
     """extract function exports"""
-    for (_, _, ea, name) in idautils.Entries():
-        yield Export(name), ea
+    for _, _, ea, name in idautils.Entries():
+        yield Export(name), AbsoluteVirtualAddress(ea)
 
 
-def extract_file_import_names():
+def extract_file_import_names() -> Iterator[Tuple[Feature, Address]]:
     """extract function imports
 
     1. imports by ordinal:
@@ -95,12 +94,13 @@ def extract_file_import_names():
      - modulename.importname
      - importname
     """
-    for (ea, info) in capa.features.extractors.ida.helpers.get_file_imports().items():
+    for ea, info in capa.features.extractors.ida.helpers.get_file_imports().items():
+        addr = AbsoluteVirtualAddress(ea)
         if info[1] and info[2]:
             # e.g. in mimikatz: ('cabinet', 'FCIAddFile', 11L)
             # extract by name here and by ordinal below
             for name in capa.features.extractors.helpers.generate_symbols(info[0], info[1]):
-                yield Import(name), ea
+                yield Import(name), addr
             dll = info[0]
             symbol = "#%d" % (info[2])
         elif info[1]:
@@ -113,10 +113,13 @@ def extract_file_import_names():
             continue
 
         for name in capa.features.extractors.helpers.generate_symbols(dll, symbol):
-            yield Import(name), ea
+            yield Import(name), addr
+
+    for ea, info in capa.features.extractors.ida.helpers.get_file_externs().items():
+        yield Import(info[1]), AbsoluteVirtualAddress(ea)
 
 
-def extract_file_section_names():
+def extract_file_section_names() -> Iterator[Tuple[Feature, Address]]:
     """extract section names
 
     IDA must load resource sections for this to be complete
@@ -124,10 +127,10 @@ def extract_file_section_names():
         - Check 'Load resource sections' when opening binary in IDA manually
     """
     for seg in capa.features.extractors.ida.helpers.get_segments(skip_header_segments=True):
-        yield Section(idaapi.get_segm_name(seg)), seg.start_ea
+        yield Section(idaapi.get_segm_name(seg)), AbsoluteVirtualAddress(seg.start_ea)
 
 
-def extract_file_strings():
+def extract_file_strings() -> Iterator[Tuple[Feature, Address]]:
     """extract ASCII and UTF-16 LE strings
 
     IDA must load resource sections for this to be complete
@@ -137,41 +140,50 @@ def extract_file_strings():
     for seg in capa.features.extractors.ida.helpers.get_segments():
         seg_buff = capa.features.extractors.ida.helpers.get_segment_buffer(seg)
 
+        # differing to common string extractor factor in segment offset here
         for s in capa.features.extractors.strings.extract_ascii_strings(seg_buff):
-            yield String(s.s), (seg.start_ea + s.offset)
+            yield String(s.s), FileOffsetAddress(seg.start_ea + s.offset)
 
         for s in capa.features.extractors.strings.extract_unicode_strings(seg_buff):
-            yield String(s.s), (seg.start_ea + s.offset)
+            yield String(s.s), FileOffsetAddress(seg.start_ea + s.offset)
 
 
-def extract_file_function_names():
+def extract_file_function_names() -> Iterator[Tuple[Feature, Address]]:
     """
     extract the names of statically-linked library functions.
     """
     for ea in idautils.Functions():
+        addr = AbsoluteVirtualAddress(ea)
         if idaapi.get_func(ea).flags & idaapi.FUNC_LIB:
             name = idaapi.get_name(ea)
-            yield FunctionName(name), ea
+            yield FunctionName(name), addr
+            if name.startswith("_"):
+                # some linkers may prefix linked routines with a `_` to avoid name collisions.
+                # extract features for both the mangled and un-mangled representations.
+                # e.g. `_fwrite` -> `fwrite`
+                # see: https://stackoverflow.com/a/2628384/87207
+                yield FunctionName(name[1:]), addr
 
 
-def extract_file_format():
-    format_name = ida_loader.get_file_type_name()
+def extract_file_format() -> Iterator[Tuple[Feature, Address]]:
+    file_info = idaapi.get_inf_structure()
 
-    if "PE" in format_name:
-        yield Format(FORMAT_PE), 0x0
-    elif "ELF64" in format_name:
-        yield Format(FORMAT_ELF), 0x0
-    elif "ELF32" in format_name:
-        yield Format(FORMAT_ELF), 0x0
+    if file_info.filetype in (idaapi.f_PE, idaapi.f_COFF):
+        yield Format(FORMAT_PE), NO_ADDRESS
+    elif file_info.filetype == idaapi.f_ELF:
+        yield Format(FORMAT_ELF), NO_ADDRESS
+    elif file_info.filetype == idaapi.f_BIN:
+        # no file type to return when processing a binary file, but we want to continue processing
+        return
     else:
-        raise NotImplementedError("file format: %s", format_name)
+        raise NotImplementedError("unexpected file format: %d" % file_info.filetype)
 
 
-def extract_features():
+def extract_features() -> Iterator[Tuple[Feature, Address]]:
     """extract file features"""
     for file_handler in FILE_HANDLERS:
-        for feature, va in file_handler():
-            yield feature, va
+        for feature, addr in file_handler():
+            yield feature, addr
 
 
 FILE_HANDLERS = (

capa/features/extractors/ida/function.py

@@ -1,35 +1,31 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
+from typing import Tuple, Iterator
 
 import idaapi
 import idautils
 
 import capa.features.extractors.ida.helpers
-from capa.features.common import Characteristic
+from capa.features.common import Feature, Characteristic
+from capa.features.address import Address, AbsoluteVirtualAddress
 from capa.features.extractors import loops
+from capa.features.extractors.base_extractor import FunctionHandle
 
 
-def extract_function_calls_to(f):
-    """extract callers to a function
-
-    args:
-        f (IDA func_t)
-    """
-    for ea in idautils.CodeRefsTo(f.start_ea, True):
-        yield Characteristic("calls to"), ea
+def extract_function_calls_to(fh: FunctionHandle):
+    """extract callers to a function"""
+    for ea in idautils.CodeRefsTo(fh.inner.start_ea, True):
+        yield Characteristic("calls to"), AbsoluteVirtualAddress(ea)
 
 
-def extract_function_loop(f):
-    """extract loop indicators from a function
-
-    args:
-        f (IDA func_t)
-    """
+def extract_function_loop(fh: FunctionHandle):
+    """extract loop indicators from a function"""
+    f: idaapi.func_t = fh.inner
     edges = []
 
     # construct control flow graph
@@ -38,28 +34,19 @@ def extract_function_loop(f):
             edges.append((bb.start_ea, succ.start_ea))
 
     if loops.has_loop(edges):
-        yield Characteristic("loop"), f.start_ea
+        yield Characteristic("loop"), fh.address
 
 
-def extract_recursive_call(f):
-    """extract recursive function call
-
-    args:
-        f (IDA func_t)
-    """
-    if capa.features.extractors.ida.helpers.is_function_recursive(f):
-        yield Characteristic("recursive call"), f.start_ea
+def extract_recursive_call(fh: FunctionHandle):
+    """extract recursive function call"""
+    if capa.features.extractors.ida.helpers.is_function_recursive(fh.inner):
+        yield Characteristic("recursive call"), fh.address
 
 
-def extract_features(f):
-    """extract function features
-
-    arg:
-        f (IDA func_t)
-    """
+def extract_features(fh: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
     for func_handler in FUNCTION_HANDLERS:
-        for (feature, ea) in func_handler(f):
-            yield feature, ea
+        for feature, addr in func_handler(fh):
+            yield feature, addr
 
 
 FUNCTION_HANDLERS = (extract_function_calls_to, extract_function_loop, extract_recursive_call)
@@ -68,8 +55,8 @@ FUNCTION_HANDLERS = (extract_function_calls_to, extract_function_loop, extract_r
 def main():
     """ """
     features = []
-    for f in capa.features.extractors.ida.get_functions(skip_thunks=True, skip_libs=True):
-        features.extend(list(extract_features(f)))
+    for fhandle in capa.features.extractors.ida.helpers.get_functions(skip_thunks=True, skip_libs=True):
+        features.extend(list(extract_features(fhandle)))
 
     import pprint
 

capa/features/extractors/ida/global_.py

@@ -1,27 +1,29 @@
 import logging
 import contextlib
+from typing import Tuple, Iterator
 
 import idaapi
 import ida_loader
 
 import capa.ida.helpers
 import capa.features.extractors.elf
-from capa.features.common import OS, ARCH_I386, ARCH_AMD64, OS_WINDOWS, Arch
+from capa.features.common import OS, ARCH_I386, ARCH_AMD64, OS_WINDOWS, Arch, Feature
+from capa.features.address import NO_ADDRESS, Address
 
 logger = logging.getLogger(__name__)
 
 
-def extract_os():
-    format_name = ida_loader.get_file_type_name()
+def extract_os() -> Iterator[Tuple[Feature, Address]]:
+    format_name: str = ida_loader.get_file_type_name()
 
     if "PE" in format_name:
-        yield OS(OS_WINDOWS), 0x0
+        yield OS(OS_WINDOWS), NO_ADDRESS
 
     elif "ELF" in format_name:
         with contextlib.closing(capa.ida.helpers.IDAIO()) as f:
             os = capa.features.extractors.elf.detect_elf_os(f)
 
-        yield OS(os), 0x0
+        yield OS(os), NO_ADDRESS
 
     else:
         # we likely end up here:
@@ -29,7 +31,7 @@ def extract_os():
         #  2. handling a new file format (e.g. macho)
         #
         # for (1) we can't do much - its shellcode and all bets are off.
-        # we could maybe accept a futher CLI argument to specify the OS,
+        # we could maybe accept a further CLI argument to specify the OS,
         # but i think this would be rarely used.
         # rules that rely on OS conditions will fail to match on shellcode.
         #
@@ -38,13 +40,13 @@ def extract_os():
         return
 
 
-def extract_arch():
-    info = idaapi.get_inf_structure()
-    if info.procName == "metapc" and info.is_64bit():
-        yield Arch(ARCH_AMD64), 0x0
-    elif info.procName == "metapc" and info.is_32bit():
-        yield Arch(ARCH_I386), 0x0
-    elif info.procName == "metapc":
+def extract_arch() -> Iterator[Tuple[Feature, Address]]:
+    info: idaapi.idainfo = idaapi.get_inf_structure()
+    if info.procname == "metapc" and info.is_64bit():
+        yield Arch(ARCH_AMD64), NO_ADDRESS
+    elif info.procname == "metapc" and info.is_32bit():
+        yield Arch(ARCH_I386), NO_ADDRESS
+    elif info.procname == "metapc":
         logger.debug("unsupported architecture: non-32-bit nor non-64-bit intel")
         return
     else:
@@ -52,5 +54,5 @@ def extract_arch():
         #  1. handling a new architecture (e.g. aarch64)
         #
         # for (1), this logic will need to be updated as the format is implemented.
-        logger.debug("unsupported architecture: %s", info.procName)
+        logger.debug("unsupported architecture: %s", info.procname)
         return

capa/features/extractors/ida/helpers.py

@@ -1,18 +1,23 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
+from typing import Any, Dict, Tuple, Iterator, Optional
 
 import idc
 import idaapi
 import idautils
 import ida_bytes
+import ida_segment
+
+from capa.features.address import AbsoluteVirtualAddress
+from capa.features.extractors.base_extractor import FunctionHandle
 
 
-def find_byte_sequence(start, end, seq):
+def find_byte_sequence(start: int, end: int, seq: bytes) -> Iterator[int]:
     """yield all ea of a given byte sequence
 
     args:
@@ -20,32 +25,32 @@ def find_byte_sequence(start, end, seq):
         end: max virtual address
         seq: bytes to search e.g. b"\x01\x03"
     """
-    seq = " ".join(["%02x" % b for b in seq])
+    seqstr = " ".join(["%02x" % b for b in seq])
     while True:
-        ea = idaapi.find_binary(start, end, seq, 0, idaapi.SEARCH_DOWN)
+        # TODO find_binary: Deprecated. Please use ida_bytes.bin_search() instead.
+        ea = idaapi.find_binary(start, end, seqstr, 0, idaapi.SEARCH_DOWN)
         if ea == idaapi.BADADDR:
             break
         start = ea + 1
         yield ea
 
 
-def get_functions(start=None, end=None, skip_thunks=False, skip_libs=False):
+def get_functions(
+    start: Optional[int] = None, end: Optional[int] = None, skip_thunks: bool = False, skip_libs: bool = False
+) -> Iterator[FunctionHandle]:
     """get functions, range optional
 
     args:
         start: min virtual address
         end: max virtual address
-
-    ret:
-        yield func_t*
     """
     for ea in idautils.Functions(start=start, end=end):
         f = idaapi.get_func(ea)
         if not (skip_thunks and (f.flags & idaapi.FUNC_THUNK) or skip_libs and (f.flags & idaapi.FUNC_LIB)):
-            yield f
+            yield FunctionHandle(address=AbsoluteVirtualAddress(ea), inner=f)
 
 
-def get_segments(skip_header_segments=False):
+def get_segments(skip_header_segments=False) -> Iterator[idaapi.segment_t]:
     """get list of segments (sections) in the binary image
 
     args:
@@ -57,7 +62,7 @@ def get_segments(skip_header_segments=False):
             yield seg
 
 
-def get_segment_buffer(seg):
+def get_segment_buffer(seg: idaapi.segment_t) -> bytes:
     """return bytes stored in a given segment
 
     decrease buffer size until IDA is able to read bytes from the segment
@@ -75,7 +80,7 @@ def get_segment_buffer(seg):
     return buff if buff else b""
 
 
-def get_file_imports():
+def get_file_imports() -> Dict[int, Tuple[str, str, int]]:
     """get file imports"""
     imports = {}
 
@@ -85,10 +90,18 @@ def get_file_imports():
         if not library:
             continue
 
+        # IDA uses section names for the library of ELF imports, like ".dynsym"
+        library = library.lstrip(".")
+
         def inspect_import(ea, function, ordinal):
             if function and function.startswith("__imp_"):
-                # handle mangled names starting
+                # handle mangled PE imports
                 function = function[len("__imp_") :]
+
+            if function and "@@" in function:
+                # handle mangled ELF imports, like "fopen@@glibc_2.2.5"
+                function, _, _ = function.partition("@@")
+
             imports[ea] = (library.lower(), function, ordinal)
             return True
 
@@ -97,14 +110,25 @@ def get_file_imports():
     return imports
 
 
-def get_instructions_in_range(start, end):
+def get_file_externs() -> Dict[int, Tuple[str, str, int]]:
+    externs = {}
+
+    for seg in get_segments(skip_header_segments=True):
+        if not (seg.type == ida_segment.SEG_XTRN):
+            continue
+
+        for ea in idautils.Functions(seg.start_ea, seg.end_ea):
+            externs[ea] = ("", idaapi.get_func_name(ea), -1)
+
+    return externs
+
+
+def get_instructions_in_range(start: int, end: int) -> Iterator[idaapi.insn_t]:
     """yield instructions in range
 
     args:
         start: virtual address (inclusive)
         end: virtual address (exclusive)
-    yield:
-        (insn_t*)
     """
     for head in idautils.Heads(start, end):
         insn = idautils.DecodeInstruction(head)
@@ -112,7 +136,7 @@ def get_instructions_in_range(start, end):
             yield insn
 
 
-def is_operand_equal(op1, op2):
+def is_operand_equal(op1: idaapi.op_t, op2: idaapi.op_t) -> bool:
     """compare two IDA op_t"""
     if op1.flags != op2.flags:
         return False
@@ -138,7 +162,7 @@ def is_operand_equal(op1, op2):
     return True
 
 
-def is_basic_block_equal(bb1, bb2):
+def is_basic_block_equal(bb1: idaapi.BasicBlock, bb2: idaapi.BasicBlock) -> bool:
     """compare two IDA BasicBlock"""
     if bb1.start_ea != bb2.start_ea:
         return False
@@ -152,12 +176,12 @@ def is_basic_block_equal(bb1, bb2):
     return True
 
 
-def basic_block_size(bb):
+def basic_block_size(bb: idaapi.BasicBlock) -> int:
     """calculate size of basic block"""
     return bb.end_ea - bb.start_ea
 
 
-def read_bytes_at(ea, count):
+def read_bytes_at(ea: int, count: int) -> bytes:
     """ """
     # check if byte has a value, see get_wide_byte doc
     if not idc.is_loaded(ea):
@@ -170,10 +194,10 @@ def read_bytes_at(ea, count):
         return idc.get_bytes(ea, count)
 
 
-def find_string_at(ea, min=4):
+def find_string_at(ea: int, min_: int = 4) -> str:
     """check if ASCII string exists at a given virtual address"""
     found = idaapi.get_strlit_contents(ea, -1, idaapi.STRTYPE_C)
-    if found and len(found) > min:
+    if found and len(found) >= min_:
         try:
             found = found.decode("ascii")
             # hacky check for IDA bug; get_strlit_contents also reads Unicode as
@@ -187,7 +211,7 @@ def find_string_at(ea, min=4):
     return ""
 
 
-def get_op_phrase_info(op):
+def get_op_phrase_info(op: idaapi.op_t) -> Dict:
     """parse phrase features from operand
 
     Pretty much dup of sark's implementation:
@@ -197,7 +221,8 @@ def get_op_phrase_info(op):
         return {}
 
     scale = 1 << ((op.specflag2 & 0xC0) >> 6)
-    offset = op.addr
+    # IDA ea_t may be 32- or 64-bit; we assume displacement can only be 32-bit
+    offset = op.addr & 0xFFFFFFFF
 
     if op.specflag1 == 0:
         index = None
@@ -224,23 +249,23 @@ def get_op_phrase_info(op):
     return {"base": base, "index": index, "scale": scale, "offset": offset}
 
 
-def is_op_write(insn, op):
+def is_op_write(insn: idaapi.insn_t, op: idaapi.op_t) -> bool:
     """Check if an operand is written to (destination operand)"""
     return idaapi.has_cf_chg(insn.get_canon_feature(), op.n)
 
 
-def is_op_read(insn, op):
+def is_op_read(insn: idaapi.insn_t, op: idaapi.op_t) -> bool:
     """Check if an operand is read from (source operand)"""
     return idaapi.has_cf_use(insn.get_canon_feature(), op.n)
 
 
-def is_op_offset(insn, op):
+def is_op_offset(insn: idaapi.insn_t, op: idaapi.op_t) -> bool:
     """Check is an operand has been marked as an offset (by auto-analysis or manually)"""
     flags = idaapi.get_flags(insn.ea)
     return ida_bytes.is_off(flags, op.n)
 
 
-def is_sp_modified(insn):
+def is_sp_modified(insn: idaapi.insn_t) -> bool:
     """determine if instruction modifies SP, ESP, RSP"""
     for op in get_insn_ops(insn, target_ops=(idaapi.o_reg,)):
         if op.reg == idautils.procregs.sp.reg and is_op_write(insn, op):
@@ -249,7 +274,7 @@ def is_sp_modified(insn):
     return False
 
 
-def is_bp_modified(insn):
+def is_bp_modified(insn: idaapi.insn_t) -> bool:
     """check if instruction modifies BP, EBP, RBP"""
     for op in get_insn_ops(insn, target_ops=(idaapi.o_reg,)):
         if op.reg == idautils.procregs.bp.reg and is_op_write(insn, op):
@@ -258,12 +283,12 @@ def is_bp_modified(insn):
     return False
 
 
-def is_frame_register(reg):
+def is_frame_register(reg: int) -> bool:
     """check if register is sp or bp"""
     return reg in (idautils.procregs.sp.reg, idautils.procregs.bp.reg)
 
 
-def get_insn_ops(insn, target_ops=()):
+def get_insn_ops(insn: idaapi.insn_t, target_ops: Optional[Tuple[Any]] = None) -> idaapi.op_t:
     """yield op_t for instruction, filter on type if specified"""
     for op in insn.ops:
         if op.type == idaapi.o_void:
@@ -274,12 +299,12 @@ def get_insn_ops(insn, target_ops=()):
         yield op
 
 
-def is_op_stack_var(ea, index):
+def is_op_stack_var(ea: int, index: int) -> bool:
     """check if operand is a stack variable"""
     return idaapi.is_stkvar(idaapi.get_flags(ea), index)
 
 
-def mask_op_val(op):
+def mask_op_val(op: idaapi.op_t) -> int:
     """mask value by data type
 
     necessary due to a bug in AMD64
@@ -299,26 +324,18 @@ def mask_op_val(op):
     return masks.get(op.dtype, op.value) & op.value
 
 
-def is_function_recursive(f):
-    """check if function is recursive
-
-    args:
-        f (IDA func_t)
-    """
+def is_function_recursive(f: idaapi.func_t) -> bool:
+    """check if function is recursive"""
     for ref in idautils.CodeRefsTo(f.start_ea, True):
         if f.contains(ref):
             return True
     return False
 
 
-def is_basic_block_tight_loop(bb):
+def is_basic_block_tight_loop(bb: idaapi.BasicBlock) -> bool:
     """check basic block loops to self
 
     true if last instruction in basic block branches to basic block start
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
     """
     bb_end = idc.prev_head(bb.end_ea)
     if bb.start_ea < bb_end:
@@ -328,7 +345,7 @@ def is_basic_block_tight_loop(bb):
     return False
 
 
-def find_data_reference_from_insn(insn, max_depth=10):
+def find_data_reference_from_insn(insn: idaapi.insn_t, max_depth: int = 10) -> int:
     """search for data reference from instruction, return address of instruction if no reference exists"""
     depth = 0
     ea = insn.ea
@@ -358,19 +375,18 @@ def find_data_reference_from_insn(insn, max_depth=10):
     return ea
 
 
-def get_function_blocks(f):
-    """yield basic blocks contained in specified function
-
-    args:
-        f (IDA func_t)
-    yield:
-        block (IDA BasicBlock)
-    """
+def get_function_blocks(f: idaapi.func_t) -> Iterator[idaapi.BasicBlock]:
+    """yield basic blocks contained in specified function"""
     # leverage idaapi.FC_NOEXT flag to ignore useless external blocks referenced by the function
     for block in idaapi.FlowChart(f, flags=(idaapi.FC_PREDS | idaapi.FC_NOEXT)):
         yield block
 
 
-def is_basic_block_return(bb):
+def is_basic_block_return(bb: idaapi.BasicBlock) -> bool:
     """check if basic block is return block"""
     return bb.type == idaapi.fcb_ret
+
+
+def has_sib(oper: idaapi.op_t) -> bool:
+    # via: https://reverseengineering.stackexchange.com/a/14300
+    return oper.specflag1 == 1

capa/features/extractors/ida/insn.py

@@ -1,10 +1,11 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
+from typing import Any, Dict, Tuple, Iterator
 
 import idc
 import idaapi
@@ -12,47 +13,29 @@ import idautils
 
 import capa.features.extractors.helpers
 import capa.features.extractors.ida.helpers
-from capa.features.insn import API, Number, Offset, Mnemonic
-from capa.features.common import (
-    BITNESS_X32,
-    BITNESS_X64,
-    MAX_BYTES_FEATURE_SIZE,
-    THUNK_CHAIN_DEPTH_DELTA,
-    Bytes,
-    String,
-    Characteristic,
-)
+from capa.features.insn import API, MAX_STRUCTURE_SIZE, Number, Offset, Mnemonic, OperandNumber, OperandOffset
+from capa.features.common import MAX_BYTES_FEATURE_SIZE, THUNK_CHAIN_DEPTH_DELTA, Bytes, String, Feature, Characteristic
+from capa.features.address import Address, AbsoluteVirtualAddress
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle
 
 # security cookie checks may perform non-zeroing XORs, these are expected within a certain
 # byte range within the first and returning basic blocks, this helps to reduce FP features
 SECURITY_COOKIE_BYTES_DELTA = 0x40
 
 
-def get_bitness(ctx):
-    """
-    fetch the BITNESS_* constant for the currently open workspace.
-
-    via Tamir Bahar/@tmr232
-    https://reverseengineering.stackexchange.com/a/11398/17194
-    """
-    if "bitness" not in ctx:
-        info = idaapi.get_inf_structure()
-        if info.is_64bit():
-            ctx["bitness"] = BITNESS_X64
-        elif info.is_32bit():
-            ctx["bitness"] = BITNESS_X32
-        else:
-            raise ValueError("unexpected bitness")
-    return ctx["bitness"]
-
-
-def get_imports(ctx):
+def get_imports(ctx: Dict[str, Any]) -> Dict[int, Any]:
     if "imports_cache" not in ctx:
         ctx["imports_cache"] = capa.features.extractors.ida.helpers.get_file_imports()
     return ctx["imports_cache"]
 
 
-def check_for_api_call(ctx, insn):
+def get_externs(ctx: Dict[str, Any]) -> Dict[int, Any]:
+    if "externs_cache" not in ctx:
+        ctx["externs_cache"] = capa.features.extractors.ida.helpers.get_file_externs()
+    return ctx["externs_cache"]
+
+
+def check_for_api_call(insn: idaapi.insn_t, funcs: Dict[int, Any]) -> Iterator[Any]:
     """check instruction for API call"""
     info = ()
     ref = insn.ea
@@ -69,7 +52,7 @@ def check_for_api_call(ctx, insn):
             except IndexError:
                 break
 
-        info = get_imports(ctx).get(ref, ())
+        info = funcs.get(ref, ())
         if info:
             break
 
@@ -78,27 +61,31 @@ def check_for_api_call(ctx, insn):
             break
 
     if info:
-        yield "%s.%s" % (info[0], info[1])
+        yield info
 
 
-def extract_insn_api_features(f, bb, insn):
-    """parse instruction API features
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-        insn (IDA insn_t)
+def extract_insn_api_features(fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """
+    parse instruction API features
 
     example:
-        call dword [0x00473038]
+       call dword [0x00473038]
     """
+    insn: idaapi.insn_t = ih.inner
+
     if not insn.get_canon_mnem() in ("call", "jmp"):
         return
 
-    for api in check_for_api_call(f.ctx, insn):
-        dll, _, symbol = api.rpartition(".")
-        for name in capa.features.extractors.helpers.generate_symbols(dll, symbol):
-            yield API(name), insn.ea
+    # check calls to imported functions
+    for api in check_for_api_call(insn, get_imports(fh.ctx)):
+        # tuple (<module>, <function>, <ordinal>)
+        for name in capa.features.extractors.helpers.generate_symbols(api[0], api[1]):
+            yield API(name), ih.address
+
+    # check calls to extern functions
+    for api in check_for_api_call(insn, get_externs(fh.ctx)):
+        # tuple (<module>, <function>, <ordinal>)
+        yield API(api[1]), ih.address
 
     # extract IDA/FLIRT recognized API functions
     targets = tuple(idautils.CodeRefsFrom(insn.ea, False))
@@ -113,20 +100,25 @@ def extract_insn_api_features(f, bb, insn):
 
     if target_func.flags & idaapi.FUNC_LIB:
         name = idaapi.get_name(target_func.start_ea)
-        yield API(name), insn.ea
+        yield API(name), ih.address
+        if name.startswith("_"):
+            # some linkers may prefix linked routines with a `_` to avoid name collisions.
+            # extract features for both the mangled and un-mangled representations.
+            # e.g. `_fwrite` -> `fwrite`
+            # see: https://stackoverflow.com/a/2628384/87207
+            yield API(name[1:]), ih.address
 
 
-def extract_insn_number_features(f, bb, insn):
-    """parse instruction number features
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-        insn (IDA insn_t)
-
+def extract_insn_number_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    """
+    parse instruction number features
     example:
         push    3136B0h         ; dwControlCode
     """
+    insn: idaapi.insn_t = ih.inner
+
     if idaapi.is_ret_insn(insn):
         # skip things like:
         #   .text:0042250E retn 8
@@ -137,7 +129,11 @@ def extract_insn_number_features(f, bb, insn):
         #   .text:00401145 add esp, 0Ch
         return
 
-    for op in capa.features.extractors.ida.helpers.get_insn_ops(insn, target_ops=(idaapi.o_imm, idaapi.o_mem)):
+    for i, op in enumerate(insn.ops):
+        if op.type == idaapi.o_void:
+            break
+        if op.type not in (idaapi.o_imm, idaapi.o_mem):
+            continue
         # skip things like:
         #   .text:00401100 shr eax, offset loc_C
         if capa.features.extractors.ida.helpers.is_op_offset(insn, op):
@@ -148,21 +144,27 @@ def extract_insn_number_features(f, bb, insn):
         else:
             const = op.addr
 
-        yield Number(const), insn.ea
-        yield Number(const, bitness=get_bitness(f.ctx)), insn.ea
+        yield Number(const), ih.address
+        yield OperandNumber(i, const), ih.address
+
+        if insn.itype == idaapi.NN_add and 0 < const < MAX_STRUCTURE_SIZE and op.type == idaapi.o_imm:
+            # for pattern like:
+            #
+            #     add eax, 0x10
+            #
+            # assume 0x10 is also an offset (imagine eax is a pointer).
+            yield Offset(const), ih.address
+            yield OperandOffset(i, const), ih.address
 
 
-def extract_insn_bytes_features(f, bb, insn):
-    """parse referenced byte sequences
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-        insn (IDA insn_t)
-
+def extract_insn_bytes_features(fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """
+    parse referenced byte sequences
     example:
         push    offset iid_004118d4_IShellLinkA ; riid
     """
+    insn: idaapi.insn_t = ih.inner
+
     if idaapi.is_call_insn(insn):
         return
 
@@ -170,43 +172,54 @@ def extract_insn_bytes_features(f, bb, insn):
     if ref != insn.ea:
         extracted_bytes = capa.features.extractors.ida.helpers.read_bytes_at(ref, MAX_BYTES_FEATURE_SIZE)
         if extracted_bytes and not capa.features.extractors.helpers.all_zeros(extracted_bytes):
-            yield Bytes(extracted_bytes), insn.ea
+            if not capa.features.extractors.ida.helpers.find_string_at(insn.ea):
+                # don't extract byte features for obvious strings
+                yield Bytes(extracted_bytes), ih.address
 
 
-def extract_insn_string_features(f, bb, insn):
-    """parse instruction string features
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-        insn (IDA insn_t)
+def extract_insn_string_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    """
+    parse instruction string features
 
     example:
         push offset aAcr     ; "ACR  > "
     """
+    insn: idaapi.insn_t = ih.inner
+
     ref = capa.features.extractors.ida.helpers.find_data_reference_from_insn(insn)
     if ref != insn.ea:
         found = capa.features.extractors.ida.helpers.find_string_at(ref)
         if found:
-            yield String(found), insn.ea
+            yield String(found), ih.address
 
 
-def extract_insn_offset_features(f, bb, insn):
-    """parse instruction structure offset features
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-        insn (IDA insn_t)
+def extract_insn_offset_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    """
+    parse instruction structure offset features
 
     example:
         .text:0040112F cmp [esi+4], ebx
     """
-    for op in capa.features.extractors.ida.helpers.get_insn_ops(insn, target_ops=(idaapi.o_phrase, idaapi.o_displ)):
+    insn: idaapi.insn_t = ih.inner
+
+    for i, op in enumerate(insn.ops):
+        if op.type == idaapi.o_void:
+            break
+        if op.type not in (idaapi.o_phrase, idaapi.o_displ):
+            continue
         if capa.features.extractors.ida.helpers.is_op_stack_var(insn.ea, op.n):
             continue
+
         p_info = capa.features.extractors.ida.helpers.get_op_phrase_info(op)
-        op_off = p_info.get("offset", 0)
+
+        op_off = p_info.get("offset", None)
+        if op_off is None:
+            continue
+
         if idaapi.is_mapped(op_off):
             # Ignore:
             #   mov esi, dword_1005B148[esi]
@@ -217,12 +230,32 @@ def extract_insn_offset_features(f, bb, insn):
         # https://stackoverflow.com/questions/31853189/x86-64-assembly-why-displacement-not-64-bits
         op_off = capa.features.extractors.helpers.twos_complement(op_off, 32)
 
-        yield Offset(op_off), insn.ea
-        yield Offset(op_off, bitness=get_bitness(f.ctx)), insn.ea
+        yield Offset(op_off), ih.address
+        yield OperandOffset(i, op_off), ih.address
+
+        if (
+            insn.itype == idaapi.NN_lea
+            and i == 1
+            # o_displ is used for both:
+            #   [eax+1]
+            #   [eax+ebx+2]
+            and op.type == idaapi.o_displ
+            # but the SIB is only present for [eax+ebx+2]
+            # which we don't want
+            and not capa.features.extractors.ida.helpers.has_sib(op)
+        ):
+            # for pattern like:
+            #
+            #     lea eax, [ebx + 1]
+            #
+            # assume 1 is also an offset (imagine ebx is a zero register).
+            yield Number(op_off), ih.address
+            yield OperandNumber(i, op_off), ih.address
 
 
-def contains_stack_cookie_keywords(s):
-    """check if string contains stack cookie keywords
+def contains_stack_cookie_keywords(s: str) -> bool:
+    """
+    check if string contains stack cookie keywords
 
     Examples:
         xor     ecx, ebp ; StackCookie
@@ -236,7 +269,7 @@ def contains_stack_cookie_keywords(s):
     return any(keyword in s for keyword in ("stack", "security"))
 
 
-def bb_stack_cookie_registers(bb):
+def bb_stack_cookie_registers(bb: idaapi.BasicBlock) -> Iterator[int]:
     """scan basic block for stack cookie operations
 
     yield registers ids that may have been used for stack cookie operations
@@ -270,7 +303,7 @@ def bb_stack_cookie_registers(bb):
                     yield op.reg
 
 
-def is_nzxor_stack_cookie_delta(f, bb, insn):
+def is_nzxor_stack_cookie_delta(f: idaapi.func_t, bb: idaapi.BasicBlock, insn: idaapi.insn_t) -> bool:
     """check if nzxor exists within stack cookie delta"""
     # security cookie check should use SP or BP
     if not capa.features.extractors.ida.helpers.is_frame_register(insn.Op2.reg):
@@ -293,7 +326,7 @@ def is_nzxor_stack_cookie_delta(f, bb, insn):
     return False
 
 
-def is_nzxor_stack_cookie(f, bb, insn):
+def is_nzxor_stack_cookie(f: idaapi.func_t, bb: idaapi.BasicBlock, insn: idaapi.insn_t) -> bool:
     """check if nzxor is related to stack cookie"""
     if contains_stack_cookie_keywords(idaapi.get_cmt(insn.ea, False)):
         # Example:
@@ -310,37 +343,49 @@ def is_nzxor_stack_cookie(f, bb, insn):
     return False
 
 
-def extract_insn_nzxor_characteristic_features(f, bb, insn):
-    """parse instruction non-zeroing XOR instruction
-
-    ignore expected non-zeroing XORs, e.g. security cookies
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-        insn (IDA insn_t)
+def extract_insn_nzxor_characteristic_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
     """
+    parse instruction non-zeroing XOR instruction
+    ignore expected non-zeroing XORs, e.g. security cookies
+    """
+    insn: idaapi.insn_t = ih.inner
+
     if insn.itype not in (idaapi.NN_xor, idaapi.NN_xorpd, idaapi.NN_xorps, idaapi.NN_pxor):
         return
     if capa.features.extractors.ida.helpers.is_operand_equal(insn.Op1, insn.Op2):
         return
-    if is_nzxor_stack_cookie(f, bb, insn):
+    if is_nzxor_stack_cookie(fh.inner, bbh.inner, insn):
         return
-    yield Characteristic("nzxor"), insn.ea
+    yield Characteristic("nzxor"), ih.address
 
 
-def extract_insn_mnemonic_features(f, bb, insn):
-    """parse instruction mnemonic features
+def extract_insn_mnemonic_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    """parse instruction mnemonic features"""
+    yield Mnemonic(idc.print_insn_mnem(ih.inner.ea)), ih.address
 
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-        insn (IDA insn_t)
+
+def extract_insn_obfs_call_plus_5_characteristic_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
     """
-    yield Mnemonic(idc.print_insn_mnem(insn.ea)), insn.ea
+    parse call $+5 instruction from the given instruction.
+    """
+    insn: idaapi.insn_t = ih.inner
+
+    if not idaapi.is_call_insn(insn):
+        return
+
+    if insn.ea + 5 == idc.get_operand_value(insn.ea, 0):
+        yield Characteristic("call $+5"), ih.address
 
 
-def extract_insn_peb_access_characteristic_features(f, bb, insn):
+def extract_insn_peb_access_characteristic_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
     """parse instruction peb access
 
     fs:[0x30] on x86, gs:[0x60] on x64
@@ -348,6 +393,8 @@ def extract_insn_peb_access_characteristic_features(f, bb, insn):
     TODO:
         IDA should be able to do this..
     """
+    insn: idaapi.insn_t = ih.inner
+
     if insn.itype not in (idaapi.NN_push, idaapi.NN_mov):
         return
 
@@ -359,15 +406,19 @@ def extract_insn_peb_access_characteristic_features(f, bb, insn):
 
     if " fs:30h" in disasm or " gs:60h" in disasm:
         # TODO: replace above with proper IDA
-        yield Characteristic("peb access"), insn.ea
+        yield Characteristic("peb access"), ih.address
 
 
-def extract_insn_segment_access_features(f, bb, insn):
+def extract_insn_segment_access_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
     """parse instruction fs or gs access
 
     TODO:
         IDA should be able to do this...
     """
+    insn: idaapi.insn_t = ih.inner
+
     if all(map(lambda op: op.type != idaapi.o_mem, insn.ops)):
         # try to optimize for only memory references
         return
@@ -376,23 +427,21 @@ def extract_insn_segment_access_features(f, bb, insn):
 
     if " fs:" in disasm:
         # TODO: replace above with proper IDA
-        yield Characteristic("fs access"), insn.ea
+        yield Characteristic("fs access"), ih.address
 
     if " gs:" in disasm:
         # TODO: replace above with proper IDA
-        yield Characteristic("gs access"), insn.ea
+        yield Characteristic("gs access"), ih.address
 
 
-def extract_insn_cross_section_cflow(f, bb, insn):
-    """inspect the instruction for a CALL or JMP that crosses section boundaries
+def extract_insn_cross_section_cflow(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    """inspect the instruction for a CALL or JMP that crosses section boundaries"""
+    insn: idaapi.insn_t = ih.inner
 
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-        insn (IDA insn_t)
-    """
     for ref in idautils.CodeRefsFrom(insn.ea, False):
-        if ref in get_imports(f.ctx).keys():
+        if ref in get_imports(fh.ctx).keys():
             # ignore API calls
             continue
         if not idaapi.getseg(ref):
@@ -400,50 +449,40 @@ def extract_insn_cross_section_cflow(f, bb, insn):
             continue
         if idaapi.getseg(ref) == idaapi.getseg(insn.ea):
             continue
-        yield Characteristic("cross section flow"), insn.ea
+        yield Characteristic("cross section flow"), ih.address
 
 
-def extract_function_calls_from(f, bb, insn):
+def extract_function_calls_from(fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
     """extract functions calls from features
 
     most relevant at the function scope, however, its most efficient to extract at the instruction scope
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-        insn (IDA insn_t)
     """
+    insn: idaapi.insn_t = ih.inner
+
     if idaapi.is_call_insn(insn):
         for ref in idautils.CodeRefsFrom(insn.ea, False):
-            yield Characteristic("calls from"), ref
+            yield Characteristic("calls from"), AbsoluteVirtualAddress(ref)
 
 
-def extract_function_indirect_call_characteristic_features(f, bb, insn):
+def extract_function_indirect_call_characteristic_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
     """extract indirect function calls (e.g., call eax or call dword ptr [edx+4])
     does not include calls like => call ds:dword_ABD4974
 
     most relevant at the function or basic block scope;
     however, its most efficient to extract at the instruction scope
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-        insn (IDA insn_t)
     """
+    insn: idaapi.insn_t = ih.inner
+
     if idaapi.is_call_insn(insn) and idc.get_operand_type(insn.ea, 0) in (idc.o_reg, idc.o_phrase, idc.o_displ):
-        yield Characteristic("indirect call"), insn.ea
+        yield Characteristic("indirect call"), ih.address
 
 
-def extract_features(f, bb, insn):
-    """extract instruction features
-
-    args:
-        f (IDA func_t)
-        bb (IDA BasicBlock)
-        insn (IDA insn_t)
-    """
+def extract_features(f: FunctionHandle, bbh: BBHandle, insn: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """extract instruction features"""
     for inst_handler in INSTRUCTION_HANDLERS:
-        for (feature, ea) in inst_handler(f, bb, insn):
+        for feature, ea in inst_handler(f, bbh, insn):
             yield feature, ea
 
 
@@ -455,6 +494,7 @@ INSTRUCTION_HANDLERS = (
     extract_insn_offset_features,
     extract_insn_nzxor_characteristic_features,
     extract_insn_mnemonic_features,
+    extract_insn_obfs_call_plus_5_characteristic_features,
     extract_insn_peb_access_characteristic_features,
     extract_insn_cross_section_cflow,
     extract_insn_segment_access_features,

capa/features/extractors/loops.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt

capa/features/extractors/null.py

@@ -0,0 +1,72 @@
+from typing import Dict, List, Tuple
+from dataclasses import dataclass
+
+from capa.features.common import Feature
+from capa.features.address import NO_ADDRESS, Address
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle, FeatureExtractor
+
+
+@dataclass
+class InstructionFeatures:
+    features: List[Tuple[Address, Feature]]
+
+
+@dataclass
+class BasicBlockFeatures:
+    features: List[Tuple[Address, Feature]]
+    instructions: Dict[Address, InstructionFeatures]
+
+
+@dataclass
+class FunctionFeatures:
+    features: List[Tuple[Address, Feature]]
+    basic_blocks: Dict[Address, BasicBlockFeatures]
+
+
+@dataclass
+class NullFeatureExtractor(FeatureExtractor):
+    """
+    An extractor that extracts some user-provided features.
+
+    This is useful for testing, as we can provide expected values and see if matching works.
+    """
+
+    base_address: Address
+    global_features: List[Feature]
+    file_features: List[Tuple[Address, Feature]]
+    functions: Dict[Address, FunctionFeatures]
+
+    def get_base_address(self):
+        return self.base_address
+
+    def extract_global_features(self):
+        for feature in self.global_features:
+            yield feature, NO_ADDRESS
+
+    def extract_file_features(self):
+        for address, feature in self.file_features:
+            yield feature, address
+
+    def get_functions(self):
+        for address in sorted(self.functions.keys()):
+            yield FunctionHandle(address, None)
+
+    def extract_function_features(self, f):
+        for address, feature in self.functions[f.address].features:
+            yield feature, address
+
+    def get_basic_blocks(self, f):
+        for address in sorted(self.functions[f.address].basic_blocks.keys()):
+            yield BBHandle(address, None)
+
+    def extract_basic_block_features(self, f, bb):
+        for address, feature in self.functions[f.address].basic_blocks[bb.address].features:
+            yield feature, address
+
+    def get_instructions(self, f, bb):
+        for address in sorted(self.functions[f.address].basic_blocks[bb.address].instructions.keys()):
+            yield InsnHandle(address, None)
+
+    def extract_insn_features(self, f, bb, insn):
+        for address, feature in self.functions[f.address].basic_blocks[bb.address].instructions[insn.address].features:
+            yield feature, address

capa/features/extractors/pefile.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -17,6 +17,7 @@ import capa.features.extractors.helpers
 import capa.features.extractors.strings
 from capa.features.file import Export, Import, Section
 from capa.features.common import OS, ARCH_I386, FORMAT_PE, ARCH_AMD64, OS_WINDOWS, Arch, Format, Characteristic
+from capa.features.address import NO_ADDRESS, FileOffsetAddress, AbsoluteVirtualAddress
 from capa.features.extractors.base_extractor import FeatureExtractor
 
 logger = logging.getLogger(__name__)
@@ -24,7 +25,7 @@ logger = logging.getLogger(__name__)
 
 def extract_file_embedded_pe(buf, **kwargs):
     for offset, _ in capa.features.extractors.helpers.carve_pe(buf, 1):
-        yield Characteristic("embedded pe"), offset
+        yield Characteristic("embedded pe"), FileOffsetAddress(offset)
 
 
 def extract_file_export_names(pe, **kwargs):
@@ -39,7 +40,7 @@ def extract_file_export_names(pe, **kwargs):
             except UnicodeDecodeError:
                 continue
             va = base_address + export.address
-            yield Export(name), va
+            yield Export(name), AbsoluteVirtualAddress(va)
 
 
 def extract_file_import_names(pe, **kwargs):
@@ -71,7 +72,7 @@ def extract_file_import_names(pe, **kwargs):
                         continue
 
                 for name in capa.features.extractors.helpers.generate_symbols(modname, impname):
-                    yield Import(name), imp.address
+                    yield Import(name), AbsoluteVirtualAddress(imp.address)
 
 
 def extract_file_section_names(pe, **kwargs):
@@ -83,7 +84,7 @@ def extract_file_section_names(pe, **kwargs):
         except UnicodeDecodeError:
             continue
 
-        yield Section(name), base_address + section.VirtualAddress
+        yield Section(name), AbsoluteVirtualAddress(base_address + section.VirtualAddress)
 
 
 def extract_file_strings(buf, **kwargs):
@@ -103,18 +104,18 @@ def extract_file_function_names(**kwargs):
 def extract_file_os(**kwargs):
     # assuming PE -> Windows
     # though i suppose they're also used by UEFI
-    yield OS(OS_WINDOWS), 0x0
+    yield OS(OS_WINDOWS), NO_ADDRESS
 
 
 def extract_file_format(**kwargs):
-    yield Format(FORMAT_PE), 0x0
+    yield Format(FORMAT_PE), NO_ADDRESS
 
 
 def extract_file_arch(pe, **kwargs):
     if pe.FILE_HEADER.Machine == pefile.MACHINE_TYPE["IMAGE_FILE_MACHINE_I386"]:
-        yield Arch(ARCH_I386), 0x0
+        yield Arch(ARCH_I386), NO_ADDRESS
     elif pe.FILE_HEADER.Machine == pefile.MACHINE_TYPE["IMAGE_FILE_MACHINE_AMD64"]:
-        yield Arch(ARCH_AMD64), 0x0
+        yield Arch(ARCH_AMD64), NO_ADDRESS
     else:
         logger.warning("unsupported architecture: %s", pefile.MACHINE_TYPE[pe.FILE_HEADER.Machine])
 
@@ -132,7 +133,8 @@ def extract_file_features(pe, buf):
     """
 
     for file_handler in FILE_HANDLERS:
-        for feature, va in file_handler(pe=pe, buf=buf):
+        # file_handler: type: (pe, bytes) -> Iterable[Tuple[Feature, Address]]
+        for feature, va in file_handler(pe=pe, buf=buf):  # type: ignore
             yield feature, va
 
 
@@ -159,7 +161,8 @@ def extract_global_features(pe, buf):
       Tuple[Feature, VA]: a feature and its location.
     """
     for handler in GLOBAL_HANDLERS:
-        for feature, va in handler(pe=pe, buf=buf):
+        # file_handler: type: (pe, bytes) -> Iterable[Tuple[Feature, Address]]
+        for feature, va in handler(pe=pe, buf=buf):  # type: ignore
             yield feature, va
 
 
@@ -171,12 +174,12 @@ GLOBAL_HANDLERS = (
 
 class PefileFeatureExtractor(FeatureExtractor):
     def __init__(self, path: str):
-        super(PefileFeatureExtractor, self).__init__()
+        super().__init__()
         self.path = path
         self.pe = pefile.PE(path)
 
     def get_base_address(self):
-        return self.pe.OPTIONAL_HEADER.ImageBase
+        return AbsoluteVirtualAddress(self.pe.OPTIONAL_HEADER.ImageBase)
 
     def extract_global_features(self):
         with open(self.path, "rb") as f:

capa/features/extractors/smda/basicblock.py

@@ -1,130 +0,0 @@
-import string
-import struct
-
-from capa.features.common import Characteristic
-from capa.features.basicblock import BasicBlock
-from capa.features.extractors.helpers import MIN_STACKSTRING_LEN
-
-
-def _bb_has_tight_loop(f, bb):
-    """
-    parse tight loops, true if last instruction in basic block branches to bb start
-    """
-    return bb.offset in f.blockrefs[bb.offset] if bb.offset in f.blockrefs else False
-
-
-def extract_bb_tight_loop(f, bb):
-    """check basic block for tight loop indicators"""
-    if _bb_has_tight_loop(f, bb):
-        yield Characteristic("tight loop"), bb.offset
-
-
-def _bb_has_stackstring(f, bb):
-    """
-    extract potential stackstring creation, using the following heuristics:
-      - basic block contains enough moves of constant bytes to the stack
-    """
-    count = 0
-    for instr in bb.getInstructions():
-        if is_mov_imm_to_stack(instr):
-            count += get_printable_len(instr.getDetailed())
-        if count > MIN_STACKSTRING_LEN:
-            return True
-    return False
-
-
-def get_operands(smda_ins):
-    return [o.strip() for o in smda_ins.operands.split(",")]
-
-
-def extract_stackstring(f, bb):
-    """check basic block for stackstring indicators"""
-    if _bb_has_stackstring(f, bb):
-        yield Characteristic("stack string"), bb.offset
-
-
-def is_mov_imm_to_stack(smda_ins):
-    """
-    Return if instruction moves immediate onto stack
-    """
-    if not smda_ins.mnemonic.startswith("mov"):
-        return False
-
-    try:
-        dst, src = get_operands(smda_ins)
-    except ValueError:
-        # not two operands
-        return False
-
-    try:
-        int(src, 16)
-    except ValueError:
-        return False
-
-    if not any(regname in dst for regname in ["ebp", "rbp", "esp", "rsp"]):
-        return False
-
-    return True
-
-
-def is_printable_ascii(chars):
-    return all(c < 127 and chr(c) in string.printable for c in chars)
-
-
-def is_printable_utf16le(chars):
-    if all(c == 0x00 for c in chars[1::2]):
-        return is_printable_ascii(chars[::2])
-
-
-def get_printable_len(instr):
-    """
-    Return string length if all operand bytes are ascii or utf16-le printable
-
-    Works on a capstone instruction
-    """
-    # should have exactly two operands for mov immediate
-    if len(instr.operands) != 2:
-        return 0
-
-    op_value = instr.operands[1].value.imm
-
-    if instr.imm_size == 1:
-        chars = struct.pack("<B", op_value & 0xFF)
-    elif instr.imm_size == 2:
-        chars = struct.pack("<H", op_value & 0xFFFF)
-    elif instr.imm_size == 4:
-        chars = struct.pack("<I", op_value & 0xFFFFFFFF)
-    elif instr.imm_size == 8:
-        chars = struct.pack("<Q", op_value & 0xFFFFFFFFFFFFFFFF)
-    else:
-        raise ValueError("Unhandled operand data type 0x%x." % instr.imm_size)
-
-    if is_printable_ascii(chars):
-        return instr.imm_size
-    if is_printable_utf16le(chars):
-        return instr.imm_size // 2
-
-    return 0
-
-
-def extract_features(f, bb):
-    """
-    extract features from the given basic block.
-
-    args:
-      f (smda.common.SmdaFunction): the function from which to extract features
-      bb (smda.common.SmdaBasicBlock): the basic block to process.
-
-    yields:
-      Tuple[Feature, int]: the features and their location found in this basic block.
-    """
-    yield BasicBlock(), bb.offset
-    for bb_handler in BASIC_BLOCK_HANDLERS:
-        for feature, va in bb_handler(f, bb):
-            yield feature, va
-
-
-BASIC_BLOCK_HANDLERS = (
-    extract_bb_tight_loop,
-    extract_stackstring,
-)

capa/features/extractors/smda/extractor.py

@@ -1,53 +0,0 @@
-from smda.common.SmdaReport import SmdaReport
-
-import capa.features.extractors.common
-import capa.features.extractors.smda.file
-import capa.features.extractors.smda.insn
-import capa.features.extractors.smda.global_
-import capa.features.extractors.smda.function
-import capa.features.extractors.smda.basicblock
-from capa.features.extractors.base_extractor import FeatureExtractor
-
-
-class SmdaFeatureExtractor(FeatureExtractor):
-    def __init__(self, smda_report: SmdaReport, path):
-        super(SmdaFeatureExtractor, self).__init__()
-        self.smda_report = smda_report
-        self.path = path
-        with open(self.path, "rb") as f:
-            self.buf = f.read()
-
-        # pre-compute these because we'll yield them at *every* scope.
-        self.global_features = []
-        self.global_features.extend(capa.features.extractors.common.extract_os(self.buf))
-        self.global_features.extend(capa.features.extractors.smda.global_.extract_arch(self.smda_report))
-
-    def get_base_address(self):
-        return self.smda_report.base_addr
-
-    def extract_global_features(self):
-        yield from self.global_features
-
-    def extract_file_features(self):
-        yield from capa.features.extractors.smda.file.extract_features(self.smda_report, self.buf)
-
-    def get_functions(self):
-        for function in self.smda_report.getFunctions():
-            yield function
-
-    def extract_function_features(self, f):
-        yield from capa.features.extractors.smda.function.extract_features(f)
-
-    def get_basic_blocks(self, f):
-        for bb in f.getBlocks():
-            yield bb
-
-    def extract_basic_block_features(self, f, bb):
-        yield from capa.features.extractors.smda.basicblock.extract_features(f, bb)
-
-    def get_instructions(self, f, bb):
-        for smda_ins in bb.getInstructions():
-            yield smda_ins
-
-    def extract_insn_features(self, f, bb, insn):
-        yield from capa.features.extractors.smda.insn.extract_features(f, bb, insn)

capa/features/extractors/smda/file.py

@@ -1,102 +0,0 @@
-# if we have SMDA we definitely have lief
-import lief
-
-import capa.features.extractors.common
-import capa.features.extractors.helpers
-import capa.features.extractors.strings
-from capa.features.file import Export, Import, Section
-from capa.features.common import String, Characteristic
-
-
-def extract_file_embedded_pe(buf, **kwargs):
-    for offset, _ in capa.features.extractors.helpers.carve_pe(buf, 1):
-        yield Characteristic("embedded pe"), offset
-
-
-def extract_file_export_names(buf, **kwargs):
-    lief_binary = lief.parse(buf)
-
-    if lief_binary is not None:
-        for function in lief_binary.exported_functions:
-            yield Export(function.name), function.address
-
-
-def extract_file_import_names(smda_report, buf):
-    # extract import table info via LIEF
-    lief_binary = lief.parse(buf)
-    if not isinstance(lief_binary, lief.PE.Binary):
-        return
-    for imported_library in lief_binary.imports:
-        library_name = imported_library.name.lower()
-        library_name = library_name[:-4] if library_name.endswith(".dll") else library_name
-        for func in imported_library.entries:
-            va = func.iat_address + smda_report.base_addr
-            if func.name:
-                for name in capa.features.extractors.helpers.generate_symbols(library_name, func.name):
-                    yield Import(name), va
-            elif func.is_ordinal:
-                for name in capa.features.extractors.helpers.generate_symbols(library_name, "#%s" % func.ordinal):
-                    yield Import(name), va
-
-
-def extract_file_section_names(buf, **kwargs):
-    lief_binary = lief.parse(buf)
-    if not isinstance(lief_binary, lief.PE.Binary):
-        return
-    if lief_binary and lief_binary.sections:
-        base_address = lief_binary.optional_header.imagebase
-        for section in lief_binary.sections:
-            yield Section(section.name), base_address + section.virtual_address
-
-
-def extract_file_strings(buf, **kwargs):
-    """
-    extract ASCII and UTF-16 LE strings from file
-    """
-    for s in capa.features.extractors.strings.extract_ascii_strings(buf):
-        yield String(s.s), s.offset
-
-    for s in capa.features.extractors.strings.extract_unicode_strings(buf):
-        yield String(s.s), s.offset
-
-
-def extract_file_function_names(smda_report, **kwargs):
-    """
-    extract the names of statically-linked library functions.
-    """
-    if False:
-        # using a `yield` here to force this to be a generator, not function.
-        yield NotImplementedError("SMDA doesn't have library matching")
-    return
-
-
-def extract_file_format(buf, **kwargs):
-    yield from capa.features.extractors.common.extract_format(buf)
-
-
-def extract_features(smda_report, buf):
-    """
-    extract file features from given workspace
-
-    args:
-      smda_report (smda.common.SmdaReport): a SmdaReport
-      buf: the raw bytes of the sample
-
-    yields:
-      Tuple[Feature, VA]: a feature and its location.
-    """
-
-    for file_handler in FILE_HANDLERS:
-        for feature, va in file_handler(smda_report=smda_report, buf=buf):
-            yield feature, va
-
-
-FILE_HANDLERS = (
-    extract_file_embedded_pe,
-    extract_file_export_names,
-    extract_file_import_names,
-    extract_file_section_names,
-    extract_file_strings,
-    extract_file_function_names,
-    extract_file_format,
-)

capa/features/extractors/smda/function.py

@@ -1,38 +0,0 @@
-from capa.features.common import Characteristic
-from capa.features.extractors import loops
-
-
-def extract_function_calls_to(f):
-    for inref in f.inrefs:
-        yield Characteristic("calls to"), inref
-
-
-def extract_function_loop(f):
-    """
-    parse if a function has a loop
-    """
-    edges = []
-    for bb_from, bb_tos in f.blockrefs.items():
-        for bb_to in bb_tos:
-            edges.append((bb_from, bb_to))
-
-    if edges and loops.has_loop(edges):
-        yield Characteristic("loop"), f.offset
-
-
-def extract_features(f):
-    """
-    extract features from the given function.
-
-    args:
-      f (smda.common.SmdaFunction): the function from which to extract features
-
-    yields:
-      Tuple[Feature, int]: the features and their location found in this function.
-    """
-    for func_handler in FUNCTION_HANDLERS:
-        for feature, va in func_handler(f):
-            yield feature, va
-
-
-FUNCTION_HANDLERS = (extract_function_calls_to, extract_function_loop)

capa/features/extractors/smda/global_.py

@@ -1,20 +0,0 @@
-import logging
-
-from capa.features.common import ARCH_I386, ARCH_AMD64, Arch
-
-logger = logging.getLogger(__name__)
-
-
-def extract_arch(smda_report):
-    if smda_report.architecture == "intel":
-        if smda_report.bitness == 32:
-            yield Arch(ARCH_I386), 0x0
-        elif smda_report.bitness == 64:
-            yield Arch(ARCH_AMD64), 0x0
-    else:
-        # we likely end up here:
-        #  1. handling a new architecture (e.g. aarch64)
-        #
-        # for (1), this logic will need to be updated as the format is implemented.
-        logger.debug("unsupported architecture: %s", smda_report.architecture)
-        return

capa/features/extractors/smda/insn.py

@@ -1,393 +0,0 @@
-import re
-import string
-import struct
-
-from smda.common.SmdaReport import SmdaReport
-
-import capa.features.extractors.helpers
-from capa.features.insn import API, Number, Offset, Mnemonic
-from capa.features.common import (
-    BITNESS_X32,
-    BITNESS_X64,
-    MAX_BYTES_FEATURE_SIZE,
-    THUNK_CHAIN_DEPTH_DELTA,
-    Bytes,
-    String,
-    Characteristic,
-)
-
-# security cookie checks may perform non-zeroing XORs, these are expected within a certain
-# byte range within the first and returning basic blocks, this helps to reduce FP features
-SECURITY_COOKIE_BYTES_DELTA = 0x40
-PATTERN_HEXNUM = re.compile(r"[+\-] (?P<num>0x[a-fA-F0-9]+)")
-PATTERN_SINGLENUM = re.compile(r"[+\-] (?P<num>[0-9])")
-
-
-def get_bitness(smda_report):
-    if smda_report.architecture == "intel":
-        if smda_report.bitness == 32:
-            return BITNESS_X32
-        elif smda_report.bitness == 64:
-            return BITNESS_X64
-    else:
-        raise NotImplementedError
-
-
-def extract_insn_api_features(f, bb, insn):
-    """parse API features from the given instruction."""
-    if insn.offset in f.apirefs:
-        api_entry = f.apirefs[insn.offset]
-        # reformat
-        dll_name, api_name = api_entry.split("!")
-        dll_name = dll_name.split(".")[0]
-        dll_name = dll_name.lower()
-        for name in capa.features.extractors.helpers.generate_symbols(dll_name, api_name):
-            yield API(name), insn.offset
-    elif insn.offset in f.outrefs:
-        current_function = f
-        current_instruction = insn
-        for index in range(THUNK_CHAIN_DEPTH_DELTA):
-            if current_function and len(current_function.outrefs[current_instruction.offset]) == 1:
-                target = current_function.outrefs[current_instruction.offset][0]
-                referenced_function = current_function.smda_report.getFunction(target)
-                if referenced_function:
-                    # TODO SMDA: implement this function for both jmp and call, checking if function has 1 instruction which refs an API
-                    if referenced_function.isApiThunk():
-                        api_entry = (
-                            referenced_function.apirefs[target] if target in referenced_function.apirefs else None
-                        )
-                        if api_entry:
-                            # reformat
-                            dll_name, api_name = api_entry.split("!")
-                            dll_name = dll_name.split(".")[0]
-                            dll_name = dll_name.lower()
-                            for name in capa.features.extractors.helpers.generate_symbols(dll_name, api_name):
-                                yield API(name), insn.offset
-                    elif referenced_function.num_instructions == 1 and referenced_function.num_outrefs == 1:
-                        current_function = referenced_function
-                        current_instruction = [i for i in referenced_function.getInstructions()][0]
-                else:
-                    return
-
-
-def extract_insn_number_features(f, bb, insn):
-    """parse number features from the given instruction."""
-    # example:
-    #
-    #     push    3136B0h         ; dwControlCode
-    operands = [o.strip() for o in insn.operands.split(",")]
-    if insn.mnemonic == "add" and operands[0] in ["esp", "rsp"]:
-        # skip things like:
-        #
-        #    .text:00401140                 call    sub_407E2B
-        #    .text:00401145                 add     esp, 0Ch
-        return
-    for operand in operands:
-        try:
-            yield Number(int(operand, 16)), insn.offset
-            yield Number(int(operand, 16), bitness=get_bitness(f.smda_report)), insn.offset
-        except:
-            continue
-
-
-def read_bytes(smda_report, va, num_bytes=None):
-    """
-    read up to MAX_BYTES_FEATURE_SIZE from the given address.
-    """
-
-    rva = va - smda_report.base_addr
-    if smda_report.buffer is None:
-        raise ValueError("buffer is empty")
-    buffer_end = len(smda_report.buffer)
-    max_bytes = num_bytes if num_bytes is not None else MAX_BYTES_FEATURE_SIZE
-    if rva + max_bytes > buffer_end:
-        return smda_report.buffer[rva:]
-    else:
-        return smda_report.buffer[rva : rva + max_bytes]
-
-
-def derefs(smda_report, p):
-    """
-    recursively follow the given pointer, yielding the valid memory addresses along the way.
-    useful when you may have a pointer to string, or pointer to pointer to string, etc.
-
-    this is a "do what i mean" type of helper function.
-
-    based on the implementation in viv/insn.py
-    """
-    depth = 0
-    while True:
-        if not smda_report.isAddrWithinMemoryImage(p):
-            return
-        yield p
-
-        bytes_ = read_bytes(smda_report, p, num_bytes=4)
-        val = struct.unpack("I", bytes_)[0]
-
-        # sanity: pointer points to self
-        if val == p:
-            return
-
-        # sanity: avoid chains of pointers that are unreasonably deep
-        depth += 1
-        if depth > 10:
-            return
-
-        p = val
-
-
-def extract_insn_bytes_features(f, bb, insn):
-    """
-    parse byte sequence features from the given instruction.
-    example:
-        #     push    offset iid_004118d4_IShellLinkA ; riid
-    """
-    for data_ref in insn.getDataRefs():
-        for v in derefs(f.smda_report, data_ref):
-            bytes_read = read_bytes(f.smda_report, v)
-            if bytes_read is None:
-                continue
-            if capa.features.extractors.helpers.all_zeros(bytes_read):
-                continue
-
-            yield Bytes(bytes_read), insn.offset
-
-
-def detect_ascii_len(smda_report, offset):
-    if smda_report.buffer is None:
-        return 0
-    ascii_len = 0
-    rva = offset - smda_report.base_addr
-    char = smda_report.buffer[rva]
-    while char < 127 and chr(char) in string.printable:
-        ascii_len += 1
-        rva += 1
-        char = smda_report.buffer[rva]
-    if char == 0:
-        return ascii_len
-    return 0
-
-
-def detect_unicode_len(smda_report, offset):
-    if smda_report.buffer is None:
-        return 0
-    unicode_len = 0
-    rva = offset - smda_report.base_addr
-    char = smda_report.buffer[rva]
-    second_char = smda_report.buffer[rva + 1]
-    while char < 127 and chr(char) in string.printable and second_char == 0:
-        unicode_len += 2
-        rva += 2
-        char = smda_report.buffer[rva]
-        second_char = smda_report.buffer[rva + 1]
-    if char == 0 and second_char == 0:
-        return unicode_len
-    return 0
-
-
-def read_string(smda_report, offset):
-    alen = detect_ascii_len(smda_report, offset)
-    if alen > 1:
-        return read_bytes(smda_report, offset, alen).decode("utf-8")
-    ulen = detect_unicode_len(smda_report, offset)
-    if ulen > 2:
-        return read_bytes(smda_report, offset, ulen).decode("utf-16")
-
-
-def extract_insn_string_features(f, bb, insn):
-    """parse string features from the given instruction."""
-    # example:
-    #
-    #     push    offset aAcr     ; "ACR  > "
-    for data_ref in insn.getDataRefs():
-        for v in derefs(f.smda_report, data_ref):
-            string_read = read_string(f.smda_report, v)
-            if string_read:
-                yield String(string_read.rstrip("\x00")), insn.offset
-
-
-def extract_insn_offset_features(f, bb, insn):
-    """parse structure offset features from the given instruction."""
-    # examples:
-    #
-    #     mov eax, [esi + 4]
-    #     mov eax, [esi + ecx + 16384]
-    operands = [o.strip() for o in insn.operands.split(",")]
-    for operand in operands:
-        if not "ptr" in operand:
-            continue
-        if "esp" in operand or "ebp" in operand or "rbp" in operand:
-            continue
-        number = 0
-        number_hex = re.search(PATTERN_HEXNUM, operand)
-        number_int = re.search(PATTERN_SINGLENUM, operand)
-        if number_hex:
-            number = int(number_hex.group("num"), 16)
-            number = -1 * number if number_hex.group().startswith("-") else number
-        elif number_int:
-            number = int(number_int.group("num"))
-            number = -1 * number if number_int.group().startswith("-") else number
-        yield Offset(number), insn.offset
-        yield Offset(number, bitness=get_bitness(f.smda_report)), insn.offset
-
-
-def is_security_cookie(f, bb, insn):
-    """
-    check if an instruction is related to security cookie checks
-    """
-    # security cookie check should use SP or BP
-    operands = [o.strip() for o in insn.operands.split(",")]
-    if operands[1] not in ["esp", "ebp", "rsp", "rbp"]:
-        return False
-    for index, block in enumerate(f.getBlocks()):
-        # expect security cookie init in first basic block within first bytes (instructions)
-        block_instructions = [i for i in block.getInstructions()]
-        if index == 0 and insn.offset < (block_instructions[0].offset + SECURITY_COOKIE_BYTES_DELTA):
-            return True
-        # ... or within last bytes (instructions) before a return
-        if block_instructions[-1].mnemonic.startswith("ret") and insn.offset > (
-            block_instructions[-1].offset - SECURITY_COOKIE_BYTES_DELTA
-        ):
-            return True
-    return False
-
-
-def extract_insn_nzxor_characteristic_features(f, bb, insn):
-    """
-    parse non-zeroing XOR instruction from the given instruction.
-    ignore expected non-zeroing XORs, e.g. security cookies.
-    """
-
-    if insn.mnemonic not in ("xor", "xorpd", "xorps", "pxor"):
-        return
-
-    operands = [o.strip() for o in insn.operands.split(",")]
-    if operands[0] == operands[1]:
-        return
-
-    if is_security_cookie(f, bb, insn):
-        return
-
-    yield Characteristic("nzxor"), insn.offset
-
-
-def extract_insn_mnemonic_features(f, bb, insn):
-    """parse mnemonic features from the given instruction."""
-    yield Mnemonic(insn.mnemonic), insn.offset
-
-
-def extract_insn_peb_access_characteristic_features(f, bb, insn):
-    """
-    parse peb access from the given function. fs:[0x30] on x86, gs:[0x60] on x64
-    """
-
-    if insn.mnemonic not in ["push", "mov"]:
-        return
-
-    operands = [o.strip() for o in insn.operands.split(",")]
-    for operand in operands:
-        if "fs:" in operand and "0x30" in operand:
-            yield Characteristic("peb access"), insn.offset
-        elif "gs:" in operand and "0x60" in operand:
-            yield Characteristic("peb access"), insn.offset
-
-
-def extract_insn_segment_access_features(f, bb, insn):
-    """parse the instruction for access to fs or gs"""
-    operands = [o.strip() for o in insn.operands.split(",")]
-    for operand in operands:
-        if "fs:" in operand:
-            yield Characteristic("fs access"), insn.offset
-        elif "gs:" in operand:
-            yield Characteristic("gs access"), insn.offset
-
-
-def extract_insn_cross_section_cflow(f, bb, insn):
-    """
-    inspect the instruction for a CALL or JMP that crosses section boundaries.
-    """
-    if insn.mnemonic in ["call", "jmp"]:
-        if insn.offset in f.apirefs:
-            return
-
-        smda_report = insn.smda_function.smda_report
-        if insn.offset in f.outrefs:
-            for target in f.outrefs[insn.offset]:
-                if smda_report.getSection(insn.offset) != smda_report.getSection(target):
-                    yield Characteristic("cross section flow"), insn.offset
-        elif insn.operands.startswith("0x"):
-            target = int(insn.operands, 16)
-            if smda_report.getSection(insn.offset) != smda_report.getSection(target):
-                yield Characteristic("cross section flow"), insn.offset
-
-
-# this is a feature that's most relevant at the function scope,
-# however, its most efficient to extract at the instruction scope.
-def extract_function_calls_from(f, bb, insn):
-    if insn.mnemonic != "call":
-        return
-
-    if insn.offset in f.outrefs:
-        for outref in f.outrefs[insn.offset]:
-            yield Characteristic("calls from"), outref
-
-            if outref == f.offset:
-                # if we found a jump target and it's the function address
-                # mark as recursive
-                yield Characteristic("recursive call"), outref
-    if insn.offset in f.apirefs:
-        yield Characteristic("calls from"), insn.offset
-
-
-# this is a feature that's most relevant at the function or basic block scope,
-# however, its most efficient to extract at the instruction scope.
-def extract_function_indirect_call_characteristic_features(f, bb, insn):
-    """
-    extract indirect function call characteristic (e.g., call eax or call dword ptr [edx+4])
-    does not include calls like => call ds:dword_ABD4974
-    """
-    if insn.mnemonic != "call":
-        return
-    if insn.operands.startswith("0x"):
-        return False
-    if "qword ptr" in insn.operands and "rip" in insn.operands:
-        return False
-    if insn.operands.startswith("dword ptr [0x"):
-        return False
-    # call edx
-    # call dword ptr [eax+50h]
-    # call qword ptr [rsp+78h]
-    yield Characteristic("indirect call"), insn.offset
-
-
-def extract_features(f, bb, insn):
-    """
-    extract features from the given insn.
-
-    args:
-      f (smda.common.SmdaFunction): the function to process.
-      bb (smda.common.SmdaBasicBlock): the basic block to process.
-      insn (smda.common.SmdaInstruction): the instruction to process.
-
-    yields:
-      Tuple[Feature, int]: the features and their location found in this insn.
-    """
-    for insn_handler in INSTRUCTION_HANDLERS:
-        for feature, va in insn_handler(f, bb, insn):
-            yield feature, va
-
-
-INSTRUCTION_HANDLERS = (
-    extract_insn_api_features,
-    extract_insn_number_features,
-    extract_insn_string_features,
-    extract_insn_bytes_features,
-    extract_insn_offset_features,
-    extract_insn_nzxor_characteristic_features,
-    extract_insn_mnemonic_features,
-    extract_insn_peb_access_characteristic_features,
-    extract_insn_cross_section_cflow,
-    extract_insn_segment_access_features,
-    extract_function_calls_from,
-    extract_function_indirect_call_characteristic_features,
-)

capa/features/extractors/strings.py

@@ -1,6 +1,6 @@
-# strings code from FLOSS, https://github.com/fireeye/flare-floss
+# strings code from FLOSS, https://github.com/mandiant/flare-floss
 #
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt

capa/features/extractors/viv/basicblock.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -8,27 +8,30 @@
 
 import string
 import struct
+from typing import Tuple, Iterator
 
 import envi
 import envi.archs.i386.disasm
 
-from capa.features.common import Characteristic
+from capa.features.common import Feature, Characteristic
+from capa.features.address import Address, AbsoluteVirtualAddress
 from capa.features.basicblock import BasicBlock
 from capa.features.extractors.helpers import MIN_STACKSTRING_LEN
+from capa.features.extractors.base_extractor import BBHandle, FunctionHandle
 
 
-def interface_extract_basic_block_XXX(f, bb):
+def interface_extract_basic_block_XXX(f: FunctionHandle, bb: BBHandle) -> Iterator[Tuple[Feature, Address]]:
     """
     parse features from the given basic block.
 
     args:
-      f (viv_utils.Function): the function to process.
-      bb (viv_utils.BasicBlock): the basic block to process.
+      f: the function to process.
+      bb: the basic block to process.
 
     yields:
-      (Feature, int): the feature and the address at which its found.
+      (Feature, Address): the feature and the address at which its found.
     """
-    yield NotImplementedError("feature"), NotImplementedError("virtual address")
+    raise NotImplementedError
 
 
 def _bb_has_tight_loop(f, bb):
@@ -44,10 +47,10 @@ def _bb_has_tight_loop(f, bb):
     return False
 
 
-def extract_bb_tight_loop(f, bb):
+def extract_bb_tight_loop(f: FunctionHandle, bb: BBHandle) -> Iterator[Tuple[Feature, Address]]:
     """check basic block for tight loop indicators"""
-    if _bb_has_tight_loop(f, bb):
-        yield Characteristic("tight loop"), bb.va
+    if _bb_has_tight_loop(f, bb.inner):
+        yield Characteristic("tight loop"), bb.address
 
 
 def _bb_has_stackstring(f, bb):
@@ -67,10 +70,10 @@ def _bb_has_stackstring(f, bb):
     return False
 
 
-def extract_stackstring(f, bb):
+def extract_stackstring(f: FunctionHandle, bb: BBHandle) -> Iterator[Tuple[Feature, Address]]:
     """check basic block for stackstring indicators"""
-    if _bb_has_stackstring(f, bb):
-        yield Characteristic("stack string"), bb.va
+    if _bb_has_stackstring(f, bb.inner):
+        yield Characteristic("stack string"), bb.address
 
 
 def is_mov_imm_to_stack(instr: envi.archs.i386.disasm.i386Opcode) -> bool:
@@ -143,7 +146,7 @@ def is_printable_utf16le(chars: bytes) -> bool:
     return False
 
 
-def extract_features(f, bb):
+def extract_features(f: FunctionHandle, bb: BBHandle) -> Iterator[Tuple[Feature, Address]]:
     """
     extract features from the given basic block.
 
@@ -154,10 +157,10 @@ def extract_features(f, bb):
     yields:
       Tuple[Feature, int]: the features and their location found in this basic block.
     """
-    yield BasicBlock(), bb.va
+    yield BasicBlock(), AbsoluteVirtualAddress(bb.inner.va)
     for bb_handler in BASIC_BLOCK_HANDLERS:
-        for feature, va in bb_handler(f, bb):
-            yield feature, va
+        for feature, addr in bb_handler(f, bb):
+            yield feature, addr
 
 
 BASIC_BLOCK_HANDLERS = (

capa/features/extractors/viv/extractor.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -6,6 +6,7 @@
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
 import logging
+from typing import List, Tuple, Iterator
 
 import viv_utils
 import viv_utils.flirt
@@ -16,40 +17,30 @@ import capa.features.extractors.viv.insn
 import capa.features.extractors.viv.global_
 import capa.features.extractors.viv.function
 import capa.features.extractors.viv.basicblock
-from capa.features.extractors.base_extractor import FeatureExtractor
+from capa.features.common import Feature
+from capa.features.address import Address, AbsoluteVirtualAddress
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle, FeatureExtractor
 
 logger = logging.getLogger(__name__)
 
 
-class InstructionHandle:
-    """this acts like a vivisect.Opcode but with an __int__() method"""
-
-    def __init__(self, inner):
-        self._inner = inner
-
-    def __int__(self):
-        return self.va
-
-    def __getattr__(self, name):
-        return getattr(self._inner, name)
-
-
 class VivisectFeatureExtractor(FeatureExtractor):
     def __init__(self, vw, path):
-        super(VivisectFeatureExtractor, self).__init__()
+        super().__init__()
         self.vw = vw
         self.path = path
         with open(self.path, "rb") as f:
             self.buf = f.read()
 
         # pre-compute these because we'll yield them at *every* scope.
-        self.global_features = []
+        self.global_features: List[Tuple[Feature, Address]] = []
+        self.global_features.extend(capa.features.extractors.viv.file.extract_file_format(self.buf))
         self.global_features.extend(capa.features.extractors.common.extract_os(self.buf))
         self.global_features.extend(capa.features.extractors.viv.global_.extract_arch(self.vw))
 
     def get_base_address(self):
         # assume there is only one file loaded into the vw
-        return list(self.vw.filemeta.values())[0]["imagebase"]
+        return AbsoluteVirtualAddress(list(self.vw.filemeta.values())[0]["imagebase"])
 
     def extract_global_features(self):
         yield from self.global_features
@@ -57,28 +48,33 @@ class VivisectFeatureExtractor(FeatureExtractor):
     def extract_file_features(self):
         yield from capa.features.extractors.viv.file.extract_features(self.vw, self.buf)
 
-    def get_functions(self):
+    def get_functions(self) -> Iterator[FunctionHandle]:
         for va in sorted(self.vw.getFunctions()):
-            yield viv_utils.Function(self.vw, va)
+            yield FunctionHandle(address=AbsoluteVirtualAddress(va), inner=viv_utils.Function(self.vw, va))
 
-    def extract_function_features(self, f):
-        yield from capa.features.extractors.viv.function.extract_features(f)
+    def extract_function_features(self, fh: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.viv.function.extract_features(fh)
 
-    def get_basic_blocks(self, f):
-        return f.basic_blocks
+    def get_basic_blocks(self, fh: FunctionHandle) -> Iterator[BBHandle]:
+        f: viv_utils.Function = fh.inner
+        for bb in f.basic_blocks:
+            yield BBHandle(address=AbsoluteVirtualAddress(bb.va), inner=bb)
 
-    def extract_basic_block_features(self, f, bb):
-        yield from capa.features.extractors.viv.basicblock.extract_features(f, bb)
+    def extract_basic_block_features(self, fh: FunctionHandle, bbh) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.viv.basicblock.extract_features(fh, bbh)
 
-    def get_instructions(self, f, bb):
+    def get_instructions(self, fh: FunctionHandle, bbh: BBHandle) -> Iterator[InsnHandle]:
+        bb: viv_utils.BasicBlock = bbh.inner
         for insn in bb.instructions:
-            yield InstructionHandle(insn)
+            yield InsnHandle(address=AbsoluteVirtualAddress(insn.va), inner=insn)
 
-    def extract_insn_features(self, f, bb, insn):
-        yield from capa.features.extractors.viv.insn.extract_features(f, bb, insn)
+    def extract_insn_features(
+        self, fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+    ) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.viv.insn.extract_features(fh, bbh, ih)
 
-    def is_library_function(self, va):
-        return viv_utils.flirt.is_library_function(self.vw, va)
+    def is_library_function(self, addr):
+        return viv_utils.flirt.is_library_function(self.vw, addr)
 
-    def get_function_name(self, va):
-        return viv_utils.get_function_name(self.vw, va)
+    def get_function_name(self, addr):
+        return viv_utils.get_function_name(self.vw, addr)

capa/features/extractors/viv/file.py

@@ -1,10 +1,11 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
+from typing import Tuple, Iterator
 
 import PE.carve as pe_carve  # vivisect PE
 import viv_utils
@@ -15,20 +16,21 @@ import capa.features.extractors.common
 import capa.features.extractors.helpers
 import capa.features.extractors.strings
 from capa.features.file import Export, Import, Section, FunctionName
-from capa.features.common import String, Characteristic
+from capa.features.common import String, Feature, Characteristic
+from capa.features.address import Address, FileOffsetAddress, AbsoluteVirtualAddress
 
 
-def extract_file_embedded_pe(buf, **kwargs):
+def extract_file_embedded_pe(buf, **kwargs) -> Iterator[Tuple[Feature, Address]]:
     for offset, _ in pe_carve.carve(buf, 1):
-        yield Characteristic("embedded pe"), offset
+        yield Characteristic("embedded pe"), FileOffsetAddress(offset)
 
 
-def extract_file_export_names(vw, **kwargs):
+def extract_file_export_names(vw, **kwargs) -> Iterator[Tuple[Feature, Address]]:
     for va, _, name, _ in vw.getExports():
-        yield Export(name), va
+        yield Export(name), AbsoluteVirtualAddress(va)
 
 
-def extract_file_import_names(vw, **kwargs):
+def extract_file_import_names(vw, **kwargs) -> Iterator[Tuple[Feature, Address]]:
     """
     extract imported function names
     1. imports by ordinal:
@@ -44,8 +46,9 @@ def extract_file_import_names(vw, **kwargs):
             # replace ord prefix with #
             impname = "#%s" % impname[len("ord") :]
 
+        addr = AbsoluteVirtualAddress(va)
         for name in capa.features.extractors.helpers.generate_symbols(modname, impname):
-            yield Import(name), va
+            yield Import(name), addr
 
 
 def is_viv_ord_impname(impname: str) -> bool:
@@ -62,30 +65,37 @@ def is_viv_ord_impname(impname: str) -> bool:
         return True
 
 
-def extract_file_section_names(vw, **kwargs):
+def extract_file_section_names(vw, **kwargs) -> Iterator[Tuple[Feature, Address]]:
     for va, _, segname, _ in vw.getSegments():
-        yield Section(segname), va
+        yield Section(segname), AbsoluteVirtualAddress(va)
 
 
-def extract_file_strings(buf, **kwargs):
+def extract_file_strings(buf, **kwargs) -> Iterator[Tuple[Feature, Address]]:
     yield from capa.features.extractors.common.extract_file_strings(buf)
 
 
-def extract_file_function_names(vw, **kwargs):
+def extract_file_function_names(vw, **kwargs) -> Iterator[Tuple[Feature, Address]]:
     """
     extract the names of statically-linked library functions.
     """
     for va in sorted(vw.getFunctions()):
+        addr = AbsoluteVirtualAddress(va)
         if viv_utils.flirt.is_library_function(vw, va):
             name = viv_utils.get_function_name(vw, va)
-            yield FunctionName(name), va
+            yield FunctionName(name), addr
+            if name.startswith("_"):
+                # some linkers may prefix linked routines with a `_` to avoid name collisions.
+                # extract features for both the mangled and un-mangled representations.
+                # e.g. `_fwrite` -> `fwrite`
+                # see: https://stackoverflow.com/a/2628384/87207
+                yield FunctionName(name[1:]), addr
 
 
-def extract_file_format(buf, **kwargs):
+def extract_file_format(buf, **kwargs) -> Iterator[Tuple[Feature, Address]]:
     yield from capa.features.extractors.common.extract_format(buf)
 
 
-def extract_features(vw, buf: bytes):
+def extract_features(vw, buf: bytes) -> Iterator[Tuple[Feature, Address]]:
     """
     extract file features from given workspace
 
@@ -94,12 +104,12 @@ def extract_features(vw, buf: bytes):
       buf: the raw input file bytes
 
     yields:
-      Tuple[Feature, VA]: a feature and its location.
+      Tuple[Feature, Address]: a feature and its location.
     """
 
     for file_handler in FILE_HANDLERS:
-        for feature, va in file_handler(vw=vw, buf=buf):  # type: ignore
-            yield feature, va
+        for feature, addr in file_handler(vw=vw, buf=buf):  # type: ignore
+            yield feature, addr
 
 
 FILE_HANDLERS = (

capa/features/extractors/viv/function.py

@@ -1,40 +1,47 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
+from typing import Tuple, Iterator
 
 import envi
+import viv_utils
 import vivisect.const
 
-from capa.features.common import Characteristic
+from capa.features.common import Feature, Characteristic
+from capa.features.address import Address, AbsoluteVirtualAddress
 from capa.features.extractors import loops
+from capa.features.extractors.base_extractor import FunctionHandle
 
 
-def interface_extract_function_XXX(f):
+def interface_extract_function_XXX(fh: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
     """
     parse features from the given function.
 
     args:
-      f (viv_utils.Function): the function to process.
+      f: the function to process.
 
     yields:
-      (Feature, int): the feature and the address at which its found.
+      (Feature, Address): the feature and the address at which its found.
     """
-    yield NotImplementedError("feature"), NotImplementedError("virtual address")
+    raise NotImplementedError
 
 
-def extract_function_calls_to(f):
+def extract_function_calls_to(fhandle: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
+    f: viv_utils.Function = fhandle.inner
     for src, _, _, _ in f.vw.getXrefsTo(f.va, rtype=vivisect.const.REF_CODE):
-        yield Characteristic("calls to"), src
+        yield Characteristic("calls to"), AbsoluteVirtualAddress(src)
 
 
-def extract_function_loop(f):
+def extract_function_loop(fhandle: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
     """
     parse if a function has a loop
     """
+    f: viv_utils.Function = fhandle.inner
+
     edges = []
 
     for bb in f.basic_blocks:
@@ -50,22 +57,22 @@ def extract_function_loop(f):
                     edges.append((bb.va, bva))
 
     if edges and loops.has_loop(edges):
-        yield Characteristic("loop"), f.va
+        yield Characteristic("loop"), fhandle.address
 
 
-def extract_features(f):
+def extract_features(fh: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
     """
     extract features from the given function.
 
     args:
-      f (viv_utils.Function): the function from which to extract features
+      fh: the function handle from which to extract features
 
     yields:
       Tuple[Feature, int]: the features and their location found in this function.
     """
     for func_handler in FUNCTION_HANDLERS:
-        for feature, va in func_handler(f):
-            yield feature, va
+        for feature, addr in func_handler(fh):
+            yield feature, addr
 
 
 FUNCTION_HANDLERS = (extract_function_calls_to, extract_function_loop)

capa/features/extractors/viv/global_.py

@@ -1,19 +1,21 @@
 import logging
+from typing import Tuple, Iterator
 
 import envi.archs.i386
 import envi.archs.amd64
 
-from capa.features.common import ARCH_I386, ARCH_AMD64, Arch
+from capa.features.common import ARCH_I386, ARCH_AMD64, Arch, Feature
+from capa.features.address import NO_ADDRESS, Address
 
 logger = logging.getLogger(__name__)
 
 
-def extract_arch(vw):
+def extract_arch(vw) -> Iterator[Tuple[Feature, Address]]:
     if isinstance(vw.arch, envi.archs.amd64.Amd64Module):
-        yield Arch(ARCH_AMD64), 0x0
+        yield Arch(ARCH_AMD64), NO_ADDRESS
 
     elif isinstance(vw.arch, envi.archs.i386.i386Module):
-        yield Arch(ARCH_I386), 0x0
+        yield Arch(ARCH_I386), NO_ADDRESS
 
     else:
         # we likely end up here:

capa/features/extractors/viv/helpers.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt

capa/features/extractors/viv/indirect_calls.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -7,7 +7,7 @@
 # See the License for the specific language governing permissions and limitations under the License.
 
 import collections
-from typing import TYPE_CHECKING, Set, List, Deque, Tuple, Union, Optional
+from typing import Set, List, Deque, Tuple, Union, Optional
 
 import envi
 import vivisect.const
@@ -15,9 +15,6 @@ import envi.archs.i386.disasm
 import envi.archs.amd64.disasm
 from vivisect import VivWorkspace
 
-if TYPE_CHECKING:
-    from capa.features.extractors.viv.extractor import InstructionHandle
-
 # pull out consts for lookup performance
 i386RegOper = envi.archs.i386.disasm.i386RegOper
 i386ImmOper = envi.archs.i386.disasm.i386ImmOper
@@ -45,7 +42,7 @@ def get_previous_instructions(vw: VivWorkspace, va: int) -> List[int]:
     ret = []
 
     # find the immediate prior instruction.
-    # ensure that it fallsthrough to this one.
+    # ensure that it falls through to this one.
     loc = vw.getPrevLocation(va, adjacent=True)
     if loc is not None:
         ploc = vw.getPrevLocation(va, adjacent=True)
@@ -62,7 +59,7 @@ def get_previous_instructions(vw: VivWorkspace, va: int) -> List[int]:
     #
     # from vivisect.const:
     # xref: (XR_FROM, XR_TO, XR_RTYPE, XR_RFLAG)
-    for (xfrom, _, _, xflag) in vw.getXrefsTo(va, REF_CODE):
+    for xfrom, _, _, xflag in vw.getXrefsTo(va, REF_CODE):
         if (xflag & FAR_BRANCH_MASK) != 0:
             continue
         ret.append(xfrom)
@@ -135,16 +132,14 @@ def find_definition(vw: VivWorkspace, va: int, reg: int) -> Tuple[int, Union[int
     raise NotFoundError()
 
 
-def is_indirect_call(vw: VivWorkspace, va: int, insn: Optional["InstructionHandle"] = None) -> bool:
+def is_indirect_call(vw: VivWorkspace, va: int, insn: envi.Opcode) -> bool:
     if insn is None:
         insn = vw.parseOpcode(va)
 
     return insn.mnem in ("call", "jmp") and isinstance(insn.opers[0], envi.archs.i386.disasm.i386RegOper)
 
 
-def resolve_indirect_call(
-    vw: VivWorkspace, va: int, insn: Optional["InstructionHandle"] = None
-) -> Tuple[int, Optional[int]]:
+def resolve_indirect_call(vw: VivWorkspace, va: int, insn: envi.Opcode) -> Tuple[int, Optional[int]]:
     """
     inspect the given indirect call instruction and attempt to resolve the target address.
 

capa/features/extractors/viv/insn.py

@@ -1,10 +1,12 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
+from typing import List, Tuple, Callable, Iterator
+
 import envi
 import envi.exc
 import viv_utils
@@ -17,16 +19,10 @@ import envi.archs.amd64.disasm
 
 import capa.features.extractors.helpers
 import capa.features.extractors.viv.helpers
-from capa.features.insn import API, Number, Offset, Mnemonic
-from capa.features.common import (
-    BITNESS_X32,
-    BITNESS_X64,
-    MAX_BYTES_FEATURE_SIZE,
-    THUNK_CHAIN_DEPTH_DELTA,
-    Bytes,
-    String,
-    Characteristic,
-)
+from capa.features.insn import API, MAX_STRUCTURE_SIZE, Number, Offset, Mnemonic, OperandNumber, OperandOffset
+from capa.features.common import MAX_BYTES_FEATURE_SIZE, THUNK_CHAIN_DEPTH_DELTA, Bytes, String, Feature, Characteristic
+from capa.features.address import Address, AbsoluteVirtualAddress
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle
 from capa.features.extractors.viv.indirect_calls import NotFoundError, resolve_indirect_call
 
 # security cookie checks may perform non-zeroing XORs, these are expected within a certain
@@ -34,27 +30,21 @@ from capa.features.extractors.viv.indirect_calls import NotFoundError, resolve_i
 SECURITY_COOKIE_BYTES_DELTA = 0x40
 
 
-def get_bitness(vw):
-    bitness = vw.getMeta("Architecture")
-    if bitness == "i386":
-        return BITNESS_X32
-    elif bitness == "amd64":
-        return BITNESS_X64
-
-
-def interface_extract_instruction_XXX(f, bb, insn):
+def interface_extract_instruction_XXX(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
     """
     parse features from the given instruction.
 
     args:
-      f (viv_utils.Function): the function to process.
-      bb (viv_utils.BasicBlock): the basic block to process.
-      insn (vivisect...Instruction): the instruction to process.
+      fh: the function handle to process.
+      bbh: the basic block handle to process.
+      ih: the instruction handle to process.
 
     yields:
-      (Feature, int): the feature and the address at which its found.
+      (Feature, Address): the feature and the address at which its found.
     """
-    yield NotImplementedError("feature"), NotImplementedError("virtual address")
+    raise NotImplementedError
 
 
 def get_imports(vw):
@@ -74,12 +64,15 @@ def get_imports(vw):
         return imports
 
 
-def extract_insn_api_features(f, bb, insn):
-    """parse API features from the given instruction."""
+def extract_insn_api_features(fh: FunctionHandle, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """
+    parse API features from the given instruction.
 
-    # example:
-    #
-    #    call dword [0x00473038]
+    example:
+       call dword [0x00473038]
+    """
+    insn: envi.Opcode = ih.inner
+    f: viv_utils.Function = fh.inner
     if insn.mnem not in ("call", "jmp"):
         return
 
@@ -96,7 +89,7 @@ def extract_insn_api_features(f, bb, insn):
         if target in imports:
             dll, symbol = imports[target]
             for name in capa.features.extractors.helpers.generate_symbols(dll, symbol):
-                yield API(name), insn.va
+                yield API(name), ih.address
 
     # call via thunk on x86,
     # see 9324d1a8ae37a36ae560c37448c9705a at 0x407985
@@ -118,14 +111,20 @@ def extract_insn_api_features(f, bb, insn):
 
         if viv_utils.flirt.is_library_function(f.vw, target):
             name = viv_utils.get_function_name(f.vw, target)
-            yield API(name), insn.va
+            yield API(name), ih.address
+            if name.startswith("_"):
+                # some linkers may prefix linked routines with a `_` to avoid name collisions.
+                # extract features for both the mangled and un-mangled representations.
+                # e.g. `_fwrite` -> `fwrite`
+                # see: https://stackoverflow.com/a/2628384/87207
+                yield API(name[1:]), ih.address
             return
 
         for _ in range(THUNK_CHAIN_DEPTH_DELTA):
             if target in imports:
                 dll, symbol = imports[target]
                 for name in capa.features.extractors.helpers.generate_symbols(dll, symbol):
-                    yield API(name), insn.va
+                    yield API(name), ih.address
 
             # if jump leads to an ENDBRANCH instruction, skip it
             if f.vw.getByteDef(target)[1].startswith(b"\xf3\x0f\x1e"):
@@ -145,7 +144,7 @@ def extract_insn_api_features(f, bb, insn):
         if target in imports:
             dll, symbol = imports[target]
             for name in capa.features.extractors.helpers.generate_symbols(dll, symbol):
-                yield API(name), insn.va
+                yield API(name), ih.address
 
     elif isinstance(insn.opers[0], envi.archs.i386.disasm.i386RegOper):
         try:
@@ -162,38 +161,7 @@ def extract_insn_api_features(f, bb, insn):
         if target in imports:
             dll, symbol = imports[target]
             for name in capa.features.extractors.helpers.generate_symbols(dll, symbol):
-                yield API(name), insn.va
-
-
-def extract_insn_number_features(f, bb, insn):
-    """parse number features from the given instruction."""
-    # example:
-    #
-    #     push    3136B0h         ; dwControlCode
-    for oper in insn.opers:
-        # this is for both x32 and x64
-        if not isinstance(oper, (envi.archs.i386.disasm.i386ImmOper, envi.archs.i386.disasm.i386ImmMemOper)):
-            continue
-
-        if isinstance(oper, envi.archs.i386.disasm.i386ImmOper):
-            v = oper.getOperValue(oper)
-        else:
-            v = oper.getOperAddr(oper)
-
-        if f.vw.probeMemory(v, 1, envi.memory.MM_READ):
-            # this is a valid address
-            # assume its not also a constant.
-            continue
-
-        if insn.mnem == "add" and insn.opers[0].isReg() and insn.opers[0].reg == envi.archs.i386.regs.REG_ESP:
-            # skip things like:
-            #
-            #    .text:00401140                 call    sub_407E2B
-            #    .text:00401145                 add     esp, 0Ch
-            return
-
-        yield Number(v), insn.va
-        yield Number(v, bitness=get_bitness(f.vw)), insn.va
+                yield API(name), ih.address
 
 
 def derefs(vw, p):
@@ -266,12 +234,15 @@ def read_bytes(vw, va: int) -> bytes:
         raise
 
 
-def extract_insn_bytes_features(f, bb, insn):
+def extract_insn_bytes_features(fh: FunctionHandle, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
     """
     parse byte sequence features from the given instruction.
     example:
         #     push    offset iid_004118d4_IShellLinkA ; riid
     """
+    insn: envi.Opcode = ih.inner
+    f: viv_utils.Function = fh.inner
+
     if insn.mnem == "call":
         return
 
@@ -300,7 +271,11 @@ def extract_insn_bytes_features(f, bb, insn):
             if capa.features.extractors.helpers.all_zeros(buf):
                 continue
 
-            yield Bytes(buf), insn.va
+            if f.vw.isProbablyString(v):
+                # don't extract byte features for obvious strings
+                continue
+
+            yield Bytes(buf), ih.address
 
 
 def read_string(vw, offset: int) -> str:
@@ -310,7 +285,12 @@ def read_string(vw, offset: int) -> str:
         pass
     else:
         if alen > 0:
-            return read_memory(vw, offset, alen).decode("utf-8")
+            buf = read_memory(vw, offset, alen)
+            if b"\x00" in buf:
+                # account for bug #1271.
+                # remove when vivisect is fixed.
+                buf = buf.partition(b"\x00")[0]
+            return buf.decode("utf-8")
 
     try:
         ulen = vw.detectUnicode(offset)
@@ -329,80 +309,13 @@ def read_string(vw, offset: int) -> str:
                 # vivisect seems to mis-detect the end unicode strings
                 # off by two, too short
                 ulen += 2
-            return read_memory(vw, offset, ulen).decode("utf-16")
+            # partition to account for bug #1271.
+            # remove when vivisect is fixed.
+            return read_memory(vw, offset, ulen).decode("utf-16").partition("\x00")[0]
 
     raise ValueError("not a string", offset)
 
 
-def extract_insn_string_features(f, bb, insn):
-    """parse string features from the given instruction."""
-    # example:
-    #
-    #     push    offset aAcr     ; "ACR  > "
-
-    for oper in insn.opers:
-        if isinstance(oper, envi.archs.i386.disasm.i386ImmOper):
-            v = oper.getOperValue(oper)
-        elif isinstance(oper, envi.archs.i386.disasm.i386ImmMemOper):
-            # like 0x10056CB4 in `lea eax, dword [0x10056CB4]`
-            v = oper.imm
-        elif isinstance(oper, envi.archs.i386.disasm.i386SibOper):
-            # like 0x401000 in `mov eax, 0x401000[2 * ebx]`
-            v = oper.imm
-        elif isinstance(oper, envi.archs.amd64.disasm.Amd64RipRelOper):
-            v = oper.getOperAddr(insn)
-        else:
-            continue
-
-        for v in derefs(f.vw, v):
-            try:
-                s = read_string(f.vw, v)
-            except ValueError:
-                continue
-            else:
-                yield String(s.rstrip("\x00")), insn.va
-
-
-def extract_insn_offset_features(f, bb, insn):
-    """parse structure offset features from the given instruction."""
-    # example:
-    #
-    #     .text:0040112F    cmp     [esi+4], ebx
-    for oper in insn.opers:
-
-        # this is for both x32 and x64
-        # like [esi + 4]
-        #       reg   ^
-        #             disp
-        if isinstance(oper, envi.archs.i386.disasm.i386RegMemOper):
-            if oper.reg == envi.archs.i386.regs.REG_ESP:
-                continue
-
-            if oper.reg == envi.archs.i386.regs.REG_EBP:
-                continue
-
-            # TODO: do x64 support for real.
-            if oper.reg == envi.archs.amd64.regs.REG_RBP:
-                continue
-
-            # viv already decodes offsets as signed
-            v = oper.disp
-
-            yield Offset(v), insn.va
-            yield Offset(v, bitness=get_bitness(f.vw)), insn.va
-
-        # like: [esi + ecx + 16384]
-        #        reg   ^     ^
-        #              index ^
-        #                    disp
-        elif isinstance(oper, envi.archs.i386.disasm.i386SibOper):
-            # viv already decodes offsets as signed
-            v = oper.disp
-
-            yield Offset(v), insn.va
-            yield Offset(v, bitness=get_bitness(f.vw)), insn.va
-
-
 def is_security_cookie(f, bb, insn) -> bool:
     """
     check if an instruction is related to security cookie checks
@@ -431,11 +344,17 @@ def is_security_cookie(f, bb, insn) -> bool:
     return False
 
 
-def extract_insn_nzxor_characteristic_features(f, bb, insn):
+def extract_insn_nzxor_characteristic_features(
+    fh: FunctionHandle, bbhandle: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
     """
     parse non-zeroing XOR instruction from the given instruction.
     ignore expected non-zeroing XORs, e.g. security cookies.
     """
+    insn: envi.Opcode = ih.inner
+    bb: viv_utils.BasicBlock = bbhandle.inner
+    f: viv_utils.Function = fh.inner
+
     if insn.mnem not in ("xor", "xorpd", "xorps", "pxor"):
         return
 
@@ -445,19 +364,40 @@ def extract_insn_nzxor_characteristic_features(f, bb, insn):
     if is_security_cookie(f, bb, insn):
         return
 
-    yield Characteristic("nzxor"), insn.va
+    yield Characteristic("nzxor"), ih.address
 
 
-def extract_insn_mnemonic_features(f, bb, insn):
+def extract_insn_mnemonic_features(f, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
     """parse mnemonic features from the given instruction."""
-    yield Mnemonic(insn.mnem), insn.va
+    yield Mnemonic(ih.inner.mnem), ih.address
 
 
-def extract_insn_peb_access_characteristic_features(f, bb, insn):
+def extract_insn_obfs_call_plus_5_characteristic_features(f, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """
+    parse call $+5 instruction from the given instruction.
+    """
+    insn: envi.Opcode = ih.inner
+
+    if insn.mnem != "call":
+        return
+
+    if isinstance(insn.opers[0], envi.archs.i386.disasm.i386PcRelOper):
+        if insn.va + 5 == insn.opers[0].getOperValue(insn):
+            yield Characteristic("call $+5"), ih.address
+
+    if isinstance(insn.opers[0], envi.archs.i386.disasm.i386ImmMemOper) or isinstance(
+        insn.opers[0], envi.archs.amd64.disasm.Amd64RipRelOper
+    ):
+        if insn.va + 5 == insn.opers[0].getOperAddr(insn):
+            yield Characteristic("call $+5"), ih.address
+
+
+def extract_insn_peb_access_characteristic_features(f, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
     """
     parse peb access from the given function. fs:[0x30] on x86, gs:[0x60] on x64
     """
     # TODO handle where fs/gs are loaded into a register or onto the stack and used later
+    insn: envi.Opcode = ih.inner
 
     if insn.mnem not in ["push", "mov"]:
         return
@@ -476,7 +416,7 @@ def extract_insn_peb_access_characteristic_features(f, bb, insn):
             if (isinstance(oper, envi.archs.i386.disasm.i386RegMemOper) and oper.disp == 0x30) or (
                 isinstance(oper, envi.archs.i386.disasm.i386ImmMemOper) and oper.imm == 0x30
             ):
-                yield Characteristic("peb access"), insn.va
+                yield Characteristic("peb access"), ih.address
     elif "gs" in prefix:
         for oper in insn.opers:
             if (
@@ -484,20 +424,22 @@ def extract_insn_peb_access_characteristic_features(f, bb, insn):
                 or (isinstance(oper, envi.archs.amd64.disasm.i386SibOper) and oper.imm == 0x60)
                 or (isinstance(oper, envi.archs.amd64.disasm.i386ImmMemOper) and oper.imm == 0x60)
             ):
-                yield Characteristic("peb access"), insn.va
+                yield Characteristic("peb access"), ih.address
     else:
         pass
 
 
-def extract_insn_segment_access_features(f, bb, insn):
+def extract_insn_segment_access_features(f, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
     """parse the instruction for access to fs or gs"""
+    insn: envi.Opcode = ih.inner
+
     prefix = insn.getPrefixName()
 
     if prefix == "fs":
-        yield Characteristic("fs access"), insn.va
+        yield Characteristic("fs access"), ih.address
 
     if prefix == "gs":
-        yield Characteristic("gs access"), insn.va
+        yield Characteristic("gs access"), ih.address
 
 
 def get_section(vw, va: int):
@@ -508,10 +450,13 @@ def get_section(vw, va: int):
     raise KeyError(va)
 
 
-def extract_insn_cross_section_cflow(f, bb, insn):
+def extract_insn_cross_section_cflow(fh: FunctionHandle, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
     """
     inspect the instruction for a CALL or JMP that crosses section boundaries.
     """
+    insn: envi.Opcode = ih.inner
+    f: viv_utils.Function = fh.inner
+
     for va, flags in insn.getBranches():
         if va is None:
             # va may be none for dynamic branches that haven't been resolved, such as `jmp eax`.
@@ -538,7 +483,7 @@ def extract_insn_cross_section_cflow(f, bb, insn):
                     continue
 
             if get_section(f.vw, insn.va) != get_section(f.vw, va):
-                yield Characteristic("cross section flow"), insn.va
+                yield Characteristic("cross section flow"), ih.address
 
         except KeyError:
             continue
@@ -546,7 +491,10 @@ def extract_insn_cross_section_cflow(f, bb, insn):
 
 # this is a feature that's most relevant at the function scope,
 # however, its most efficient to extract at the instruction scope.
-def extract_function_calls_from(f, bb, insn):
+def extract_function_calls_from(fh: FunctionHandle, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    insn: envi.Opcode = ih.inner
+    f: viv_utils.Function = fh.inner
+
     if insn.mnem != "call":
         return
 
@@ -556,7 +504,8 @@ def extract_function_calls_from(f, bb, insn):
     if isinstance(insn.opers[0], envi.archs.i386.disasm.i386ImmMemOper):
         oper = insn.opers[0]
         target = oper.getOperAddr(insn)
-        yield Characteristic("calls from"), target
+        if target >= 0:
+            yield Characteristic("calls from"), AbsoluteVirtualAddress(target)
 
     # call via thunk on x86,
     # see 9324d1a8ae37a36ae560c37448c9705a at 0x407985
@@ -565,43 +514,193 @@ def extract_function_calls_from(f, bb, insn):
     # see Lab21-01.exe_:0x140001178
     elif isinstance(insn.opers[0], envi.archs.i386.disasm.i386PcRelOper):
         target = insn.opers[0].getOperValue(insn)
-        yield Characteristic("calls from"), target
+        if target >= 0:
+            yield Characteristic("calls from"), AbsoluteVirtualAddress(target)
 
     # call via IAT, x64
     elif isinstance(insn.opers[0], envi.archs.amd64.disasm.Amd64RipRelOper):
         op = insn.opers[0]
         target = op.getOperAddr(insn)
-        yield Characteristic("calls from"), target
+        if target >= 0:
+            yield Characteristic("calls from"), AbsoluteVirtualAddress(target)
 
     if target and target == f.va:
         # if we found a jump target and it's the function address
         # mark as recursive
-        yield Characteristic("recursive call"), target
+        yield Characteristic("recursive call"), AbsoluteVirtualAddress(target)
 
 
 # this is a feature that's most relevant at the function or basic block scope,
 # however, its most efficient to extract at the instruction scope.
-def extract_function_indirect_call_characteristic_features(f, bb, insn):
+def extract_function_indirect_call_characteristic_features(f, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
     """
     extract indirect function call characteristic (e.g., call eax or call dword ptr [edx+4])
     does not include calls like => call ds:dword_ABD4974
     """
+    insn: envi.Opcode = ih.inner
+
     if insn.mnem != "call":
         return
 
     # Checks below work for x86 and x64
     if isinstance(insn.opers[0], envi.archs.i386.disasm.i386RegOper):
         # call edx
-        yield Characteristic("indirect call"), insn.va
+        yield Characteristic("indirect call"), ih.address
     elif isinstance(insn.opers[0], envi.archs.i386.disasm.i386RegMemOper):
         # call dword ptr [eax+50h]
-        yield Characteristic("indirect call"), insn.va
+        yield Characteristic("indirect call"), ih.address
     elif isinstance(insn.opers[0], envi.archs.i386.disasm.i386SibOper):
         # call qword ptr [rsp+78h]
-        yield Characteristic("indirect call"), insn.va
+        yield Characteristic("indirect call"), ih.address
 
 
-def extract_features(f, bb, insn):
+def extract_op_number_features(
+    fh: FunctionHandle, bb, ih: InsnHandle, i, oper: envi.Operand
+) -> Iterator[Tuple[Feature, Address]]:
+    """parse number features from the given operand.
+
+    example:
+        push    3136B0h         ; dwControlCode
+    """
+    insn: envi.Opcode = ih.inner
+    f: viv_utils.Function = fh.inner
+
+    # this is for both x32 and x64
+    if not isinstance(oper, (envi.archs.i386.disasm.i386ImmOper, envi.archs.i386.disasm.i386ImmMemOper)):
+        return
+
+    if isinstance(oper, envi.archs.i386.disasm.i386ImmOper):
+        v = oper.getOperValue(oper)
+    else:
+        v = oper.getOperAddr(oper)
+
+    if f.vw.probeMemory(v, 1, envi.memory.MM_READ):
+        # this is a valid address
+        # assume its not also a constant.
+        return
+
+    if insn.mnem == "add" and insn.opers[0].isReg() and insn.opers[0].reg == envi.archs.i386.regs.REG_ESP:
+        # skip things like:
+        #
+        #    .text:00401140                 call    sub_407E2B
+        #    .text:00401145                 add     esp, 0Ch
+        return
+
+    yield Number(v), ih.address
+    yield OperandNumber(i, v), ih.address
+
+    if insn.mnem == "add" and 0 < v < MAX_STRUCTURE_SIZE and isinstance(oper, envi.archs.i386.disasm.i386ImmOper):
+        # for pattern like:
+        #
+        #     add eax, 0x10
+        #
+        # assume 0x10 is also an offset (imagine eax is a pointer).
+        yield Offset(v), ih.address
+        yield OperandOffset(i, v), ih.address
+
+
+def extract_op_offset_features(
+    fh: FunctionHandle, bb, ih: InsnHandle, i, oper: envi.Operand
+) -> Iterator[Tuple[Feature, Address]]:
+    """parse structure offset features from the given operand."""
+    # example:
+    #
+    #     .text:0040112F    cmp     [esi+4], ebx
+    insn: envi.Opcode = ih.inner
+    f: viv_utils.Function = fh.inner
+
+    # this is for both x32 and x64
+    # like [esi + 4]
+    #       reg   ^
+    #             disp
+    if isinstance(oper, envi.archs.i386.disasm.i386RegMemOper):
+        if oper.reg == envi.archs.i386.regs.REG_ESP:
+            return
+
+        if oper.reg == envi.archs.i386.regs.REG_EBP:
+            return
+
+        # TODO: do x64 support for real.
+        if oper.reg == envi.archs.amd64.regs.REG_RBP:
+            return
+
+        # viv already decodes offsets as signed
+        v = oper.disp
+
+        yield Offset(v), ih.address
+        yield OperandOffset(i, v), ih.address
+
+        if insn.mnem == "lea" and i == 1 and not f.vw.probeMemory(v, 1, envi.memory.MM_READ):
+            # for pattern like:
+            #
+            #     lea eax, [ebx + 1]
+            #
+            # assume 1 is also an offset (imagine ebx is a zero register).
+            yield Number(v), ih.address
+            yield OperandNumber(i, v), ih.address
+
+    # like: [esi + ecx + 16384]
+    #        reg   ^     ^
+    #              index ^
+    #                    disp
+    elif isinstance(oper, envi.archs.i386.disasm.i386SibOper):
+        # viv already decodes offsets as signed
+        v = oper.disp
+
+        yield Offset(v), ih.address
+        yield OperandOffset(i, v), ih.address
+
+
+def extract_op_string_features(
+    fh: FunctionHandle, bb, ih: InsnHandle, i, oper: envi.Operand
+) -> Iterator[Tuple[Feature, Address]]:
+    """parse string features from the given operand."""
+    # example:
+    #
+    #     push    offset aAcr     ; "ACR  > "
+    insn: envi.Opcode = ih.inner
+    f: viv_utils.Function = fh.inner
+
+    if isinstance(oper, envi.archs.i386.disasm.i386ImmOper):
+        v = oper.getOperValue(oper)
+    elif isinstance(oper, envi.archs.i386.disasm.i386ImmMemOper):
+        # like 0x10056CB4 in `lea eax, dword [0x10056CB4]`
+        v = oper.imm
+    elif isinstance(oper, envi.archs.i386.disasm.i386SibOper):
+        # like 0x401000 in `mov eax, 0x401000[2 * ebx]`
+        v = oper.imm
+    elif isinstance(oper, envi.archs.amd64.disasm.Amd64RipRelOper):
+        v = oper.getOperAddr(insn)
+    else:
+        return
+
+    for v in derefs(f.vw, v):
+        try:
+            s = read_string(f.vw, v).rstrip("\x00")
+        except ValueError:
+            continue
+        else:
+            if len(s) >= 4:
+                yield String(s), ih.address
+
+
+def extract_operand_features(f: FunctionHandle, bb, insn: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    for i, oper in enumerate(insn.inner.opers):
+        for op_handler in OPERAND_HANDLERS:
+            for feature, addr in op_handler(f, bb, insn, i, oper):
+                yield feature, addr
+
+
+OPERAND_HANDLERS: List[
+    Callable[[FunctionHandle, BBHandle, InsnHandle, int, envi.Operand], Iterator[Tuple[Feature, Address]]]
+] = [
+    extract_op_number_features,
+    extract_op_offset_features,
+    extract_op_string_features,
+]
+
+
+def extract_features(f, bb, insn) -> Iterator[Tuple[Feature, Address]]:
     """
     extract features from the given insn.
 
@@ -611,24 +710,23 @@ def extract_features(f, bb, insn):
       insn (vivisect...Instruction): the instruction to process.
 
     yields:
-      Tuple[Feature, int]: the features and their location found in this insn.
+      Tuple[Feature, Address]: the features and their location found in this insn.
     """
     for insn_handler in INSTRUCTION_HANDLERS:
-        for feature, va in insn_handler(f, bb, insn):
-            yield feature, va
+        for feature, addr in insn_handler(f, bb, insn):
+            yield feature, addr
 
 
-INSTRUCTION_HANDLERS = (
+INSTRUCTION_HANDLERS: List[Callable[[FunctionHandle, BBHandle, InsnHandle], Iterator[Tuple[Feature, Address]]]] = [
     extract_insn_api_features,
-    extract_insn_number_features,
-    extract_insn_string_features,
     extract_insn_bytes_features,
-    extract_insn_offset_features,
     extract_insn_nzxor_characteristic_features,
     extract_insn_mnemonic_features,
+    extract_insn_obfs_call_plus_5_characteristic_features,
     extract_insn_peb_access_characteristic_features,
     extract_insn_cross_section_cflow,
     extract_insn_segment_access_features,
     extract_function_calls_from,
     extract_function_indirect_call_characteristic_features,
-)
+    extract_operand_features,
+]

capa/features/file.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -12,19 +12,19 @@ from capa.features.common import Feature
 class Export(Feature):
     def __init__(self, value: str, description=None):
         # value is export name
-        super(Export, self).__init__(value, description=description)
+        super().__init__(value, description=description)
 
 
 class Import(Feature):
     def __init__(self, value: str, description=None):
         # value is import name
-        super(Import, self).__init__(value, description=description)
+        super().__init__(value, description=description)
 
 
 class Section(Feature):
     def __init__(self, value: str, description=None):
         # value is section name
-        super(Section, self).__init__(value, description=description)
+        super().__init__(value, description=description)
 
 
 class FunctionName(Feature):
@@ -32,7 +32,7 @@ class FunctionName(Feature):
 
     def __init__(self, name: str, description=None):
         # value is function name
-        super(FunctionName, self).__init__(name, description=description)
+        super().__init__(name, description=description)
         # override the name property set by `capa.features.Feature`
         # that would be `functionname` (note missing dash)
         self.name = "function-name"

capa/features/freeze.py

@@ -1,284 +0,0 @@
-"""
-capa freeze file format: `| capa0000 | + zlib(utf-8(json(...)))`
-
-json format:
-
-    {
-      'version': 1,
-      'base address': int(base address),
-      'functions': {
-        int(function va): {
-          'basic blocks': {
-            int(basic block va): {
-              'instructions': [instruction va, ...]
-            },
-            ...
-          },
-          ...
-        },
-        ...
-      },
-      'scopes': {
-        'global': [
-          (str(name), [any(arg), ...], int(va), ()),
-          ...
-        },
-        'file': [
-          (str(name), [any(arg), ...], int(va), ()),
-          ...
-        },
-        'function': [
-          (str(name), [any(arg), ...], int(va), (int(function va), )),
-          ...
-        ],
-        'basic block': [
-          (str(name), [any(arg), ...], int(va), (int(function va),
-                                                 int(basic block va))),
-          ...
-        ],
-        'instruction': [
-          (str(name), [any(arg), ...], int(va), (int(function va),
-                                                 int(basic block va),
-                                                 int(instruction va))),
-          ...
-        ],
-      }
-    }
-
-Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
-Licensed under the Apache License, Version 2.0 (the "License");
- you may not use this file except in compliance with the License.
-You may obtain a copy of the License at: [package root]/LICENSE.txt
-Unless required by applicable law or agreed to in writing, software distributed under the License
- is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-See the License for the specific language governing permissions and limitations under the License.
-"""
-import json
-import zlib
-import logging
-
-import capa.features.file
-import capa.features.insn
-import capa.features.common
-import capa.features.basicblock
-import capa.features.extractors.base_extractor
-from capa.helpers import hex
-
-logger = logging.getLogger(__name__)
-
-
-def serialize_feature(feature):
-    return feature.freeze_serialize()
-
-
-KNOWN_FEATURES = {F.__name__: F for F in capa.features.common.Feature.__subclasses__()}
-
-
-def deserialize_feature(doc):
-    F = KNOWN_FEATURES[doc[0]]
-    return F.freeze_deserialize(doc[1])
-
-
-def dumps(extractor):
-    """
-    serialize the given extractor to a string
-
-    args:
-      extractor: capa.features.extractors.base_extractor.FeatureExtractor:
-
-    returns:
-      str: the serialized features.
-    """
-    ret = {
-        "version": 1,
-        "base address": extractor.get_base_address(),
-        "functions": {},
-        "scopes": {
-            "global": [],
-            "file": [],
-            "function": [],
-            "basic block": [],
-            "instruction": [],
-        },
-    }
-    for feature, va in extractor.extract_global_features():
-        ret["scopes"]["global"].append(serialize_feature(feature) + (hex(va), ()))
-
-    for feature, va in extractor.extract_file_features():
-        ret["scopes"]["file"].append(serialize_feature(feature) + (hex(va), ()))
-
-    for f in extractor.get_functions():
-        ret["functions"][hex(f)] = {}
-
-        for feature, va in extractor.extract_function_features(f):
-            ret["scopes"]["function"].append(serialize_feature(feature) + (hex(va), (hex(f),)))
-
-        for bb in extractor.get_basic_blocks(f):
-            ret["functions"][hex(f)][hex(bb)] = []
-
-            for feature, va in extractor.extract_basic_block_features(f, bb):
-                ret["scopes"]["basic block"].append(
-                    serialize_feature(feature)
-                    + (
-                        hex(va),
-                        (
-                            hex(f),
-                            hex(bb),
-                        ),
-                    )
-                )
-
-            for insnva, insn in sorted(
-                [(int(insn), insn) for insn in extractor.get_instructions(f, bb)], key=lambda p: p[0]
-            ):
-                ret["functions"][hex(f)][hex(bb)].append(hex(insnva))
-
-                for feature, va in extractor.extract_insn_features(f, bb, insn):
-                    ret["scopes"]["instruction"].append(
-                        serialize_feature(feature)
-                        + (
-                            hex(va),
-                            (
-                                hex(f),
-                                hex(bb),
-                                hex(insnva),
-                            ),
-                        )
-                    )
-    return json.dumps(ret)
-
-
-def loads(s):
-    """deserialize a set of features (as a NullFeatureExtractor) from a string."""
-    doc = json.loads(s)
-
-    if doc.get("version") != 1:
-        raise ValueError("unsupported freeze format version: %d" % (doc.get("version")))
-
-    features = {
-        "base address": doc.get("base address"),
-        "global features": [],
-        "file features": [],
-        "functions": {},
-    }
-
-    for fva, function in doc.get("functions", {}).items():
-        fva = int(fva, 0x10)
-        features["functions"][fva] = {
-            "features": [],
-            "basic blocks": {},
-        }
-
-        for bbva, bb in function.items():
-            bbva = int(bbva, 0x10)
-            features["functions"][fva]["basic blocks"][bbva] = {
-                "features": [],
-                "instructions": {},
-            }
-
-            for insnva in bb:
-                insnva = int(insnva, 0x10)
-                features["functions"][fva]["basic blocks"][bbva]["instructions"][insnva] = {
-                    "features": [],
-                }
-
-    # in the following blocks, each entry looks like:
-    #
-    #     ('MatchedRule', ('foo', ), '0x401000', ('0x401000', ))
-    #      ^^^^^^^^^^^^^  ^^^^^^^^^  ^^^^^^^^^^  ^^^^^^^^^^^^^^
-    #      feature name   args       addr         func/bb/insn
-    for feature in doc.get("scopes", {}).get("global", []):
-        va, loc = feature[2:]
-        va = int(va, 0x10)
-        feature = deserialize_feature(feature[:2])
-        features["global features"].append((va, feature))
-
-    for feature in doc.get("scopes", {}).get("file", []):
-        va, loc = feature[2:]
-        va = int(va, 0x10)
-        feature = deserialize_feature(feature[:2])
-        features["file features"].append((va, feature))
-
-    for feature in doc.get("scopes", {}).get("function", []):
-        # fetch the pair like:
-        #
-        #     ('0x401000', ('0x401000', ))
-        #      ^^^^^^^^^^  ^^^^^^^^^^^^^^
-        #      addr         func/bb/insn
-        va, loc = feature[2:]
-        va = int(va, 0x10)
-        loc = [int(lo, 0x10) for lo in loc]
-
-        # decode the feature from the pair like:
-        #
-        #     ('MatchedRule', ('foo', ))
-        #      ^^^^^^^^^^^^^  ^^^^^^^^^
-        #      feature name   args
-        feature = deserialize_feature(feature[:2])
-        features["functions"][loc[0]]["features"].append((va, feature))
-
-    for feature in doc.get("scopes", {}).get("basic block", []):
-        va, loc = feature[2:]
-        va = int(va, 0x10)
-        loc = [int(lo, 0x10) for lo in loc]
-        feature = deserialize_feature(feature[:2])
-        features["functions"][loc[0]]["basic blocks"][loc[1]]["features"].append((va, feature))
-
-    for feature in doc.get("scopes", {}).get("instruction", []):
-        va, loc = feature[2:]
-        va = int(va, 0x10)
-        loc = [int(lo, 0x10) for lo in loc]
-        feature = deserialize_feature(feature[:2])
-        features["functions"][loc[0]]["basic blocks"][loc[1]]["instructions"][loc[2]]["features"].append((va, feature))
-
-    return capa.features.extractors.base_extractor.NullFeatureExtractor(features)
-
-
-MAGIC = "capa0000".encode("ascii")
-
-
-def dump(extractor):
-    """serialize the given extractor to a byte array."""
-    return MAGIC + zlib.compress(dumps(extractor).encode("utf-8"))
-
-
-def is_freeze(buf: bytes) -> bool:
-    return buf[: len(MAGIC)] == MAGIC
-
-
-def load(buf):
-    """deserialize a set of features (as a NullFeatureExtractor) from a byte array."""
-    if not is_freeze(buf):
-        raise ValueError("missing magic header")
-    return loads(zlib.decompress(buf[len(MAGIC) :]).decode("utf-8"))
-
-
-def main(argv=None):
-    import sys
-    import argparse
-
-    import capa.main
-
-    if argv is None:
-        argv = sys.argv[1:]
-
-    parser = argparse.ArgumentParser(description="save capa features to a file")
-    capa.main.install_common_args(parser, {"sample", "format", "backend", "signatures"})
-    parser.add_argument("output", type=str, help="Path to output file")
-    args = parser.parse_args(args=argv)
-    capa.main.handle_common_args(args)
-
-    sigpaths = capa.main.get_signatures(args.signatures)
-
-    extractor = capa.main.get_extractor(args.sample, args.format, args.backend, sigpaths, False)
-
-    with open(args.output, "wb") as f:
-        f.write(dump(extractor))
-
-    return 0
-
-
-if __name__ == "__main__":
-    import sys
-
-    sys.exit(main())

capa/features/freeze/__init__.py

@@ -0,0 +1,395 @@
+"""
+capa freeze file format: `| capa0000 | + zlib(utf-8(json(...)))`
+
+Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
+Licensed under the Apache License, Version 2.0 (the "License");
+ you may not use this file except in compliance with the License.
+You may obtain a copy of the License at: [package root]/LICENSE.txt
+Unless required by applicable law or agreed to in writing, software distributed under the License
+ is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and limitations under the License.
+"""
+import zlib
+import logging
+from enum import Enum
+from typing import Any, List, Tuple
+
+from pydantic import Field, BaseModel
+
+import capa.helpers
+import capa.version
+import capa.features.file
+import capa.features.insn
+import capa.features.common
+import capa.features.address
+import capa.features.basicblock
+import capa.features.extractors.base_extractor
+from capa.helpers import assert_never
+from capa.features.freeze.features import Feature, feature_from_capa
+
+logger = logging.getLogger(__name__)
+
+
+class HashableModel(BaseModel):
+    class Config:
+        frozen = True
+
+
+class AddressType(str, Enum):
+    ABSOLUTE = "absolute"
+    RELATIVE = "relative"
+    FILE = "file"
+    DN_TOKEN = "dn token"
+    DN_TOKEN_OFFSET = "dn token offset"
+    NO_ADDRESS = "no address"
+
+
+class Address(HashableModel):
+    type: AddressType
+    value: Any
+
+    @classmethod
+    def from_capa(cls, a: capa.features.address.Address) -> "Address":
+        if isinstance(a, capa.features.address.AbsoluteVirtualAddress):
+            return cls(type=AddressType.ABSOLUTE, value=int(a))
+
+        elif isinstance(a, capa.features.address.RelativeVirtualAddress):
+            return cls(type=AddressType.RELATIVE, value=int(a))
+
+        elif isinstance(a, capa.features.address.FileOffsetAddress):
+            return cls(type=AddressType.FILE, value=int(a))
+
+        elif isinstance(a, capa.features.address.DNTokenAddress):
+            return cls(type=AddressType.DN_TOKEN, value=int(a))
+
+        elif isinstance(a, capa.features.address.DNTokenOffsetAddress):
+            return cls(type=AddressType.DN_TOKEN_OFFSET, value=(a.token, a.offset))
+
+        elif a == capa.features.address.NO_ADDRESS or isinstance(a, capa.features.address._NoAddress):
+            return cls(type=AddressType.NO_ADDRESS, value=None)
+
+        elif isinstance(a, capa.features.address.Address) and not issubclass(type(a), capa.features.address.Address):
+            raise ValueError("don't use an Address instance directly")
+
+        elif isinstance(a, capa.features.address.Address):
+            raise ValueError("don't use an Address instance directly")
+
+        else:
+            assert_never(a)
+
+    def to_capa(self) -> capa.features.address.Address:
+        if self.type is AddressType.ABSOLUTE:
+            return capa.features.address.AbsoluteVirtualAddress(self.value)
+
+        elif self.type is AddressType.RELATIVE:
+            return capa.features.address.RelativeVirtualAddress(self.value)
+
+        elif self.type is AddressType.FILE:
+            return capa.features.address.FileOffsetAddress(self.value)
+
+        elif self.type is AddressType.DN_TOKEN:
+            return capa.features.address.DNTokenAddress(self.value)
+
+        elif self.type is AddressType.DN_TOKEN_OFFSET:
+            token, offset = self.value
+            return capa.features.address.DNTokenOffsetAddress(token, offset)
+
+        elif self.type is AddressType.NO_ADDRESS:
+            return capa.features.address.NO_ADDRESS
+
+        else:
+            assert_never(self.type)
+
+    def __lt__(self, other: "Address") -> bool:
+        if self.type != other.type:
+            return self.type < other.type
+
+        if self.type is AddressType.NO_ADDRESS:
+            return True
+
+        else:
+            return self.value < other.value
+
+
+class GlobalFeature(HashableModel):
+    feature: Feature
+
+
+class FileFeature(HashableModel):
+    address: Address
+    feature: Feature
+
+
+class FunctionFeature(HashableModel):
+    """
+    args:
+        function: the address of the function to which this feature belongs.
+        address: the address at which this feature is found.
+
+    function != address because, e.g., the feature may be found *within* the scope (function).
+    versus right at its starting address.
+    """
+
+    function: Address
+    address: Address
+    feature: Feature
+
+
+class BasicBlockFeature(HashableModel):
+    """
+    args:
+        basic_block: the address of the basic block to which this feature belongs.
+        address: the address at which this feature is found.
+
+    basic_block != address because, e.g., the feature may be found *within* the scope (basic block).
+    versus right at its starting address.
+    """
+
+    basic_block: Address = Field(alias="basic block")
+    address: Address
+    feature: Feature
+
+    class Config:
+        allow_population_by_field_name = True
+
+
+class InstructionFeature(HashableModel):
+    """
+    args:
+        instruction: the address of the instruction to which this feature belongs.
+        address: the address at which this feature is found.
+
+    instruction != address because, e.g., the feature may be found *within* the scope (basic block),
+    versus right at its starting address.
+    """
+
+    instruction: Address
+    address: Address
+    feature: Feature
+
+
+class InstructionFeatures(BaseModel):
+    address: Address
+    features: Tuple[InstructionFeature, ...]
+
+
+class BasicBlockFeatures(BaseModel):
+    address: Address
+    features: Tuple[BasicBlockFeature, ...]
+    instructions: Tuple[InstructionFeatures, ...]
+
+
+class FunctionFeatures(BaseModel):
+    address: Address
+    features: Tuple[FunctionFeature, ...]
+    basic_blocks: Tuple[BasicBlockFeatures, ...] = Field(alias="basic blocks")
+
+    class Config:
+        allow_population_by_field_name = True
+
+
+class Features(BaseModel):
+    global_: Tuple[GlobalFeature, ...] = Field(alias="global")
+    file: Tuple[FileFeature, ...]
+    functions: Tuple[FunctionFeatures, ...]
+
+    class Config:
+        allow_population_by_field_name = True
+
+
+class Extractor(BaseModel):
+    name: str
+    version: str = capa.version.__version__
+
+    class Config:
+        allow_population_by_field_name = True
+
+
+class Freeze(BaseModel):
+    version: int = 2
+    base_address: Address = Field(alias="base address")
+    extractor: Extractor
+    features: Features
+
+    class Config:
+        allow_population_by_field_name = True
+
+
+def dumps(extractor: capa.features.extractors.base_extractor.FeatureExtractor) -> str:
+    """
+    serialize the given extractor to a string
+    """
+
+    global_features: List[GlobalFeature] = []
+    for feature, _ in extractor.extract_global_features():
+        global_features.append(
+            GlobalFeature(
+                feature=feature_from_capa(feature),
+            )
+        )
+
+    file_features: List[FileFeature] = []
+    for feature, address in extractor.extract_file_features():
+        file_features.append(
+            FileFeature(
+                feature=feature_from_capa(feature),
+                address=Address.from_capa(address),
+            )
+        )
+
+    function_features: List[FunctionFeatures] = []
+    for f in extractor.get_functions():
+        faddr = Address.from_capa(f.address)
+        ffeatures = [
+            FunctionFeature(
+                function=faddr,
+                address=Address.from_capa(addr),
+                feature=feature_from_capa(feature),
+            )
+            for feature, addr in extractor.extract_function_features(f)
+        ]
+
+        basic_blocks = []
+        for bb in extractor.get_basic_blocks(f):
+            bbaddr = Address.from_capa(bb.address)
+            bbfeatures = [
+                BasicBlockFeature(
+                    basic_block=bbaddr,
+                    address=Address.from_capa(addr),
+                    feature=feature_from_capa(feature),
+                )
+                for feature, addr in extractor.extract_basic_block_features(f, bb)
+            ]
+
+            instructions = []
+            for insn in extractor.get_instructions(f, bb):
+                iaddr = Address.from_capa(insn.address)
+                ifeatures = [
+                    InstructionFeature(
+                        instruction=iaddr,
+                        address=Address.from_capa(addr),
+                        feature=feature_from_capa(feature),
+                    )
+                    for feature, addr in extractor.extract_insn_features(f, bb, insn)
+                ]
+
+                instructions.append(
+                    InstructionFeatures(
+                        address=iaddr,
+                        features=ifeatures,
+                    )
+                )
+
+            basic_blocks.append(
+                BasicBlockFeatures(
+                    address=bbaddr,
+                    features=bbfeatures,
+                    instructions=instructions,
+                )
+            )
+
+        function_features.append(
+            FunctionFeatures(
+                address=faddr,
+                features=ffeatures,
+                basic_blocks=basic_blocks,
+            )
+        )
+
+    features = Features(
+        global_=global_features,
+        file=file_features,
+        functions=function_features,
+    )
+
+    freeze = Freeze(
+        version=2,
+        base_address=Address.from_capa(extractor.get_base_address()),
+        extractor=Extractor(name=extractor.__class__.__name__),
+        features=features,
+    )
+
+    return freeze.json()
+
+
+def loads(s: str) -> capa.features.extractors.base_extractor.FeatureExtractor:
+    """deserialize a set of features (as a NullFeatureExtractor) from a string."""
+    import capa.features.extractors.null as null
+
+    freeze = Freeze.parse_raw(s)
+    if freeze.version != 2:
+        raise ValueError("unsupported freeze format version: %d", freeze.version)
+
+    return null.NullFeatureExtractor(
+        base_address=freeze.base_address.to_capa(),
+        global_features=[f.feature.to_capa() for f in freeze.features.global_],
+        file_features=[(f.address.to_capa(), f.feature.to_capa()) for f in freeze.features.file],
+        functions={
+            f.address.to_capa(): null.FunctionFeatures(
+                features=[(fe.address.to_capa(), fe.feature.to_capa()) for fe in f.features],
+                basic_blocks={
+                    bb.address.to_capa(): null.BasicBlockFeatures(
+                        features=[(fe.address.to_capa(), fe.feature.to_capa()) for fe in bb.features],
+                        instructions={
+                            i.address.to_capa(): null.InstructionFeatures(
+                                features=[(fe.address.to_capa(), fe.feature.to_capa()) for fe in i.features]
+                            )
+                            for i in bb.instructions
+                        },
+                    )
+                    for bb in f.basic_blocks
+                },
+            )
+            for f in freeze.features.functions
+        },
+    )
+
+
+MAGIC = "capa0000".encode("ascii")
+
+
+def dump(extractor: capa.features.extractors.base_extractor.FeatureExtractor) -> bytes:
+    """serialize the given extractor to a byte array."""
+    return MAGIC + zlib.compress(dumps(extractor).encode("utf-8"))
+
+
+def is_freeze(buf: bytes) -> bool:
+    return buf[: len(MAGIC)] == MAGIC
+
+
+def load(buf: bytes) -> capa.features.extractors.base_extractor.FeatureExtractor:
+    """deserialize a set of features (as a NullFeatureExtractor) from a byte array."""
+    if not is_freeze(buf):
+        raise ValueError("missing magic header")
+    return loads(zlib.decompress(buf[len(MAGIC) :]).decode("utf-8"))
+
+
+def main(argv=None):
+    import sys
+    import argparse
+
+    import capa.main
+
+    if argv is None:
+        argv = sys.argv[1:]
+
+    parser = argparse.ArgumentParser(description="save capa features to a file")
+    capa.main.install_common_args(parser, {"sample", "format", "backend", "signatures"})
+    parser.add_argument("output", type=str, help="Path to output file")
+    args = parser.parse_args(args=argv)
+    capa.main.handle_common_args(args)
+
+    sigpaths = capa.main.get_signatures(args.signatures)
+
+    extractor = capa.main.get_extractor(args.sample, args.format, args.backend, sigpaths, False)
+
+    with open(args.output, "wb") as f:
+        f.write(dump(extractor))
+
+    return 0
+
+
+if __name__ == "__main__":
+    import sys
+
+    sys.exit(main())

capa/features/freeze/features.py

@@ -0,0 +1,345 @@
+import binascii
+from typing import Union, Optional
+
+from pydantic import Field, BaseModel
+
+import capa.features.file
+import capa.features.insn
+import capa.features.common
+import capa.features.basicblock
+
+
+class FeatureModel(BaseModel):
+    class Config:
+        frozen = True
+        allow_population_by_field_name = True
+
+    def to_capa(self) -> capa.features.common.Feature:
+        if isinstance(self, OSFeature):
+            return capa.features.common.OS(self.os, description=self.description)
+
+        elif isinstance(self, ArchFeature):
+            return capa.features.common.Arch(self.arch, description=self.description)
+
+        elif isinstance(self, FormatFeature):
+            return capa.features.common.Format(self.format, description=self.description)
+
+        elif isinstance(self, MatchFeature):
+            return capa.features.common.MatchedRule(self.match, description=self.description)
+
+        elif isinstance(
+            self,
+            CharacteristicFeature,
+        ):
+            return capa.features.common.Characteristic(self.characteristic, description=self.description)
+
+        elif isinstance(self, ExportFeature):
+            return capa.features.file.Export(self.export, description=self.description)
+
+        elif isinstance(self, ImportFeature):
+            return capa.features.file.Import(self.import_, description=self.description)
+
+        elif isinstance(self, SectionFeature):
+            return capa.features.file.Section(self.section, description=self.description)
+
+        elif isinstance(self, FunctionNameFeature):
+            return capa.features.file.FunctionName(self.function_name, description=self.description)
+
+        elif isinstance(self, SubstringFeature):
+            return capa.features.common.Substring(self.substring, description=self.description)
+
+        elif isinstance(self, RegexFeature):
+            return capa.features.common.Regex(self.regex, description=self.description)
+
+        elif isinstance(self, StringFeature):
+            return capa.features.common.String(self.string, description=self.description)
+
+        elif isinstance(self, ClassFeature):
+            return capa.features.common.Class(self.class_, description=self.description)
+
+        elif isinstance(self, NamespaceFeature):
+            return capa.features.common.Namespace(self.namespace, description=self.description)
+
+        elif isinstance(self, BasicBlockFeature):
+            return capa.features.basicblock.BasicBlock(description=self.description)
+
+        elif isinstance(self, APIFeature):
+            return capa.features.insn.API(self.api, description=self.description)
+
+        elif isinstance(self, PropertyFeature):
+            return capa.features.insn.Property(self.property, access=self.access, description=self.description)
+
+        elif isinstance(self, NumberFeature):
+            return capa.features.insn.Number(self.number, description=self.description)
+
+        elif isinstance(self, BytesFeature):
+            return capa.features.common.Bytes(binascii.unhexlify(self.bytes), description=self.description)
+
+        elif isinstance(self, OffsetFeature):
+            return capa.features.insn.Offset(self.offset, description=self.description)
+
+        elif isinstance(self, MnemonicFeature):
+            return capa.features.insn.Mnemonic(self.mnemonic, description=self.description)
+
+        elif isinstance(self, OperandNumberFeature):
+            return capa.features.insn.OperandNumber(
+                self.index,
+                self.operand_number,
+                description=self.description,
+            )
+
+        elif isinstance(self, OperandOffsetFeature):
+            return capa.features.insn.OperandOffset(
+                self.index,
+                self.operand_offset,
+                description=self.description,
+            )
+
+        else:
+            raise NotImplementedError(f"Feature.to_capa({type(self)}) not implemented")
+
+
+def feature_from_capa(f: capa.features.common.Feature) -> "Feature":
+    if isinstance(f, capa.features.common.OS):
+        return OSFeature(os=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.common.Arch):
+        return ArchFeature(arch=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.common.Format):
+        return FormatFeature(format=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.common.MatchedRule):
+        return MatchFeature(match=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.common.Characteristic):
+        return CharacteristicFeature(characteristic=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.file.Export):
+        return ExportFeature(export=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.file.Import):
+        return ImportFeature(import_=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.file.Section):
+        return SectionFeature(section=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.file.FunctionName):
+        return FunctionNameFeature(function_name=f.value, description=f.description)
+
+    # must come before check for String due to inheritance
+    elif isinstance(f, capa.features.common.Substring):
+        return SubstringFeature(substring=f.value, description=f.description)
+
+    # must come before check for String due to inheritance
+    elif isinstance(f, capa.features.common.Regex):
+        return RegexFeature(regex=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.common.String):
+        return StringFeature(string=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.common.Class):
+        return ClassFeature(class_=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.common.Namespace):
+        return NamespaceFeature(namespace=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.basicblock.BasicBlock):
+        return BasicBlockFeature(description=f.description)
+
+    elif isinstance(f, capa.features.insn.API):
+        return APIFeature(api=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.insn.Property):
+        return PropertyFeature(property=f.value, access=f.access, description=f.description)
+
+    elif isinstance(f, capa.features.insn.Number):
+        return NumberFeature(number=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.common.Bytes):
+        buf = f.value
+        assert isinstance(buf, bytes)
+        return BytesFeature(bytes=binascii.hexlify(buf).decode("ascii"), description=f.description)
+
+    elif isinstance(f, capa.features.insn.Offset):
+        return OffsetFeature(offset=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.insn.Mnemonic):
+        return MnemonicFeature(mnemonic=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.insn.OperandNumber):
+        return OperandNumberFeature(index=f.index, operand_number=f.value, description=f.description)
+
+    elif isinstance(f, capa.features.insn.OperandOffset):
+        return OperandOffsetFeature(index=f.index, operand_offset=f.value, description=f.description)
+
+    else:
+        raise NotImplementedError(f"feature_from_capa({type(f)}) not implemented")
+
+
+class OSFeature(FeatureModel):
+    type: str = "os"
+    os: str
+    description: Optional[str]
+
+
+class ArchFeature(FeatureModel):
+    type: str = "arch"
+    arch: str
+    description: Optional[str]
+
+
+class FormatFeature(FeatureModel):
+    type: str = "format"
+    format: str
+    description: Optional[str]
+
+
+class MatchFeature(FeatureModel):
+    type: str = "match"
+    match: str
+    description: Optional[str]
+
+
+class CharacteristicFeature(FeatureModel):
+    type: str = "characteristic"
+    characteristic: str
+    description: Optional[str]
+
+
+class ExportFeature(FeatureModel):
+    type: str = "export"
+    export: str
+    description: Optional[str]
+
+
+class ImportFeature(FeatureModel):
+    type: str = "import"
+    import_: str = Field(alias="import")
+    description: Optional[str]
+
+
+class SectionFeature(FeatureModel):
+    type: str = "section"
+    section: str
+    description: Optional[str]
+
+
+class FunctionNameFeature(FeatureModel):
+    type: str = "function name"
+    function_name: str = Field(alias="function name")
+    description: Optional[str]
+
+
+class SubstringFeature(FeatureModel):
+    type: str = "substring"
+    substring: str
+    description: Optional[str]
+
+
+class RegexFeature(FeatureModel):
+    type: str = "regex"
+    regex: str
+    description: Optional[str]
+
+
+class StringFeature(FeatureModel):
+    type: str = "string"
+    string: str
+    description: Optional[str]
+
+
+class ClassFeature(FeatureModel):
+    type: str = "class"
+    class_: str = Field(alias="class")
+    description: Optional[str]
+
+
+class NamespaceFeature(FeatureModel):
+    type: str = "namespace"
+    namespace: str
+    description: Optional[str]
+
+
+class BasicBlockFeature(FeatureModel):
+    type: str = "basic block"
+    description: Optional[str]
+
+
+class APIFeature(FeatureModel):
+    type: str = "api"
+    api: str
+    description: Optional[str]
+
+
+class PropertyFeature(FeatureModel):
+    type: str = "property"
+    access: Optional[str]
+    property: str
+    description: Optional[str]
+
+
+class NumberFeature(FeatureModel):
+    type: str = "number"
+    number: Union[int, float]
+    description: Optional[str]
+
+
+class BytesFeature(FeatureModel):
+    type: str = "bytes"
+    bytes: str
+    description: Optional[str]
+
+
+class OffsetFeature(FeatureModel):
+    type: str = "offset"
+    offset: int
+    description: Optional[str]
+
+
+class MnemonicFeature(FeatureModel):
+    type: str = "mnemonic"
+    mnemonic: str
+    description: Optional[str]
+
+
+class OperandNumberFeature(FeatureModel):
+    type: str = "operand number"
+    index: int
+    operand_number: int = Field(alias="operand number")
+    description: Optional[str]
+
+
+class OperandOffsetFeature(FeatureModel):
+    type: str = "operand offset"
+    index: int
+    operand_offset: int = Field(alias="operand offset")
+    description: Optional[str]
+
+
+Feature = Union[
+    OSFeature,
+    ArchFeature,
+    FormatFeature,
+    MatchFeature,
+    CharacteristicFeature,
+    ExportFeature,
+    ImportFeature,
+    SectionFeature,
+    FunctionNameFeature,
+    SubstringFeature,
+    RegexFeature,
+    StringFeature,
+    ClassFeature,
+    NamespaceFeature,
+    APIFeature,
+    PropertyFeature,
+    NumberFeature,
+    BytesFeature,
+    OffsetFeature,
+    MnemonicFeature,
+    OperandNumberFeature,
+    OperandOffsetFeature,
+    # Note! this must be last, see #1161
+    BasicBlockFeature,
+]

capa/features/insn.py

@@ -1,41 +1,131 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
+import abc
+from typing import Union, Optional
 
-import capa.render.utils
-from capa.features.common import Feature
+import capa.helpers
+from capa.features.common import VALID_FEATURE_ACCESS, Feature
+
+
+def hex(n: int) -> str:
+    """render the given number using upper case hex, like: 0x123ABC"""
+    if n < 0:
+        return "-0x%X" % (-n)
+    else:
+        return "0x%X" % n
 
 
 class API(Feature):
     def __init__(self, name: str, description=None):
-        # Downcase library name if given
-        if "." in name:
-            modname, _, impname = name.rpartition(".")
-            name = modname.lower() + "." + impname
+        super().__init__(name, description=description)
 
-        super(API, self).__init__(name, description=description)
+
+class _AccessFeature(Feature, abc.ABC):
+    # superclass: don't use directly
+    def __init__(self, value: str, access: Optional[str] = None, description: Optional[str] = None):
+        super().__init__(value, description=description)
+        if access is not None:
+            if access not in VALID_FEATURE_ACCESS:
+                raise ValueError("%s access type %s not valid" % (self.name, access))
+        self.access = access
+
+    def __hash__(self):
+        return hash((self.name, self.value, self.access))
+
+    def __eq__(self, other):
+        return super().__eq__(other) and self.access == other.access
+
+    def get_name_str(self) -> str:
+        if self.access is not None:
+            return f"{self.name}/{self.access}"
+        return self.name
+
+
+class Property(_AccessFeature):
+    def __init__(self, value: str, access: Optional[str] = None, description=None):
+        super().__init__(value, access=access, description=description)
 
 
 class Number(Feature):
-    def __init__(self, value: int, bitness=None, description=None):
-        super(Number, self).__init__(value, bitness=bitness, description=description)
+    def __init__(self, value: Union[int, float], description=None):
+        super().__init__(value, description=description)
 
     def get_value_str(self):
-        return capa.render.utils.hex(self.value)
+        if isinstance(self.value, int):
+            return capa.helpers.hex(self.value)
+        elif isinstance(self.value, float):
+            return str(self.value)
+        else:
+            raise ValueError("invalid value type")
+
+
+# max recognized structure size (and therefore, offset size)
+MAX_STRUCTURE_SIZE = 0x10000
 
 
 class Offset(Feature):
-    def __init__(self, value: int, bitness=None, description=None):
-        super(Offset, self).__init__(value, bitness=bitness, description=description)
+    def __init__(self, value: int, description=None):
+        super().__init__(value, description=description)
 
     def get_value_str(self):
-        return capa.render.utils.hex(self.value)
+        assert isinstance(self.value, int)
+        return hex(self.value)
 
 
 class Mnemonic(Feature):
     def __init__(self, value: str, description=None):
-        super(Mnemonic, self).__init__(value, description=description)
+        super().__init__(value, description=description)
+
+
+# max number of operands to consider for a given instruction.
+# since we only support Intel and .NET, we can assume this is 3
+# which covers cases up to e.g. "vinserti128 ymm0,ymm0,ymm5,1"
+MAX_OPERAND_COUNT = 4
+MAX_OPERAND_INDEX = MAX_OPERAND_COUNT - 1
+
+
+class _Operand(Feature, abc.ABC):
+    # superclass: don't use directly
+    # subclasses should set self.name and provide the value string formatter
+    def __init__(self, index: int, value: int, description=None):
+        super().__init__(value, description=description)
+        self.index = index
+
+    def __hash__(self):
+        return hash((self.name, self.value))
+
+    def __eq__(self, other):
+        return super().__eq__(other) and self.index == other.index
+
+
+class OperandNumber(_Operand):
+    # cached names so we don't do extra string formatting every ctor
+    NAMES = ["operand[%d].number" % i for i in range(MAX_OPERAND_COUNT)]
+
+    # operand[i].number: 0x12
+    def __init__(self, index: int, value: int, description=None):
+        super().__init__(index, value, description=description)
+        self.name = self.NAMES[index]
+
+    def get_value_str(self) -> str:
+        assert isinstance(self.value, int)
+        return hex(self.value)
+
+
+class OperandOffset(_Operand):
+    # cached names so we don't do extra string formatting every ctor
+    NAMES = ["operand[%d].offset" % i for i in range(MAX_OPERAND_COUNT)]
+
+    # operand[i].offset: 0x12
+    def __init__(self, index: int, value: int, description=None):
+        super().__init__(index, value, description=description)
+        self.name = self.NAMES[index]
+
+    def get_value_str(self) -> str:
+        assert isinstance(self.value, int)
+        return hex(self.value)

capa/helpers.py

@@ -1,18 +1,30 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
-
 import os
+import logging
+from typing import NoReturn
 
-_hex = hex
+from capa.exceptions import UnsupportedFormatError
+from capa.features.common import FORMAT_PE, FORMAT_SC32, FORMAT_SC64, FORMAT_DOTNET, FORMAT_UNKNOWN, Format
+
+EXTENSIONS_SHELLCODE_32 = ("sc32", "raw32")
+EXTENSIONS_SHELLCODE_64 = ("sc64", "raw64")
+EXTENSIONS_ELF = "elf_"
+
+logger = logging.getLogger("capa")
 
 
-def hex(i):
-    return _hex(int(i))
+def hex(n: int) -> str:
+    """render the given number using upper case hex, like: 0x123ABC"""
+    if n < 0:
+        return "-0x%X" % (-n)
+    else:
+        return "0x%X" % n
 
 
 def get_file_taste(sample_path: str) -> bytes:
@@ -30,3 +42,85 @@ def is_runtime_ida():
         return False
     else:
         return True
+
+
+def assert_never(value: NoReturn) -> NoReturn:
+    assert False, f"Unhandled value: {value} ({type(value).__name__})"
+
+
+def get_format_from_extension(sample: str) -> str:
+    if sample.endswith(EXTENSIONS_SHELLCODE_32):
+        return FORMAT_SC32
+    elif sample.endswith(EXTENSIONS_SHELLCODE_64):
+        return FORMAT_SC64
+    return FORMAT_UNKNOWN
+
+
+def get_auto_format(path: str) -> str:
+    format_ = get_format(path)
+    if format_ == FORMAT_UNKNOWN:
+        format_ = get_format_from_extension(path)
+    if format_ == FORMAT_UNKNOWN:
+        raise UnsupportedFormatError()
+    return format_
+
+
+def get_format(sample: str) -> str:
+    # imported locally to avoid import cycle
+    from capa.features.extractors.common import extract_format
+    from capa.features.extractors.dnfile_ import DnfileFeatureExtractor
+
+    with open(sample, "rb") as f:
+        buf = f.read()
+
+    for feature, _ in extract_format(buf):
+        if feature == Format(FORMAT_PE):
+            dnfile_extractor = DnfileFeatureExtractor(sample)
+            if dnfile_extractor.is_dotnet_file():
+                feature = Format(FORMAT_DOTNET)
+
+        assert isinstance(feature.value, str)
+        return feature.value
+
+    return FORMAT_UNKNOWN
+
+
+def log_unsupported_format_error():
+    logger.error("-" * 80)
+    logger.error(" Input file does not appear to be a PE or ELF file.")
+    logger.error(" ")
+    logger.error(
+        " capa currently only supports analyzing PE and ELF files (or shellcode, when using --format sc32|sc64)."
+    )
+    logger.error(" If you don't know the input file type, you can try using the `file` utility to guess it.")
+    logger.error("-" * 80)
+
+
+def log_unsupported_os_error():
+    logger.error("-" * 80)
+    logger.error(" Input file does not appear to target a supported OS.")
+    logger.error(" ")
+    logger.error(
+        " capa currently only supports analyzing executables for some operating systems (including Windows and Linux)."
+    )
+    logger.error("-" * 80)
+
+
+def log_unsupported_arch_error():
+    logger.error("-" * 80)
+    logger.error(" Input file does not appear to target a supported architecture.")
+    logger.error(" ")
+    logger.error(" capa currently only supports analyzing x86 (32- and 64-bit).")
+    logger.error("-" * 80)
+
+
+def log_unsupported_runtime_error():
+    logger.error("-" * 80)
+    logger.error(" Unsupported runtime or Python interpreter.")
+    logger.error(" ")
+    logger.error(" capa supports running under Python 3.7 and higher.")
+    logger.error(" ")
+    logger.error(
+        " If you're seeing this message on the command line, please ensure you're running a supported Python version."
+    )
+    logger.error("-" * 80)

capa/ida/helpers.py

@@ -1,57 +1,59 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
-
+import json
 import logging
 import datetime
+import contextlib
+from typing import Optional
 
 import idc
 import idaapi
 import idautils
 import ida_bytes
 import ida_loader
+from netnode import netnode
 
 import capa
 import capa.version
+import capa.render.utils as rutils
 import capa.features.common
+import capa.render.result_document
+from capa.features.address import AbsoluteVirtualAddress
 
 logger = logging.getLogger("capa")
 
-# IDA version as returned by idaapi.get_kernel_version()
-SUPPORTED_IDA_VERSIONS = (
-    "7.4",
-    "7.5",
-    "7.6",
-)
-
 # file type as returned by idainfo.file_type
 SUPPORTED_FILE_TYPES = (
     idaapi.f_PE,
     idaapi.f_ELF,
     idaapi.f_BIN,
+    idaapi.f_COFF,
     # idaapi.f_MACHO,
 )
 
 # arch type as returned by idainfo.procname
 SUPPORTED_ARCH_TYPES = ("metapc",)
 
+CAPA_NETNODE = f"$ com.mandiant.capa.v{capa.version.__version__}"
+NETNODE_RESULTS = "results"
+NETNODE_RULES_CACHE_ID = "rules-cache-id"
+
 
 def inform_user_ida_ui(message):
     idaapi.info("%s. Please refer to IDA Output window for more information." % message)
 
 
 def is_supported_ida_version():
-    version = idaapi.get_kernel_version()
-    if version not in SUPPORTED_IDA_VERSIONS:
+    version = float(idaapi.get_kernel_version())
+    if version < 7.4 or version >= 9:
         warning_msg = "This plugin does not support your IDA Pro version"
         logger.warning(warning_msg)
-        logger.warning(
-            "Your IDA Pro version is: %s. Supported versions are: %s." % (version, ", ".join(SUPPORTED_IDA_VERSIONS))
-        )
+        logger.warning("Your IDA Pro version is: %s. Supported versions are: IDA >= 7.4 and IDA < 9.0." % version)
         return False
     return True
 
@@ -116,14 +118,31 @@ def get_file_sha256():
     return sha256
 
 
-def collect_metadata():
+def collect_metadata(rules):
     """ """
     md5 = get_file_md5()
     sha256 = get_file_sha256()
 
+    info: idaapi.idainfo = idaapi.get_inf_structure()
+    if info.procname == "metapc" and info.is_64bit():
+        arch = "x86_64"
+    elif info.procname == "metapc" and info.is_32bit():
+        arch = "x86"
+    else:
+        arch = "unknown arch"
+
+    format_name: str = ida_loader.get_file_type_name()
+    if "PE" in format_name:
+        os = "windows"
+    elif "ELF" in format_name:
+        with contextlib.closing(capa.ida.helpers.IDAIO()) as f:
+            os = capa.features.extractors.elf.detect_elf_os(f)
+    else:
+        os = "unknown os"
+
     return {
         "timestamp": datetime.datetime.now().isoformat(),
-        # "argv" is not relevant here
+        "argv": [],
         "sample": {
             "md5": md5,
             "sha1": "",  # not easily accessible
@@ -132,8 +151,23 @@ def collect_metadata():
         },
         "analysis": {
             "format": idaapi.get_file_type_name(),
+            "arch": arch,
+            "os": os,
             "extractor": "ida",
+            "rules": rules,
             "base_address": idaapi.get_imagebase(),
+            "layout": {
+                # this is updated after capabilities have been collected.
+                # will look like:
+                #
+                # "functions": { 0x401000: { "matched_basic_blocks": [ 0x401000, 0x401005, ... ] }, ... }
+            },
+            # ignore these for now - not used by IDA plugin.
+            "feature_counts": {
+                "file": {},
+                "functions": {},
+            },
+            "library_functions": {},
         },
         "version": capa.version.__version__,
     }
@@ -146,7 +180,7 @@ class IDAIO:
     """
 
     def __init__(self):
-        super(IDAIO, self).__init__()
+        super().__init__()
         self.offset = 0
 
     def seek(self, offset, whence=0):
@@ -156,11 +190,63 @@ class IDAIO:
     def read(self, size):
         ea = ida_loader.get_fileregion_ea(self.offset)
         if ea == idc.BADADDR:
-            # best guess, such as if file is mapped at address 0x0.
-            ea = self.offset
+            logger.debug("cannot read 0x%x bytes at 0x%x (ea: BADADDR)", size, self.offset)
+            return b""
 
         logger.debug("reading 0x%x bytes at 0x%x (ea: 0x%x)", size, self.offset, ea)
-        return ida_bytes.get_bytes(ea, size)
+
+        # get_bytes returns None on error, for consistency with read always return bytes
+        return ida_bytes.get_bytes(ea, size) or b""
 
     def close(self):
         return
+
+
+def save_cached_results(resdoc):
+    logger.debug("saving cached capa results to netnode '%s'", CAPA_NETNODE)
+    n = netnode.Netnode(CAPA_NETNODE)
+    n[NETNODE_RESULTS] = resdoc.json()
+
+
+def idb_contains_cached_results() -> bool:
+    try:
+        n = netnode.Netnode(CAPA_NETNODE)
+        return bool(n.get(NETNODE_RESULTS))
+    except netnode.NetnodeCorruptError as e:
+        logger.error("%s", e, exc_info=True)
+        return False
+
+
+def load_and_verify_cached_results() -> Optional[capa.render.result_document.ResultDocument]:
+    """verifies that cached results have valid (mapped) addresses for the current database"""
+    logger.debug("loading cached capa results from netnode '%s'", CAPA_NETNODE)
+
+    n = netnode.Netnode(CAPA_NETNODE)
+    doc = capa.render.result_document.ResultDocument.parse_obj(json.loads(n[NETNODE_RESULTS]))
+
+    for rule in rutils.capability_rules(doc):
+        for location_, _ in rule.matches:
+            location = location_.to_capa()
+            if isinstance(location, AbsoluteVirtualAddress):
+                ea = int(location)
+                if not idaapi.is_mapped(ea):
+                    logger.error("cached address %s is not a valid location in this database", hex(ea))
+                    return None
+    return doc
+
+
+def save_rules_cache_id(ruleset_id):
+    logger.debug("saving ruleset ID to netnode '%s'", CAPA_NETNODE)
+    n = netnode.Netnode(CAPA_NETNODE)
+    n[NETNODE_RULES_CACHE_ID] = ruleset_id
+
+
+def load_rules_cache_id():
+    n = netnode.Netnode(CAPA_NETNODE)
+    return n[NETNODE_RULES_CACHE_ID]
+
+
+def delete_cached_results():
+    logger.debug("deleting cached capa data")
+    n = netnode.Netnode(CAPA_NETNODE)
+    del n[NETNODE_RESULTS]

capa/ida/plugin/README.md

@@ -1,8 +1,8 @@
 ![capa explorer](../../../.github/capa-explorer-logo.png)
 
 capa explorer is an IDAPython plugin that integrates the FLARE team's open-source framework, capa, with IDA Pro. capa is a framework that uses a well-defined collection of rules to 
-identify capabilities in a program. You can run capa against a PE file or shellcode and it tells you what it thinks the program can do. For example, it might suggest that 
-the program is a backdoor, can install services, or relies on HTTP to communicate. capa explorer runs capa directly against your IDA Pro database (IDB) without requiring access
+identify capabilities in a program. You can run capa against a PE file, ELF file, or shellcode and it tells you what it thinks the program can do. For example, it might suggest that 
+the program is a backdoor, can install services, or relies on HTTP to communicate. capa explorer runs capa analysis on your IDA Pro database (IDB) without needing access
 to the original binary file. Once a database has been analyzed, capa explorer helps you identify interesting areas of a program and build new capa rules using features extracted from your IDB.
 
 We love using capa explorer during malware analysis because it teaches us what parts of a program suggest a behavior. As we click on rows, capa explorer jumps directly 
@@ -21,73 +21,43 @@ We can use capa explorer to navigate our Disassembly view directly to the suspec
 Using the `Rule Information` and `Details` columns capa explorer shows us that the suspect function matched `self delete via COMSPEC environment variable` because it contains capa rule matches for `create process`, `get COMSPEC environment variable`,
 and `query environment variable`, references to the strings `COMSPEC`, ` > nul`, and `/c del `, and calls to the Windows API functions `GetEnvironmentVariableA` and `ShellExecuteEx`.
 
-capa explorer also helps you build new capa rules. To start select the `Rule Generator` tab, navigate to a function in your Disassembly view,
+capa explorer also helps you build and test new capa rules. To start, select the `Rule Generator` tab, navigate to a function in your Disassembly view,
 and click `Analyze`. capa explorer will extract features from the function and display them in the `Features` pane. You can add features listed in this pane to the `Editor` pane
 by either double-clicking a feature or using multi-select + right-click to add multiple features at once. The `Preview` and `Editor` panes help edit your rule. Use the `Preview` pane
-to modify the rule text directly and the `Editor` pane to construct and rearrange your hierarchy of statements and features. When you finish a rule you can save it directly to a file by clicking `Save`.
+to modify rule text directly and the `Editor` pane to construct and rearrange your hierarchy of statements and features. When you finish a rule you can save it directly to a file by clicking `Save`.
 
 ![](../../../doc/img/rulegen_expanded.png)
 
-For more information on the FLARE team's open-source framework, capa, check out the overview in our first [blog](https://www.fireeye.com/blog/threat-research/2020/07/capa-automatically-identify-malware-capabilities.html).
+For more information on the FLARE team's open-source framework, capa, check out the overview in our first [blog](https://www.mandiant.com/resources/capa-automatically-identify-malware-capabilities).
 
 ## Getting Started
 
-### Requirements
-
-capa explorer supports Python versions >= 3.6.x and the following IDA Pro versions:
-
-* IDA 7.4
-* IDA 7.5
-* IDA 7.6 (caveat below)
-
-capa explorer is however limited to the Python versions supported by your IDA installation (which may not include all Python versions >= 3.6.x). Based on our testing the following matrix shows the Python versions supported
-by each supported IDA version:
-
-| | IDA 7.4 | IDA 7.5 | IDA 7.6 |
-| --- | --- | --- | --- |
-| Python 3.6.x | Yes | Yes | Yes |
-| Python 3.7.x | Yes | Yes | Yes |
-| Python 3.8.x | Partial (see below) | Yes | Yes |
-| Python 3.9.x | No | Partial (see below) | Yes |
-
-To use capa explorer with IDA 7.4 and Python 3.8.x you must follow the instructions provided by hex-rays [here](https://hex-rays.com/blog/ida-7-4-and-python-3-8/).
-
-To use capa explorer with IDA 7.5 and Python 3.9.x you must follow the instructions provided by hex-rays [here](https://hex-rays.com/blog/python-3-9-support-for-ida-7-5/).
-
-If you encounter issues with your specific setup, please open a new [Issue](https://github.com/fireeye/capa/issues).
-
-#### IDA 7.6 caveat: IDA 7.6sp1 or patch required
-
-As described [here](https://www.hex-rays.com/blog/ida-7-6-empty-qtreeview-qtreewidget/):
-
-> A rather nasty issue evaded our testing and found its way into IDA 7.6: using the PyQt5 modules that are shipped with IDA, QTreeView (or QTreeWidget) instances will always fail to display contents.
-
-Therefore, in order to use capa under IDA 7.6 you need the [Service Pack 1 for IDA 7.6](https://www.hex-rays.com/products/ida/news/7_6sp1). Alternatively, you can download and install the fix corresponding to your IDA installation, replacing the original QtWidgets DLL with the one contained in the .zip file (links to Hex-Rays):
-
-
-  - Windows: [pyqt5_qtwidgets_win](https://www.hex-rays.com/wp-content/uploads/2021/04/pyqt5_qtwidgets_win.zip)
-  - Linux: [pyqt5_qtwidgets_linux](https://www.hex-rays.com/wp-content/uploads/2021/04/pyqt5_qtwidgets_linux.zip)
-  - MacOS (Intel): [pyqt5_qtwidgets_mac_x64](https://www.hex-rays.com/wp-content/uploads/2021/04/pyqt5_qtwidgets_mac_x64.zip)
-  - MacOS (AppleSilicon): [pyqt5_qtwidgets_mac_arm](https://www.hex-rays.com/wp-content/uploads/2021/04/pyqt5_qtwidgets_mac_arm.zip)
-
-
-### Supported File Types
-
-capa explorer is limited to the file types supported by capa, which include:
-
-* Windows x86 (32- and 64-bit) PE and ELF files
-* Windows x86 (32- and 64-bit) shellcode
-
 ### Installation
 
 You can install capa explorer using the following steps:
 
-1. Install capa and its dependencies from PyPI for the Python interpreter used by your IDA installation:
+1. Install capa and its dependencies from PyPI using the Python interpreter configured for your IDA installation:
     ```
     $ pip install flare-capa
     ```
-3. Download the [standard collection of capa rules](https://github.com/fireeye/capa-rules) (capa explorer needs capa rules to analyze a database)
-4. Copy [capa_explorer.py](https://raw.githubusercontent.com/fireeye/capa/master/capa/ida/plugin/capa_explorer.py) to your IDA plugins directory
+2. Download and extract the [official capa rules](https://github.com/mandiant/capa-rules/releases) that match the version of capa you have installed
+   1. Use the following command to view the version of capa you have installed:
+   ```commandline
+    $ pip show flare-capa
+    OR
+    $ capa --version
+    ```
+3. Copy [capa_explorer.py](https://raw.githubusercontent.com/mandiant/capa/master/capa/ida/plugin/capa_explorer.py) to your IDA plugins directory
+   - find your plugin directories via `idaapi.get_ida_subdirs("plugins")` or see this [Hex-Rays blog](https://hex-rays.com/blog/igors-tip-of-the-week-103-sharing-plugins-between-ida-installs/)
+   - common paths are `%APPDATA%\Hex-Rays\IDA Pro\plugins` (Windows) or `$HOME/.idapro/plugins` on Linux/Mac
+
+### Supported File Types
+
+capa explorer is limited to the file types supported by capa, which include:
+
+* Windows x86 (32- and 64-bit) PE files
+* Windows x86 (32- and 64-bit) shellcode
+* ELF files on various operating systems
 
 ### Usage
 
@@ -97,19 +67,20 @@ You can install capa explorer using the following steps:
 3. Select the `Program Analysis` tab
 4. Click the `Analyze` button
 
-When running capa explorer for the first time you are prompted to select a file directory containing capa rules. The plugin conveniently
-remembers your selection for future runs; you can change this selection and other default settings by clicking `Settings`. We recommend 
-downloading and using the [standard collection of capa rules](https://github.com/fireeye/capa-rules) when getting started with the plugin.
+The first time you run capa explorer you will be asked to specify a local directory containing capa rules to use for analysis. We recommend downloading and extracting the [official capa rules](https://github.com/mandiant/capa-rules/releases) that match 
+the version of capa you have installed (see installation instructions above for more details). capa explorer remembers your selection for future analysis which you
+can update using the `Settings` button.
 
 #### Tips for Program Analysis
 
 * Start analysis by clicking the `Analyze` button
+  * capa explorer caches results to the database and reuses them across IDA sessions 
 * Reset the plugin user interface and remove highlighting from your Disassembly view by clicking the `Reset` button
-* Change your capa rules directory and other default settings by clicking `Settings`
+* Change your local capa rules directory, auto analysis settings, and other default settings by clicking the `Settings` button
 * Hover your cursor over a rule match to view the source content of the rule
 * Double-click the `Address` column to navigate your Disassembly view to the address of the associated feature
 * Double-click a result in the `Rule Information` column to expand its children
-* Select a checkbox in the `Rule Information` column to highlight the address of the associated feature in your Dissasembly view
+* Select a checkbox in the `Rule Information` column to highlight the address of the associated feature in your Disassembly view
 
 #### Tips for Rule Generator
 
@@ -122,18 +93,34 @@ downloading and using the [standard collection of capa rules](https://github.com
 * Directly edit rule text and metadata fields using the `Preview` pane
 * Change the default rule author and default rule scope displayed in the `Preview` pane by clicking `Settings`
 
+### Requirements
+
+capa explorer supports Python versions >= 3.7.x and IDA Pro versions >= 7.4. The following IDA Pro versions have been tested:
+
+* IDA 7.4
+* IDA 7.5
+* IDA 7.6 Service Pack 1
+* IDA 7.7
+* IDA 8.0
+* IDA 8.1
+* IDA 8.2
+
+capa explorer is however limited to the Python versions supported by your IDA installation (which may not include all Python versions >= 3.7.x).
+
+If you encounter issues with your specific setup, please open a new [Issue](https://github.com/mandiant/capa/issues).
+
 ## Development
 
 capa explorer is packaged with capa so you will need to install capa locally for development. You can install capa locally by following the steps outlined in `Method 3: Inspecting the capa source code` of the [capa 
-installation guide](https://github.com/fireeye/capa/blob/master/doc/installation.md#method-3-inspecting-the-capa-source-code). Once installed, copy [capa_explorer.py](https://raw.githubusercontent.com/fireeye/capa/master/capa/ida/plugin/capa_explorer.py) 
+installation guide](https://github.com/mandiant/capa/blob/master/doc/installation.md#method-3-inspecting-the-capa-source-code). Once installed, copy [capa_explorer.py](https://raw.githubusercontent.com/mandiant/capa/master/capa/ida/plugin/capa_explorer.py) 
 to your plugins directory to install capa explorer in IDA.
 
 ### Components
 
 capa explorer consists of two main components:
 
-* An [feature extractor](https://github.com/fireeye/capa/tree/master/capa/features/extractors/ida) built on top of IDA's binary analysis engine
-  * This component uses IDAPython to extract [capa features](https://github.com/fireeye/capa-rules/blob/master/doc/format.md#extracted-features) from your IDBs such as strings, 
+* An [feature extractor](https://github.com/mandiant/capa/tree/master/capa/features/extractors/ida) built on top of IDA's binary analysis engine
+  * This component uses IDAPython to extract [capa features](https://github.com/mandiant/capa-rules/blob/master/doc/format.md#extracted-features) from your IDBs such as strings, 
 disassembly, and control flow; these extracted features are used by capa to find feature combinations that result in a rule match
-* An [interactive user interface](https://github.com/fireeye/capa/tree/master/capa/ida/plugin) for displaying and exploring capa rule matches
+* An [interactive user interface](https://github.com/mandiant/capa/tree/master/capa/ida/plugin) for displaying and exploring capa rule matches
   * This component integrates the feature extractor and capa, providing an interactive user interface to dissect rule matches found by capa using features extracted directly from your IDBs

capa/ida/plugin/__init__.py

@@ -1,11 +1,10 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
-
 import logging
 
 import idaapi
@@ -18,7 +17,6 @@ logger = logging.getLogger(__name__)
 
 
 class CapaExplorerPlugin(idaapi.plugin_t):
-
     # Mandatory definitions
     PLUGIN_NAME = "FLARE capa explorer"
     PLUGIN_VERSION = "1.0.0"
@@ -27,8 +25,8 @@ class CapaExplorerPlugin(idaapi.plugin_t):
     wanted_name = PLUGIN_NAME
     wanted_hotkey = "ALT-F5"
     comment = "IDA Pro plugin for the FLARE team's capa tool to identify capabilities in executable files."
-    website = "https://github.com/fireeye/capa"
-    help = "See https://github.com/fireeye/capa/blob/master/doc/usage.md"
+    website = "https://github.com/mandiant/capa"
+    help = "See https://github.com/mandiant/capa/blob/master/doc/usage.md"
     version = ""
     flags = 0
 
@@ -86,7 +84,7 @@ class CapaExplorerPlugin(idaapi.plugin_t):
 # so we need to register a callback that's invoked from the main thread after the plugin is registered.
 #
 # after a lot of guess-and-check, we can use `UI_Hooks.updated_actions` to
-#  receive notications after IDA has created an action for each plugin.
+#  receive notifications after IDA has created an action for each plugin.
 # so, create this hook, wait for capa plugin to load, set the icon, and unhook.
 
 
@@ -94,7 +92,7 @@ class OnUpdatedActionsHook(ida_kernwin.UI_Hooks):
     """register a callback to be invoked each time the UI actions are updated"""
 
     def __init__(self, cb):
-        super(OnUpdatedActionsHook, self).__init__()
+        super().__init__()
         self.cb = cb
 
     def updated_actions(self):

capa/ida/plugin/cache.py

@@ -0,0 +1,220 @@
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
+# Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at: [package root]/LICENSE.txt
+# Unless required by applicable law or agreed to in writing, software distributed under the License
+#  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and limitations under the License.
+
+from __future__ import annotations
+
+import itertools
+import collections
+from typing import Set, Dict, List, Tuple, Union, Optional
+
+import capa.engine
+from capa.rules import Scope, RuleSet
+from capa.engine import FeatureSet, MatchResults
+from capa.features.address import NO_ADDRESS, Address
+from capa.ida.plugin.extractor import CapaExplorerFeatureExtractor
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle
+
+
+class CapaRuleGenFeatureCacheNode:
+    def __init__(
+        self,
+        inner: Optional[Union[FunctionHandle, BBHandle, InsnHandle]],
+        parent: Optional[CapaRuleGenFeatureCacheNode],
+    ):
+        self.inner: Optional[Union[FunctionHandle, BBHandle, InsnHandle]] = inner
+        self.address = NO_ADDRESS if self.inner is None else self.inner.address
+        self.parent: Optional[CapaRuleGenFeatureCacheNode] = parent
+
+        if self.parent is not None:
+            self.parent.children.add(self)
+
+        self.features: FeatureSet = collections.defaultdict(set)
+        self.children: Set[CapaRuleGenFeatureCacheNode] = set()
+
+    def __hash__(self):
+        # TODO: unique enough?
+        return hash((self.address,))
+
+    def __eq__(self, other):
+        if not isinstance(other, type(self)):
+            return NotImplemented
+        # TODO: unique enough?
+        return self.address == other.address
+
+
+class CapaRuleGenFeatureCache:
+    def __init__(self, fh_list: List[FunctionHandle], extractor: CapaExplorerFeatureExtractor):
+        self.global_features: FeatureSet = collections.defaultdict(set)
+
+        self.file_node: CapaRuleGenFeatureCacheNode = CapaRuleGenFeatureCacheNode(None, None)
+        self.func_nodes: Dict[Address, CapaRuleGenFeatureCacheNode] = {}
+        self.bb_nodes: Dict[Address, CapaRuleGenFeatureCacheNode] = {}
+        self.insn_nodes: Dict[Address, CapaRuleGenFeatureCacheNode] = {}
+
+        self._find_global_features(extractor)
+        self._find_file_features(extractor)
+        self._find_function_and_below_features(fh_list, extractor)
+
+    def _find_global_features(self, extractor: CapaExplorerFeatureExtractor):
+        for feature, addr in extractor.extract_global_features():
+            # not all global features may have virtual addresses.
+            # if not, then at least ensure the feature shows up in the index.
+            # the set of addresses will still be empty.
+            if addr is not None:
+                self.global_features[feature].add(addr)
+            else:
+                if feature not in self.global_features:
+                    self.global_features[feature] = set()
+
+    def _find_file_features(self, extractor: CapaExplorerFeatureExtractor):
+        # not all file features may have virtual addresses.
+        # if not, then at least ensure the feature shows up in the index.
+        # the set of addresses will still be empty.
+        for feature, addr in extractor.extract_file_features():
+            if addr is not None:
+                self.file_node.features[feature].add(addr)
+            else:
+                if feature not in self.file_node.features:
+                    self.file_node.features[feature] = set()
+
+    def _find_function_and_below_features(self, fh_list: List[FunctionHandle], extractor: CapaExplorerFeatureExtractor):
+        for fh in fh_list:
+            f_node: CapaRuleGenFeatureCacheNode = CapaRuleGenFeatureCacheNode(fh, self.file_node)
+
+            # extract basic block and below features
+            for bbh in extractor.get_basic_blocks(fh):
+                bb_node: CapaRuleGenFeatureCacheNode = CapaRuleGenFeatureCacheNode(bbh, f_node)
+
+                # extract instruction features
+                for ih in extractor.get_instructions(fh, bbh):
+                    inode: CapaRuleGenFeatureCacheNode = CapaRuleGenFeatureCacheNode(ih, bb_node)
+
+                    for feature, addr in extractor.extract_insn_features(fh, bbh, ih):
+                        inode.features[feature].add(addr)
+
+                    self.insn_nodes[inode.address] = inode
+
+                # extract basic block features
+                for feature, addr in extractor.extract_basic_block_features(fh, bbh):
+                    bb_node.features[feature].add(addr)
+
+                # store basic block features in cache and function parent
+                self.bb_nodes[bb_node.address] = bb_node
+
+            # extract function features
+            for feature, addr in extractor.extract_function_features(fh):
+                f_node.features[feature].add(addr)
+
+            self.func_nodes[f_node.address] = f_node
+
+    def _find_instruction_capabilities(
+        self, ruleset: RuleSet, insn: CapaRuleGenFeatureCacheNode
+    ) -> Tuple[FeatureSet, MatchResults]:
+        features: FeatureSet = collections.defaultdict(set)
+
+        for feature, locs in itertools.chain(insn.features.items(), self.global_features.items()):
+            features[feature].update(locs)
+
+        _, matches = ruleset.match(Scope.INSTRUCTION, features, insn.address)
+        for name, result in matches.items():
+            rule = ruleset[name]
+            for addr, _ in result:
+                capa.engine.index_rule_matches(features, rule, [addr])
+
+        return features, matches
+
+    def _find_basic_block_capabilities(
+        self, ruleset: RuleSet, bb: CapaRuleGenFeatureCacheNode
+    ) -> Tuple[FeatureSet, MatchResults, MatchResults]:
+        features: FeatureSet = collections.defaultdict(set)
+        insn_matches: MatchResults = collections.defaultdict(list)
+
+        for insn in bb.children:
+            ifeatures, imatches = self._find_instruction_capabilities(ruleset, insn)
+            for feature, locs in ifeatures.items():
+                features[feature].update(locs)
+            for name, result in imatches.items():
+                insn_matches[name].extend(result)
+
+        for feature, locs in itertools.chain(bb.features.items(), self.global_features.items()):
+            features[feature].update(locs)
+
+        _, matches = ruleset.match(Scope.BASIC_BLOCK, features, bb.address)
+        for name, result in matches.items():
+            rule = ruleset[name]
+            for loc, _ in result:
+                capa.engine.index_rule_matches(features, rule, [loc])
+
+        return features, matches, insn_matches
+
+    def find_code_capabilities(
+        self, ruleset: RuleSet, fh: FunctionHandle
+    ) -> Tuple[FeatureSet, MatchResults, MatchResults, MatchResults]:
+        f_node: Optional[CapaRuleGenFeatureCacheNode] = self.func_nodes.get(fh.address, None)
+        if f_node is None:
+            return {}, {}, {}, {}
+
+        insn_matches: MatchResults = collections.defaultdict(list)
+        bb_matches: MatchResults = collections.defaultdict(list)
+        function_features: FeatureSet = collections.defaultdict(set)
+
+        for bb in f_node.children:
+            features, bmatches, imatches = self._find_basic_block_capabilities(ruleset, bb)
+            for feature, locs in features.items():
+                function_features[feature].update(locs)
+            for name, result in bmatches.items():
+                bb_matches[name].extend(result)
+            for name, result in imatches.items():
+                insn_matches[name].extend(result)
+
+        for feature, locs in itertools.chain(f_node.features.items(), self.global_features.items()):
+            function_features[feature].update(locs)
+
+        _, function_matches = ruleset.match(Scope.FUNCTION, function_features, f_node.address)
+        return function_features, function_matches, bb_matches, insn_matches
+
+    def find_file_capabilities(self, ruleset: RuleSet) -> Tuple[FeatureSet, MatchResults]:
+        features: FeatureSet = collections.defaultdict(set)
+
+        for func_node in self.file_node.children:
+            assert func_node.inner is not None
+            assert isinstance(func_node.inner, FunctionHandle)
+
+            func_features, _, _, _ = self.find_code_capabilities(ruleset, func_node.inner)
+            for feature, locs in func_features.items():
+                features[feature].update(locs)
+
+        for feature, locs in itertools.chain(self.file_node.features.items(), self.global_features.items()):
+            features[feature].update(locs)
+
+        _, matches = ruleset.match(Scope.FILE, features, NO_ADDRESS)
+        return features, matches
+
+    def get_all_function_features(self, fh: FunctionHandle) -> FeatureSet:
+        f_node: Optional[CapaRuleGenFeatureCacheNode] = self.func_nodes.get(fh.address, None)
+        if f_node is None:
+            return {}
+
+        all_function_features: FeatureSet = collections.defaultdict(set)
+        all_function_features.update(f_node.features)
+
+        for bb_node in f_node.children:
+            for i_node in bb_node.children:
+                for feature, locs in i_node.features.items():
+                    all_function_features[feature].update(locs)
+            for feature, locs in bb_node.features.items():
+                all_function_features[feature].update(locs)
+
+        # include global features just once
+        for feature, locs in self.global_features.items():
+            all_function_features[feature].update(locs)
+
+        return all_function_features
+
+    def get_all_file_features(self):
+        yield from itertools.chain(self.file_node.features.items(), self.global_features.items())

capa/ida/plugin/capa_explorer.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt

capa/ida/plugin/error.py

@@ -0,0 +1,13 @@
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
+# Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at: [package root]/LICENSE.txt
+# Unless required by applicable law or agreed to in writing, software distributed under the License
+#  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and limitations under the License.
+
+
+class UserCancelledError(Exception):
+    """throw exception when user cancels action"""
+
+    pass

capa/ida/plugin/extractor.py

@@ -0,0 +1,44 @@
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
+# Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at: [package root]/LICENSE.txt
+# Unless required by applicable law or agreed to in writing, software distributed under the License
+#  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and limitations under the License.
+
+import ida_kernwin
+from PyQt5 import QtCore
+
+from capa.ida.plugin.error import UserCancelledError
+from capa.features.extractors.ida.extractor import IdaFeatureExtractor
+from capa.features.extractors.base_extractor import FunctionHandle
+
+
+class CapaExplorerProgressIndicator(QtCore.QObject):
+    """implement progress signal, used during feature extraction"""
+
+    progress = QtCore.pyqtSignal(str)
+
+    def update(self, text):
+        """emit progress update
+
+        check if user cancelled action, raise exception for parent function to catch
+        """
+        if ida_kernwin.user_cancelled():
+            raise UserCancelledError("user cancelled")
+        self.progress.emit("extracting features from %s" % text)
+
+
+class CapaExplorerFeatureExtractor(IdaFeatureExtractor):
+    """subclass the IdaFeatureExtractor
+
+    track progress during feature extraction, also allow user to cancel feature extraction
+    """
+
+    def __init__(self):
+        super().__init__()
+        self.indicator = CapaExplorerProgressIndicator()
+
+    def extract_function_features(self, fh: FunctionHandle):
+        self.indicator.update("function at 0x%X" % fh.inner.start_ea)
+        return super().extract_function_features(fh)

capa/ida/plugin/hooks.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -16,7 +16,7 @@ class CapaExplorerIdaHooks(idaapi.UI_Hooks):
         @param screen_ea_changed_hook: function hook for IDA screen ea changed
         @param action_hooks: dict of IDA action handles
         """
-        super(CapaExplorerIdaHooks, self).__init__()
+        super().__init__()
 
         self.screen_ea_changed_hook = screen_ea_changed_hook
         self.process_action_hooks = action_hooks

capa/ida/plugin/item.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -7,12 +7,14 @@
 # See the License for the specific language governing permissions and limitations under the License.
 
 import codecs
+from typing import List, Iterator, Optional
 
 import idc
 import idaapi
 from PyQt5 import QtCore
 
 import capa.ida.helpers
+from capa.features.address import Address, FileOffsetAddress, AbsoluteVirtualAddress
 
 
 def info_to_name(display):
@@ -26,19 +28,19 @@ def info_to_name(display):
         return ""
 
 
-def location_to_hex(location):
-    """convert location to hex for display"""
-    return "%08X" % location
+def ea_to_hex(ea):
+    """convert effective address (ea) to hex for display"""
+    return "%08X" % ea
 
 
 class CapaExplorerDataItem:
     """store data for CapaExplorerDataModel"""
 
-    def __init__(self, parent, data, can_check=True):
+    def __init__(self, parent: Optional["CapaExplorerDataItem"], data: List[str], can_check=True):
         """initialize item"""
         self.pred = parent
         self._data = data
-        self.children = []
+        self._children: List["CapaExplorerDataItem"] = []
         self._checked = False
         self._can_check = can_check
 
@@ -76,29 +78,29 @@ class CapaExplorerDataItem:
         """get item is checked"""
         return self._checked
 
-    def appendChild(self, item):
+    def appendChild(self, item: "CapaExplorerDataItem"):
         """add a new child to specified item
 
         @param item: CapaExplorerDataItem
         """
-        self.children.append(item)
+        self._children.append(item)
 
-    def child(self, row):
+    def child(self, row: int) -> "CapaExplorerDataItem":
         """get child row
 
         @param row: row number
         """
-        return self.children[row]
+        return self._children[row]
 
-    def childCount(self):
+    def childCount(self) -> int:
         """get child count"""
-        return len(self.children)
+        return len(self._children)
 
-    def columnCount(self):
+    def columnCount(self) -> int:
         """get column count"""
         return len(self._data)
 
-    def data(self, column):
+    def data(self, column: int) -> Optional[str]:
         """get data at column
 
         @param: column number
@@ -108,17 +110,17 @@ class CapaExplorerDataItem:
         except IndexError:
             return None
 
-    def parent(self):
+    def parent(self) -> Optional["CapaExplorerDataItem"]:
         """get parent"""
         return self.pred
 
-    def row(self):
+    def row(self) -> int:
         """get row location"""
         if self.pred:
-            return self.pred.children.index(self)
+            return self.pred._children.index(self)
         return 0
 
-    def setData(self, column, value):
+    def setData(self, column: int, value: str):
         """set data in column
 
         @param column: column number
@@ -126,14 +128,14 @@ class CapaExplorerDataItem:
         """
         self._data[column] = value
 
-    def children(self):
+    def children(self) -> Iterator["CapaExplorerDataItem"]:
         """yield children"""
-        for child in self.children:
+        for child in self._children:
             yield child
 
     def removeChildren(self):
         """remove children"""
-        del self.children[:]
+        del self._children[:]
 
     def __str__(self):
         """get string representation of columns
@@ -148,7 +150,7 @@ class CapaExplorerDataItem:
         return self._data[0]
 
     @property
-    def location(self):
+    def location(self) -> Optional[int]:
         """return data stored in location column"""
         try:
             # address stored as str, convert to int before return
@@ -167,7 +169,9 @@ class CapaExplorerRuleItem(CapaExplorerDataItem):
 
     fmt = "%s (%d matches)"
 
-    def __init__(self, parent, name, namespace, count, source, can_check=True):
+    def __init__(
+        self, parent: CapaExplorerDataItem, name: str, namespace: str, count: int, source: str, can_check=True
+    ):
         """initialize item
 
         @param parent: parent node
@@ -177,7 +181,7 @@ class CapaExplorerRuleItem(CapaExplorerDataItem):
         @param source: rule source (tooltip)
         """
         display = self.fmt % (name, count) if count > 1 else name
-        super(CapaExplorerRuleItem, self).__init__(parent, [display, "", namespace], can_check)
+        super().__init__(parent, [display, "", namespace], can_check)
         self._source = source
 
     @property
@@ -189,14 +193,14 @@ class CapaExplorerRuleItem(CapaExplorerDataItem):
 class CapaExplorerRuleMatchItem(CapaExplorerDataItem):
     """store data for rule match"""
 
-    def __init__(self, parent, display, source=""):
+    def __init__(self, parent: CapaExplorerDataItem, display: str, source=""):
         """initialize item
 
         @param parent: parent node
         @param display: text to display in UI
         @param source: rule match source to display (tooltip)
         """
-        super(CapaExplorerRuleMatchItem, self).__init__(parent, [display, "", ""])
+        super().__init__(parent, [display, "", ""])
         self._source = source
 
     @property
@@ -210,20 +214,20 @@ class CapaExplorerFunctionItem(CapaExplorerDataItem):
 
     fmt = "function(%s)"
 
-    def __init__(self, parent, location, can_check=True):
+    def __init__(self, parent: CapaExplorerDataItem, location: Address, can_check=True):
         """initialize item
 
         @param parent: parent node
         @param location: virtual address of function as seen by IDA
         """
-        super(CapaExplorerFunctionItem, self).__init__(
-            parent, [self.fmt % idaapi.get_name(location), location_to_hex(location), ""], can_check
-        )
+        assert isinstance(location, AbsoluteVirtualAddress)
+        ea = int(location)
+        super().__init__(parent, [self.fmt % idaapi.get_name(ea), ea_to_hex(ea), ""], can_check)
 
     @property
     def info(self):
         """return function name"""
-        info = super(CapaExplorerFunctionItem, self).info
+        info = super().info
         display = info_to_name(info)
         return display if display else info
 
@@ -243,13 +247,13 @@ class CapaExplorerSubscopeItem(CapaExplorerDataItem):
 
     fmt = "subscope(%s)"
 
-    def __init__(self, parent, scope):
+    def __init__(self, parent: CapaExplorerDataItem, scope):
         """initialize item
 
         @param parent: parent node
         @param scope: subscope name
         """
-        super(CapaExplorerSubscopeItem, self).__init__(parent, [self.fmt % scope, "", ""])
+        super().__init__(parent, [self.fmt % scope, "", ""])
 
 
 class CapaExplorerBlockItem(CapaExplorerDataItem):
@@ -257,19 +261,29 @@ class CapaExplorerBlockItem(CapaExplorerDataItem):
 
     fmt = "basic block(loc_%08X)"
 
-    def __init__(self, parent, location):
+    def __init__(self, parent: CapaExplorerDataItem, location: Address):
         """initialize item
 
         @param parent: parent node
         @param location: virtual address of basic block as seen by IDA
         """
-        super(CapaExplorerBlockItem, self).__init__(parent, [self.fmt % location, location_to_hex(location), ""])
+        assert isinstance(location, AbsoluteVirtualAddress)
+        ea = int(location)
+        super().__init__(parent, [self.fmt % ea, ea_to_hex(ea), ""])
+
+
+class CapaExplorerInstructionItem(CapaExplorerBlockItem):
+    """store data for instruction match"""
+
+    fmt = "instruction(loc_%08X)"
 
 
 class CapaExplorerDefaultItem(CapaExplorerDataItem):
     """store data for default match e.g. statement (and, or)"""
 
-    def __init__(self, parent, display, details="", location=None):
+    def __init__(
+        self, parent: CapaExplorerDataItem, display: str, details: str = "", location: Optional[Address] = None
+    ):
         """initialize item
 
         @param parent: parent node
@@ -277,14 +291,20 @@ class CapaExplorerDefaultItem(CapaExplorerDataItem):
         @param details: text to display in details section of UI
         @param location: virtual address as seen by IDA
         """
-        location = location_to_hex(location) if location else ""
-        super(CapaExplorerDefaultItem, self).__init__(parent, [display, location, details])
+        ea = None
+        if location:
+            assert isinstance(location, AbsoluteVirtualAddress)
+            ea = int(location)
+
+        super().__init__(parent, [display, ea_to_hex(ea) if ea is not None else "", details])
 
 
 class CapaExplorerFeatureItem(CapaExplorerDataItem):
     """store data for feature match"""
 
-    def __init__(self, parent, display, location="", details=""):
+    def __init__(
+        self, parent: CapaExplorerDataItem, display: str, location: Optional[Address] = None, details: str = ""
+    ):
         """initialize item
 
         @param parent: parent node
@@ -292,14 +312,18 @@ class CapaExplorerFeatureItem(CapaExplorerDataItem):
         @param details: text to display in details section of UI
         @param location: virtual address as seen by IDA
         """
-        location = location_to_hex(location) if location else ""
-        super(CapaExplorerFeatureItem, self).__init__(parent, [display, location, details])
+        if location:
+            assert isinstance(location, (AbsoluteVirtualAddress, FileOffsetAddress))
+            ea = int(location)
+            super().__init__(parent, [display, ea_to_hex(ea), details])
+        else:
+            super().__init__(parent, [display, "", details])
 
 
 class CapaExplorerInstructionViewItem(CapaExplorerFeatureItem):
     """store data for instruction match"""
 
-    def __init__(self, parent, display, location):
+    def __init__(self, parent: CapaExplorerDataItem, display: str, location: Address):
         """initialize item
 
         details section shows disassembly view for match
@@ -308,15 +332,17 @@ class CapaExplorerInstructionViewItem(CapaExplorerFeatureItem):
         @param display: text to display in UI
         @param location: virtual address as seen by IDA
         """
-        details = capa.ida.helpers.get_disasm_line(location)
-        super(CapaExplorerInstructionViewItem, self).__init__(parent, display, location=location, details=details)
-        self.ida_highlight = idc.get_color(location, idc.CIC_ITEM)
+        assert isinstance(location, AbsoluteVirtualAddress)
+        ea = int(location)
+        details = capa.ida.helpers.get_disasm_line(ea)
+        super().__init__(parent, display, location=location, details=details)
+        self.ida_highlight = idc.get_color(ea, idc.CIC_ITEM)
 
 
 class CapaExplorerByteViewItem(CapaExplorerFeatureItem):
     """store data for byte match"""
 
-    def __init__(self, parent, display, location):
+    def __init__(self, parent: CapaExplorerDataItem, display: str, location: Address):
         """initialize item
 
         details section shows byte preview for match
@@ -325,26 +351,32 @@ class CapaExplorerByteViewItem(CapaExplorerFeatureItem):
         @param display: text to display in UI
         @param location: virtual address as seen by IDA
         """
-        byte_snap = idaapi.get_bytes(location, 32)
+        assert isinstance(location, (AbsoluteVirtualAddress, FileOffsetAddress))
+        ea = int(location)
+
+        byte_snap = idaapi.get_bytes(ea, 32)
 
         details = ""
         if byte_snap:
             byte_snap = codecs.encode(byte_snap, "hex").upper()
             details = " ".join([byte_snap[i : i + 2].decode() for i in range(0, len(byte_snap), 2)])
 
-        super(CapaExplorerByteViewItem, self).__init__(parent, display, location=location, details=details)
-        self.ida_highlight = idc.get_color(location, idc.CIC_ITEM)
+        super().__init__(parent, display, location=location, details=details)
+        self.ida_highlight = idc.get_color(ea, idc.CIC_ITEM)
 
 
 class CapaExplorerStringViewItem(CapaExplorerFeatureItem):
     """store data for string match"""
 
-    def __init__(self, parent, display, location, value):
+    def __init__(self, parent: CapaExplorerDataItem, display: str, location: Address, value: str):
         """initialize item
 
         @param parent: parent node
         @param display: text to display in UI
         @param location: virtual address as seen by IDA
         """
-        super(CapaExplorerStringViewItem, self).__init__(parent, display, location=location, details=value)
-        self.ida_highlight = idc.get_color(location, idc.CIC_ITEM)
+        assert isinstance(location, (AbsoluteVirtualAddress, FileOffsetAddress))
+        ea = int(location)
+
+        super().__init__(parent, display, location=location, details=value)
+        self.ida_highlight = idc.get_color(ea, idc.CIC_ITEM)

capa/ida/plugin/model.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -6,7 +6,8 @@
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
 
-from collections import deque, defaultdict
+from typing import Set, Dict, List, Tuple, Optional
+from collections import deque
 
 import idc
 import idaapi
@@ -16,6 +17,9 @@ import capa.rules
 import capa.ida.helpers
 import capa.render.utils as rutils
 import capa.features.common
+import capa.features.freeze as frz
+import capa.render.result_document as rd
+import capa.features.freeze.features as frzf
 from capa.ida.plugin.item import (
     CapaExplorerDataItem,
     CapaExplorerRuleItem,
@@ -27,8 +31,10 @@ from capa.ida.plugin.item import (
     CapaExplorerSubscopeItem,
     CapaExplorerRuleMatchItem,
     CapaExplorerStringViewItem,
+    CapaExplorerInstructionItem,
     CapaExplorerInstructionViewItem,
 )
+from capa.features.address import Address, AbsoluteVirtualAddress
 
 # default highlight color used in IDA window
 DEFAULT_HIGHLIGHT = 0xE6C700
@@ -45,7 +51,7 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
 
     def __init__(self, parent=None):
         """initialize model"""
-        super(CapaExplorerDataModel, self).__init__(parent)
+        super().__init__(parent)
         # root node does not have parent, contains header columns
         self.root_node = CapaExplorerDataItem(None, ["Rule Information", "Address", "Details"])
 
@@ -137,6 +143,7 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
                     CapaExplorerFunctionItem,
                     CapaExplorerFeatureItem,
                     CapaExplorerSubscopeItem,
+                    CapaExplorerInstructionItem,
                 ),
             )
             and column == CapaExplorerDataModel.COLUMN_INDEX_RULE_INFORMATION
@@ -342,7 +349,14 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
 
         return item.childCount()
 
-    def render_capa_doc_statement_node(self, parent, statement, locations, doc):
+    def render_capa_doc_statement_node(
+        self,
+        parent: CapaExplorerDataItem,
+        match: rd.Match,
+        statement: rd.Statement,
+        locations: List[Address],
+        doc: rd.ResultDocument,
+    ):
         """render capa statement read from doc
 
         @param parent: parent to which new child is assigned
@@ -350,132 +364,154 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
         @param locations: locations of children (applies to range only?)
         @param doc: result doc
         """
-        if statement["type"] in ("and", "or", "optional"):
-            display = statement["type"]
-            if statement.get("description"):
-                display += " (%s)" % statement["description"]
-            return CapaExplorerDefaultItem(parent, display)
-        elif statement["type"] == "not":
+
+        if isinstance(statement, rd.CompoundStatement):
+            if statement.type != rd.CompoundStatementType.NOT:
+                display = statement.type
+                if statement.description:
+                    display += " (%s)" % statement.description
+                return CapaExplorerDefaultItem(parent, display)
+        elif isinstance(statement, rd.CompoundStatement) and statement.type == rd.CompoundStatementType.NOT:
             # TODO: do we display 'not'
             pass
-        elif statement["type"] == "some":
-            display = "%d or more" % statement["count"]
-            if statement.get("description"):
-                display += " (%s)" % statement["description"]
+        elif isinstance(statement, rd.SomeStatement):
+            display = "%d or more" % statement.count
+            if statement.description:
+                display += " (%s)" % statement.description
             return CapaExplorerDefaultItem(parent, display)
-        elif statement["type"] == "range":
+        elif isinstance(statement, rd.RangeStatement):
             # `range` is a weird node, its almost a hybrid of statement + feature.
             # it is a specific feature repeated multiple times.
             # there's no additional logic in the feature part, just the existence of a feature.
             # so, we have to inline some of the feature rendering here.
-            display = "count(%s): " % self.capa_doc_feature_to_display(statement["child"])
+            display = "count(%s): " % self.capa_doc_feature_to_display(statement.child)
 
-            if statement["max"] == statement["min"]:
-                display += "%d" % (statement["min"])
-            elif statement["min"] == 0:
-                display += "%d or fewer" % (statement["max"])
-            elif statement["max"] == (1 << 64 - 1):
-                display += "%d or more" % (statement["min"])
+            if statement.max == statement.min:
+                display += "%d" % (statement.min)
+            elif statement.min == 0:
+                display += "%d or fewer" % (statement.max)
+            elif statement.max == (1 << 64 - 1):
+                display += "%d or more" % (statement.min)
             else:
-                display += "between %d and %d" % (statement["min"], statement["max"])
+                display += "between %d and %d" % (statement.min, statement.max)
 
-            if statement.get("description"):
-                display += " (%s)" % statement["description"]
+            if statement.description:
+                display += " (%s)" % statement.description
 
             parent2 = CapaExplorerFeatureItem(parent, display=display)
 
             for location in locations:
                 # for each location render child node for range statement
-                self.render_capa_doc_feature(parent2, statement["child"], location, doc)
+                self.render_capa_doc_feature(parent2, match, statement.child, location, doc)
 
             return parent2
-        elif statement["type"] == "subscope":
-            display = statement[statement["type"]]
-            if statement.get("description"):
-                display += " (%s)" % statement["description"]
+        elif isinstance(statement, rd.SubscopeStatement):
+            display = str(statement.scope)
+            if statement.description:
+                display += " (%s)" % statement.description
             return CapaExplorerSubscopeItem(parent, display)
         else:
             raise RuntimeError("unexpected match statement type: " + str(statement))
 
-    def render_capa_doc_match(self, parent, match, doc):
+    def render_capa_doc_match(self, parent: CapaExplorerDataItem, match: rd.Match, doc: rd.ResultDocument):
         """render capa match read from doc
 
         @param parent: parent node to which new child is assigned
         @param match: match read from doc
         @param doc: result doc
         """
-        if not match["success"]:
+        if not match.success:
             # TODO: display failed branches at some point? Help with debugging rules?
             return
 
         # optional statement with no successful children is empty
-        if match["node"].get("statement", {}).get("type") == "optional" and not any(
-            map(lambda m: m["success"], match["children"])
-        ):
-            return
+        if isinstance(match.node, rd.StatementNode) and match.node.statement.type == rd.CompoundStatementType.OPTIONAL:
+            if not any(map(lambda m: m.success, match.children)):
+                return
 
-        if match["node"]["type"] == "statement":
+        if isinstance(match.node, rd.StatementNode):
             parent2 = self.render_capa_doc_statement_node(
-                parent, match["node"]["statement"], match.get("locations", []), doc
+                parent, match, match.node.statement, [addr.to_capa() for addr in match.locations], doc
             )
-        elif match["node"]["type"] == "feature":
+        elif isinstance(match.node, rd.FeatureNode):
             parent2 = self.render_capa_doc_feature_node(
-                parent, match["node"]["feature"], match.get("locations", []), doc
+                parent, match, match.node.feature, [addr.to_capa() for addr in match.locations], doc
             )
         else:
-            raise RuntimeError("unexpected node type: " + str(match["node"]["type"]))
+            raise RuntimeError("unexpected node type: " + str(match.node.type))
 
-        for child in match.get("children", []):
+        for child in match.children:
             self.render_capa_doc_match(parent2, child, doc)
 
-    def render_capa_doc_by_function(self, doc):
-        """ """
-        matches_by_function = {}
+    def render_capa_doc_by_function(self, doc: rd.ResultDocument):
+        """render rule matches by function meaning each rule match is nested under function where it was found"""
+        matches_by_function: Dict[AbsoluteVirtualAddress, Tuple[CapaExplorerFunctionItem, Set[str]]] = {}
         for rule in rutils.capability_rules(doc):
-            for ea in rule["matches"].keys():
-                ea = capa.ida.helpers.get_func_start_ea(ea)
-                if ea is None:
-                    # file scope, skip rendering in this mode
+            match_eas: List[int] = []
+
+            # initial pass of rule matches
+            for addr_, _ in rule.matches:
+                addr: Address = addr_.to_capa()
+                if isinstance(addr, AbsoluteVirtualAddress):
+                    match_eas.append(int(addr))
+
+            for ea in match_eas:
+                func_ea: Optional[int] = capa.ida.helpers.get_func_start_ea(ea)
+                if func_ea is None:
+                    # rule match address is not located in a defined function
                     continue
-                if not matches_by_function.get(ea, ()):
-                    # new function root
-                    matches_by_function[ea] = (CapaExplorerFunctionItem(self.root_node, ea, can_check=False), [])
-                function_root, match_cache = matches_by_function[ea]
-                if rule["meta"]["name"] in match_cache:
-                    # rule match already rendered for this function root, skip it
+
+                func_address: AbsoluteVirtualAddress = AbsoluteVirtualAddress(func_ea)
+                if not matches_by_function.get(func_address, ()):
+                    # create a new function root to nest its rule matches; Note: we must use the address of the
+                    # function here so everything is displayed properly
+                    matches_by_function[func_address] = (
+                        CapaExplorerFunctionItem(self.root_node, func_address, can_check=False),
+                        set(),
+                    )
+
+                func_root, func_match_cache = matches_by_function[func_address]
+                if rule.meta.name in func_match_cache:
+                    # only nest each rule once, so if found, skip
                     continue
-                match_cache.append(rule["meta"]["name"])
+
+                # add matched rule to its function cache; create a new rule node whose parent is the matched
+                # function node
+                func_match_cache.add(rule.meta.name)
                 CapaExplorerRuleItem(
-                    function_root,
-                    rule["meta"]["name"],
-                    rule["meta"].get("namespace"),
-                    len(rule["matches"]),
-                    rule["source"],
+                    func_root,
+                    rule.meta.name,
+                    rule.meta.namespace or "",
+                    len([ea for ea in match_eas if capa.ida.helpers.get_func_start_ea(ea) == func_ea]),
+                    rule.source,
                     can_check=False,
                 )
 
-    def render_capa_doc_by_program(self, doc):
+    def render_capa_doc_by_program(self, doc: rd.ResultDocument):
         """ """
         for rule in rutils.capability_rules(doc):
-            rule_name = rule["meta"]["name"]
-            rule_namespace = rule["meta"].get("namespace")
-            parent = CapaExplorerRuleItem(
-                self.root_node, rule_name, rule_namespace, len(rule["matches"]), rule["source"]
-            )
+            rule_name = rule.meta.name
+            rule_namespace = rule.meta.namespace or ""
+            parent = CapaExplorerRuleItem(self.root_node, rule_name, rule_namespace, len(rule.matches), rule.source)
 
-            for (location, match) in doc["rules"][rule["meta"]["name"]]["matches"].items():
-                if rule["meta"]["scope"] == capa.rules.FILE_SCOPE:
+            for location_, match in rule.matches:
+                location = location_.to_capa()
+
+                parent2: CapaExplorerDataItem
+                if rule.meta.scope == capa.rules.FILE_SCOPE:
                     parent2 = parent
-                elif rule["meta"]["scope"] == capa.rules.FUNCTION_SCOPE:
+                elif rule.meta.scope == capa.rules.FUNCTION_SCOPE:
                     parent2 = CapaExplorerFunctionItem(parent, location)
-                elif rule["meta"]["scope"] == capa.rules.BASIC_BLOCK_SCOPE:
+                elif rule.meta.scope == capa.rules.BASIC_BLOCK_SCOPE:
                     parent2 = CapaExplorerBlockItem(parent, location)
+                elif rule.meta.scope == capa.rules.INSTRUCTION_SCOPE:
+                    parent2 = CapaExplorerInstructionItem(parent, location)
                 else:
-                    raise RuntimeError("unexpected rule scope: " + str(rule["meta"]["scope"]))
+                    raise RuntimeError("unexpected rule scope: " + str(rule.meta.scope))
 
                 self.render_capa_doc_match(parent2, match, doc)
 
-    def render_capa_doc(self, doc, by_function):
+    def render_capa_doc(self, doc: rd.ResultDocument, by_function: bool):
         """render capa features specified in doc
 
         @param doc: capa result doc
@@ -491,27 +527,44 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
         # inform model changes have ended
         self.endResetModel()
 
-    def capa_doc_feature_to_display(self, feature):
+    def capa_doc_feature_to_display(self, feature: frzf.Feature):
         """convert capa doc feature type string to display string for ui
 
         @param feature: capa feature read from doc
         """
-        key = feature["type"]
-        value = feature[feature["type"]]
+        key = feature.type
+        value = feature.dict(by_alias=True).get(feature.type)
+
         if value:
-            if key == "string":
+            if isinstance(feature, frzf.StringFeature):
                 value = '"%s"' % capa.features.common.escape_string(value)
-            if feature.get("description", ""):
-                return "%s(%s = %s)" % (key, value, feature["description"])
+
+            if isinstance(feature, frzf.PropertyFeature) and feature.access is not None:
+                key = f"property/{feature.access}"
+            elif isinstance(feature, frzf.OperandNumberFeature):
+                key = f"operand[{feature.index}].number"
+            elif isinstance(feature, frzf.OperandOffsetFeature):
+                key = f"operand[{feature.index}].offset"
+
+            if feature.description:
+                return "%s(%s = %s)" % (key, value, feature.description)
             else:
                 return "%s(%s)" % (key, value)
         else:
             return "%s" % key
 
-    def render_capa_doc_feature_node(self, parent, feature, locations, doc):
+    def render_capa_doc_feature_node(
+        self,
+        parent: CapaExplorerDataItem,
+        match: rd.Match,
+        feature: frzf.Feature,
+        locations: List[Address],
+        doc: rd.ResultDocument,
+    ):
         """process capa doc feature node
 
         @param parent: parent node to which child is assigned
+        @param match: match information
         @param feature: capa doc feature node
         @param locations: locations identified for feature
         @param doc: capa doc
@@ -522,6 +575,7 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
             # only one location for feature so no need to nest children
             parent2 = self.render_capa_doc_feature(
                 parent,
+                match,
                 feature,
                 next(iter(locations)),
                 doc,
@@ -532,81 +586,114 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
             parent2 = CapaExplorerFeatureItem(parent, display)
 
             for location in sorted(locations):
-                self.render_capa_doc_feature(parent2, feature, location, doc)
+                self.render_capa_doc_feature(parent2, match, feature, location, doc)
 
         return parent2
 
-    def render_capa_doc_feature(self, parent, feature, location, doc, display="-"):
+    def render_capa_doc_feature(
+        self,
+        parent: CapaExplorerDataItem,
+        match: rd.Match,
+        feature: frzf.Feature,
+        location: Address,
+        doc: rd.ResultDocument,
+        display="-",
+    ):
         """render capa feature read from doc
 
         @param parent: parent node to which new child is assigned
+        @param match: match information
         @param feature: feature read from doc
         @param doc: capa feature doc
         @param location: address of feature
         @param display: text to display in plugin UI
         """
+
         # special handling for characteristic pending type
-        if feature["type"] == "characteristic":
-            if feature[feature["type"]] in ("embedded pe",):
+        if isinstance(feature, frzf.CharacteristicFeature):
+            characteristic = feature.characteristic
+            if characteristic in ("embedded pe",):
                 return CapaExplorerByteViewItem(parent, display, location)
 
-            if feature[feature["type"]] in ("loop", "recursive call", "tight loop"):
+            if characteristic in ("loop", "recursive call", "tight loop"):
                 return CapaExplorerFeatureItem(parent, display=display)
 
             # default to instruction view for all other characteristics
             return CapaExplorerInstructionViewItem(parent, display, location)
 
-        if feature["type"] == "match":
+        elif isinstance(feature, frzf.MatchFeature):
             # display content of rule for all rule matches
-            return CapaExplorerRuleMatchItem(
-                parent, display, source=doc["rules"].get(feature[feature["type"]], {}).get("source", "")
-            )
+            matched_rule_source = ""
 
-        if feature["type"] in ("regex", "substring"):
-            for s, locations in feature["matches"].items():
-                if location in locations:
-                    return CapaExplorerStringViewItem(
-                        parent, display, location, '"' + capa.features.common.escape_string(s) + '"'
-                    )
+            # check if match is a matched rule
+            matched_rule = doc.rules.get(feature.match, None)
+            if matched_rule is not None:
+                matched_rule_source = matched_rule.source
+
+            return CapaExplorerRuleMatchItem(parent, display, source=matched_rule_source)
+
+        elif isinstance(feature, (frzf.RegexFeature, frzf.SubstringFeature)):
+            for capture, addrs in sorted(match.captures.items()):
+                for addr in addrs:
+                    assert isinstance(addr, frz.Address)
+                    if location == addr.value:
+                        return CapaExplorerStringViewItem(
+                            parent, display, location, '"' + capa.features.common.escape_string(capture) + '"'
+                        )
 
             # programming error: the given location should always be found in the regex matches
             raise ValueError("regex match at location not found")
 
-        if feature["type"] == "basicblock":
+        elif isinstance(feature, frzf.BasicBlockFeature):
             return CapaExplorerBlockItem(parent, location)
 
-        if feature["type"] in (
-            "bytes",
-            "api",
-            "mnemonic",
-            "number",
-            "offset",
-            "number/x32",
-            "number/x64",
-            "offset/x32",
-            "offset/x64",
+        elif isinstance(
+            feature,
+            (
+                frzf.BytesFeature,
+                frzf.APIFeature,
+                frzf.MnemonicFeature,
+                frzf.NumberFeature,
+                frzf.OffsetFeature,
+                frzf.OperandNumberFeature,
+                frzf.OperandOffsetFeature,
+            ),
         ):
             # display instruction preview
             return CapaExplorerInstructionViewItem(parent, display, location)
 
-        if feature["type"] in ("section",):
+        elif isinstance(feature, frzf.SectionFeature):
             # display byte preview
             return CapaExplorerByteViewItem(parent, display, location)
 
-        if feature["type"] in ("string",):
+        elif isinstance(feature, frzf.StringFeature):
             # display string preview
             return CapaExplorerStringViewItem(
-                parent, display, location, '"%s"' % capa.features.common.escape_string(feature[feature["type"]])
+                parent, display, location, '"%s"' % capa.features.common.escape_string(feature.string)
             )
 
-        if feature["type"] in ("import", "export", "function-name"):
+        elif isinstance(
+            feature,
+            (
+                frzf.ImportFeature,
+                frzf.ExportFeature,
+                frzf.FunctionNameFeature,
+            ),
+        ):
             # display no preview
             return CapaExplorerFeatureItem(parent, location=location, display=display)
 
-        if feature["type"] in ("arch", "os", "format"):
+        elif isinstance(
+            feature,
+            (
+                frzf.ArchFeature,
+                frzf.OSFeature,
+                frzf.FormatFeature,
+            ),
+        ):
             return CapaExplorerFeatureItem(parent, display=display)
 
-        raise RuntimeError("unexpected feature type: " + str(feature["type"]))
+        raise RuntimeError("unexpected feature type: " + str(feature.type))
 
     def update_function_name(self, old_name, new_name):
         """update all instances of old function name with new function name

capa/ida/plugin/proxy.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -22,7 +22,7 @@ class CapaExplorerRangeProxyModel(QtCore.QSortFilterProxyModel):
 
     def __init__(self, parent=None):
         """initialize proxy filter"""
-        super(CapaExplorerRangeProxyModel, self).__init__(parent)
+        super().__init__(parent)
         self.min_ea = None
         self.max_ea = None
 
@@ -92,7 +92,7 @@ class CapaExplorerRangeProxyModel(QtCore.QSortFilterProxyModel):
         @param parent: QModelIndex of parent
         """
         # filter not set
-        if self.min_ea is None and self.max_ea is None:
+        if self.min_ea is None or self.max_ea is None:
             return True
 
         index = self.sourceModel().index(row, 0, parent)
@@ -145,7 +145,7 @@ class CapaExplorerSearchProxyModel(QtCore.QSortFilterProxyModel):
 
     def __init__(self, parent=None):
         """ """
-        super(CapaExplorerSearchProxyModel, self).__init__(parent)
+        super().__init__(parent)
         self.query = ""
         self.setFilterKeyColumn(-1)  # all columns
 

capa/ida/plugin/view.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -18,6 +18,7 @@ import capa.ida.helpers
 import capa.features.common
 import capa.features.basicblock
 from capa.ida.plugin.item import CapaExplorerFunctionItem
+from capa.features.address import AbsoluteVirtualAddress, _NoAddress
 from capa.ida.plugin.model import CapaExplorerDataModel
 
 MAX_SECTION_SIZE = 750
@@ -73,7 +74,7 @@ def parse_node_for_feature(feature, description, comment, depth):
     if feature.startswith("#"):
         display += "%s%s\n" % (" " * depth, feature)
     elif description:
-        if feature.startswith(("- and", "- or", "- optional", "- basic block", "- not")):
+        if feature.startswith(("- and", "- or", "- optional", "- basic block", "- not", "- instruction:")):
             display += "%s%s" % (" " * depth, feature)
             if comment:
                 display += " # %s" % comment
@@ -172,17 +173,17 @@ def resize_columns_to_content(header):
         header.resizeSection(0, MAX_SECTION_SIZE)
 
 
-class CapaExplorerRulgenPreview(QtWidgets.QTextEdit):
-
+class CapaExplorerRulegenPreview(QtWidgets.QTextEdit):
     INDENT = " " * 2
 
     def __init__(self, parent=None):
         """ """
-        super(CapaExplorerRulgenPreview, self).__init__(parent)
+        super().__init__(parent)
 
         self.setFont(QtGui.QFont("Courier", weight=QtGui.QFont.Bold))
         self.setLineWrapMode(QtWidgets.QTextEdit.NoWrap)
         self.setHorizontalScrollBarPolicy(QtCore.Qt.ScrollBarAsNeeded)
+        self.setAcceptRichText(False)
 
     def reset_view(self):
         """ """
@@ -196,9 +197,11 @@ class CapaExplorerRulgenPreview(QtWidgets.QTextEdit):
             "  meta:",
             "    name: <insert_name>",
             "    namespace: <insert_namespace>",
-            "    author: %s" % author,
+            "    authors:",
+            "      - %s" % author,
             "    scope: %s" % scope,
-            "    references: <insert_references>",
+            "    references:",
+            "      - <insert_references>",
             "    examples:",
             "      - %s:0x%X" % (capa.ida.helpers.get_file_md5().upper(), ea)
             if ea
@@ -251,7 +254,7 @@ class CapaExplorerRulgenPreview(QtWidgets.QTextEdit):
                     lines_modified = 0
                     first_modified = False
                     change = []
-                    for (lineno, line) in enumerate(plain[start_lineno : end_lineno + 1]):
+                    for lineno, line in enumerate(plain[start_lineno : end_lineno + 1]):
                         if line.startswith(self.INDENT):
                             if lineno == 0:
                                 # keep track if first line is modified, so we can properly display
@@ -282,7 +285,7 @@ class CapaExplorerRulgenPreview(QtWidgets.QTextEdit):
                 self.set_selection(select_start_ppos, select_end_ppos, len(self.toPlainText()))
                 self.verticalScrollBar().setSliderPosition(scroll_ppos)
         else:
-            super(CapaExplorerRulgenPreview, self).keyPressEvent(e)
+            super().keyPressEvent(e)
 
     def count_previous_lines_from_block(self, block):
         """calculate number of lines preceding block"""
@@ -302,13 +305,12 @@ class CapaExplorerRulgenPreview(QtWidgets.QTextEdit):
         self.setTextCursor(cursor)
 
 
-class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
-
+class CapaExplorerRulegenEditor(QtWidgets.QTreeWidget):
     updated = QtCore.pyqtSignal()
 
     def __init__(self, preview, parent=None):
         """ """
-        super(CapaExplorerRulgenEditor, self).__init__(parent)
+        super().__init__(parent)
 
         self.preview = preview
 
@@ -372,18 +374,18 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
 
     def dragMoveEvent(self, e):
         """ """
-        super(CapaExplorerRulgenEditor, self).dragMoveEvent(e)
+        super().dragMoveEvent(e)
 
     def dragEventEnter(self, e):
         """ """
-        super(CapaExplorerRulgenEditor, self).dragEventEnter(e)
+        super().dragEventEnter(e)
 
     def dropEvent(self, e):
         """ """
         if not self.indexAt(e.pos()).isValid():
             return
 
-        super(CapaExplorerRulgenEditor, self).dropEvent(e)
+        super().dropEvent(e)
 
         self.update_preview()
         expand_tree(self.invisibleRootItem())
@@ -425,6 +427,10 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
             # add default child expression when nesting under basic block
             new_parent.setExpanded(True)
             new_parent = self.new_expression_node(new_parent, ("- or:", ""))
+        elif "instruction" in action.data()[0]:
+            # add default child expression when nesting under instruction
+            new_parent.setExpanded(True)
+            new_parent = self.new_expression_node(new_parent, ("- or:", ""))
 
         for o in self.get_features(selected=True):
             # take child from its parent by index, add to new parent
@@ -437,7 +443,7 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         """ """
         expression, o = action.data()
         if "basic block" in expression and "basic block" not in o.text(
-            CapaExplorerRulgenEditor.get_column_feature_index()
+            CapaExplorerRulegenEditor.get_column_feature_index()
         ):
             # current expression is "basic block", and not changing to "basic block" expression
             children = o.takeChildren()
@@ -445,7 +451,17 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
             for child in children:
                 new_parent.addChild(child)
             new_parent.setExpanded(True)
-        o.setText(CapaExplorerRulgenEditor.get_column_feature_index(), expression)
+        elif "instruction" in expression and "instruction" not in o.text(
+            CapaExplorerRulegenEditor.get_column_feature_index()
+        ):
+            # current expression is "instruction", and not changing to "instruction" expression
+            children = o.takeChildren()
+            new_parent = self.new_expression_node(o, ("- or:", ""))
+            for child in children:
+                new_parent.addChild(child)
+            new_parent.setExpanded(True)
+
+        o.setText(CapaExplorerRulegenEditor.get_column_feature_index(), expression)
 
     def slot_clear_all(self, action):
         """ """
@@ -456,7 +472,7 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         if not self.indexAt(pos).isValid():
             # user selected invalid index
             self.load_custom_context_menu_invalid_index(pos)
-        elif self.itemAt(pos).capa_type == CapaExplorerRulgenEditor.get_node_type_expression():
+        elif self.itemAt(pos).capa_type == CapaExplorerRulegenEditor.get_node_type_expression():
             # user selected expression node
             self.load_custom_context_menu_expression(pos)
         else:
@@ -468,8 +484,8 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
     def slot_item_double_clicked(self, o, column):
         """ """
         if column in (
-            CapaExplorerRulgenEditor.get_column_comment_index(),
-            CapaExplorerRulgenEditor.get_column_description_index(),
+            CapaExplorerRulegenEditor.get_column_comment_index(),
+            CapaExplorerRulegenEditor.get_column_description_index(),
         ):
             o.setFlags(o.flags() | QtCore.Qt.ItemIsEditable)
             self.editItem(o, column)
@@ -518,6 +534,7 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
             ("not", ("- not:",), self.slot_nest_features),
             ("optional", ("- optional:",), self.slot_nest_features),
             ("basic block", ("- basic block:",), self.slot_nest_features),
+            ("instruction", ("- instruction:",), self.slot_nest_features),
         )
 
         # build submenu with modify actions
@@ -539,6 +556,7 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
             ("not", ("- not:", self.itemAt(pos)), self.slot_edit_expression),
             ("optional", ("- optional:", self.itemAt(pos)), self.slot_edit_expression),
             ("basic block", ("- basic block:", self.itemAt(pos)), self.slot_edit_expression),
+            ("instruction", ("- instruction:", self.itemAt(pos)), self.slot_edit_expression),
         )
 
         # build submenu with modify actions
@@ -555,7 +573,7 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         font = QtGui.QFont()
         font.setBold(True)
 
-        o.setFont(CapaExplorerRulgenEditor.get_column_feature_index(), font)
+        o.setFont(CapaExplorerRulegenEditor.get_column_feature_index(), font)
 
     def style_feature_node(self, o):
         """ """
@@ -566,8 +584,8 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         font.setWeight(QtGui.QFont.Medium)
         brush.setColor(QtGui.QColor(*COLOR_GREEN_RGB))
 
-        o.setFont(CapaExplorerRulgenEditor.get_column_feature_index(), font)
-        o.setForeground(CapaExplorerRulgenEditor.get_column_feature_index(), brush)
+        o.setFont(CapaExplorerRulegenEditor.get_column_feature_index(), font)
+        o.setForeground(CapaExplorerRulegenEditor.get_column_feature_index(), brush)
 
     def style_comment_node(self, o):
         """ """
@@ -575,22 +593,22 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         font.setBold(True)
         font.setFamily("Courier")
 
-        o.setFont(CapaExplorerRulgenEditor.get_column_feature_index(), font)
+        o.setFont(CapaExplorerRulegenEditor.get_column_feature_index(), font)
 
     def set_expression_node(self, o):
         """ """
-        setattr(o, "capa_type", CapaExplorerRulgenEditor.get_node_type_expression())
+        setattr(o, "capa_type", CapaExplorerRulegenEditor.get_node_type_expression())
         self.style_expression_node(o)
 
     def set_feature_node(self, o):
         """ """
-        setattr(o, "capa_type", CapaExplorerRulgenEditor.get_node_type_feature())
+        setattr(o, "capa_type", CapaExplorerRulegenEditor.get_node_type_feature())
         o.setFlags(o.flags() & ~QtCore.Qt.ItemIsDropEnabled)
         self.style_feature_node(o)
 
     def set_comment_node(self, o):
         """ """
-        setattr(o, "capa_type", CapaExplorerRulgenEditor.get_node_type_comment())
+        setattr(o, "capa_type", CapaExplorerRulegenEditor.get_node_type_comment())
         o.setFlags(o.flags() & ~QtCore.Qt.ItemIsDropEnabled)
 
         self.style_comment_node(o)
@@ -599,7 +617,7 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         """ """
         o = QtWidgets.QTreeWidgetItem(parent)
         self.set_expression_node(o)
-        for (i, v) in enumerate(values):
+        for i, v in enumerate(values):
             o.setText(i, v)
         return o
 
@@ -607,7 +625,7 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         """ """
         o = QtWidgets.QTreeWidgetItem(parent)
         self.set_feature_node(o)
-        for (i, v) in enumerate(values):
+        for i, v in enumerate(values):
             o.setText(i, v)
         return o
 
@@ -615,7 +633,7 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         """ """
         o = QtWidgets.QTreeWidgetItem(parent)
         self.set_comment_node(o)
-        for (i, v) in enumerate(values):
+        for i, v in enumerate(values):
             o.setText(i, v)
         return o
 
@@ -634,7 +652,7 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         counted = list(zip(Counter(features).keys(), Counter(features).values()))
 
         # single features
-        for (k, v) in filter(lambda t: t[1] == 1, counted):
+        for k, v in filter(lambda t: t[1] == 1, counted):
             if isinstance(k, (capa.features.common.String,)):
                 value = '"%s"' % capa.features.common.escape_string(k.get_value_str())
             else:
@@ -642,7 +660,7 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
             self.new_feature_node(top_node, ("- %s: %s" % (k.name.lower(), value), ""))
 
         # n > 1 features
-        for (k, v) in filter(lambda t: t[1] > 1, counted):
+        for k, v in filter(lambda t: t[1] > 1, counted):
             if k.value:
                 if isinstance(k, (capa.features.common.String,)):
                     value = '"%s"' % capa.features.common.escape_string(k.get_value_str())
@@ -687,16 +705,16 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         node = QtWidgets.QTreeWidgetItem(parent)
 
         # set node text to data parsed from feature
-        for (idx, text) in enumerate((feature, comment, description)):
+        for idx, text in enumerate((feature, comment, description)):
             node.setText(idx, text)
 
         # we need to set our own type so we can control the GUI accordingly
-        if feature.startswith(("- and:", "- or:", "- not:", "- basic block:", "- optional:")):
-            setattr(node, "capa_type", CapaExplorerRulgenEditor.get_node_type_expression())
+        if feature.startswith(("- and:", "- or:", "- not:", "- basic block:", "- instruction:", "- optional:")):
+            setattr(node, "capa_type", CapaExplorerRulegenEditor.get_node_type_expression())
         elif feature.startswith("#"):
-            setattr(node, "capa_type", CapaExplorerRulgenEditor.get_node_type_comment())
+            setattr(node, "capa_type", CapaExplorerRulegenEditor.get_node_type_comment())
         else:
-            setattr(node, "capa_type", CapaExplorerRulgenEditor.get_node_type_feature())
+            setattr(node, "capa_type", CapaExplorerRulegenEditor.get_node_type_feature())
 
         # format the node based on its type
         (self.set_expression_node, self.set_feature_node, self.set_comment_node)[node.capa_type](node)
@@ -758,7 +776,7 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         """ """
         for feature in filter(
             lambda o: o.capa_type
-            in (CapaExplorerRulgenEditor.get_node_type_feature(), CapaExplorerRulgenEditor.get_node_type_comment()),
+            in (CapaExplorerRulegenEditor.get_node_type_feature(), CapaExplorerRulegenEditor.get_node_type_comment()),
             tuple(iterate_tree(self)),
         ):
             if feature in ignore:
@@ -770,7 +788,7 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
     def get_expressions(self, selected=False, ignore=()):
         """ """
         for expression in filter(
-            lambda o: o.capa_type == CapaExplorerRulgenEditor.get_node_type_expression(), tuple(iterate_tree(self))
+            lambda o: o.capa_type == CapaExplorerRulegenEditor.get_node_type_expression(), tuple(iterate_tree(self))
         ):
             if expression in ignore:
                 continue
@@ -782,12 +800,12 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
 class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
     def __init__(self, editor, parent=None):
         """ """
-        super(CapaExplorerRulegenFeatures, self).__init__(parent)
+        super().__init__(parent)
 
         self.parent_items = {}
         self.editor = editor
 
-        self.setHeaderLabels(["Feature", "Virtual Address"])
+        self.setHeaderLabels(["Feature", "Address"])
         self.setStyleSheet("QTreeView::item {padding-right: 15 px;padding-bottom: 2 px;}")
 
         # configure view columns to auto-resize
@@ -979,7 +997,7 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
         o = QtWidgets.QTreeWidgetItem(parent)
 
         self.set_parent_node(o)
-        for (i, v) in enumerate(data):
+        for i, v in enumerate(data):
             o.setText(i, v)
         if feature:
             o.setData(0, 0x100, feature)
@@ -991,7 +1009,7 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
         o = QtWidgets.QTreeWidgetItem(parent)
 
         self.set_leaf_node(o)
-        for (i, v) in enumerate(data):
+        for i, v in enumerate(data):
             o.setText(i, v)
         if feature:
             o.setData(0, 0x100, feature)
@@ -1010,7 +1028,10 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
         self.parent_items = {}
 
         def format_address(e):
-            return "%X" % e if e else ""
+            if isinstance(e, AbsoluteVirtualAddress):
+                return "%X" % int(e)
+            else:
+                return ""
 
         def format_feature(feature):
             """ """
@@ -1020,7 +1041,7 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
                 value = '"%s"' % capa.features.common.escape_string(value)
             return "%s(%s)" % (name, value)
 
-        for (feature, eas) in sorted(features.items(), key=lambda k: sorted(k[1])):
+        for feature, addrs in sorted(features.items(), key=lambda k: sorted(k[1])):
             if isinstance(feature, capa.features.basicblock.BasicBlock):
                 # filter basic blocks for now, we may want to add these back in some time
                 # in the future
@@ -1032,7 +1053,7 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
 
             # level 1
             if feature not in self.parent_items:
-                if len(eas) > 1:
+                if len(addrs) > 1:
                     self.parent_items[feature] = self.new_parent_node(
                         self.parent_items[type(feature)], (format_feature(feature),), feature=feature
                     )
@@ -1042,18 +1063,18 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
                     )
 
             # level n > 1
-            if len(eas) > 1:
-                for ea in sorted(eas):
+            if len(addrs) > 1:
+                for addr in sorted(addrs):
                     self.new_leaf_node(
-                        self.parent_items[feature], (format_feature(feature), format_address(ea)), feature=feature
+                        self.parent_items[feature], (format_feature(feature), format_address(addr)), feature=feature
                     )
             else:
-                if eas:
-                    ea = eas.pop()
+                if addrs:
+                    addr = addrs.pop()
                 else:
                     # some features may not have an address e.g. "format"
-                    ea = ""
-                for (i, v) in enumerate((format_feature(feature), format_address(ea))):
+                    addr = _NoAddress()
+                for i, v in enumerate((format_feature(feature), format_address(addr))):
                     self.parent_items[feature].setText(i, v)
                 self.parent_items[feature].setData(0, 0x100, feature)
 
@@ -1069,7 +1090,7 @@ class CapaExplorerQtreeView(QtWidgets.QTreeView):
 
     def __init__(self, model, parent=None):
         """initialize view"""
-        super(CapaExplorerQtreeView, self).__init__(parent)
+        super().__init__(parent)
 
         self.setModel(model)
 

capa/optimizer.py

@@ -0,0 +1,70 @@
+import logging
+
+import capa.engine as ceng
+import capa.features.common
+
+logger = logging.getLogger(__name__)
+
+
+def get_node_cost(node):
+    if isinstance(node, (capa.features.common.OS, capa.features.common.Arch, capa.features.common.Format)):
+        # we assume these are the most restrictive features:
+        # authors commonly use them at the start of rules to restrict the category of samples to inspect
+        return 0
+
+    # elif "everything else":
+    #   return 1
+    #
+    # this should be all hash-lookup features.
+    # see below.
+
+    elif isinstance(node, (capa.features.common.Substring, capa.features.common.Regex, capa.features.common.Bytes)):
+        # substring and regex features require a full scan of each string
+        # which we anticipate is more expensive then a hash lookup feature (e.g. mnemonic or count).
+        #
+        # TODO: compute the average cost of these feature relative to hash feature
+        # and adjust the factor accordingly.
+        return 2
+
+    elif isinstance(node, (ceng.Not, ceng.Range)):
+        # the cost of these nodes are defined by the complexity of their single child.
+        return 1 + get_node_cost(node.child)
+
+    elif isinstance(node, (ceng.And, ceng.Or, ceng.Some)):
+        # the cost of these nodes is the full cost of their children
+        # as this is the worst-case scenario.
+        return 1 + sum(map(get_node_cost, node.children))
+
+    else:
+        # this should be all hash-lookup features.
+        # we give this a arbitrary weight of 1.
+        # the only thing more "important" than this is checking OS/Arch/Format.
+        return 1
+
+
+def optimize_statement(statement):
+    # this routine operates in-place
+
+    if isinstance(statement, (ceng.And, ceng.Or, ceng.Some)):
+        # has .children
+        statement.children = sorted(statement.children, key=get_node_cost)
+        return
+    elif isinstance(statement, (ceng.Not, ceng.Range)):
+        # has .child
+        optimize_statement(statement.child)
+        return
+    else:
+        # appears to be "simple"
+        return
+
+
+def optimize_rule(rule):
+    # this routine operates in-place
+    optimize_statement(rule.statement)
+
+
+def optimize_rules(rules):
+    logger.debug("optimizing %d rules", len(rules))
+    for rule in rules:
+        optimize_rule(rule)
+    return rules

capa/perf.py

@@ -0,0 +1,10 @@
+import typing
+import collections
+
+# this structure is unstable and may change before the next major release.
+counters: typing.Counter[str] = collections.Counter()
+
+
+def reset():
+    global counters
+    counters = collections.Counter()

capa/render/default.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -11,7 +11,9 @@ import collections
 import tabulate
 
 import capa.render.utils as rutils
-import capa.render.result_document
+import capa.features.freeze as frz
+import capa.render.result_document as rd
+import capa.features.freeze.features as frzf
 from capa.rules import RuleSet
 from capa.engine import MatchResults
 from capa.render.utils import StringIO
@@ -27,50 +29,49 @@ def width(s: str, character_count: int) -> str:
         return s
 
 
-def render_meta(doc, ostream: StringIO):
+def render_meta(doc: rd.ResultDocument, ostream: StringIO):
     rows = [
-        (width("md5", 22), width(doc["meta"]["sample"]["md5"], 82)),
-        ("sha1", doc["meta"]["sample"]["sha1"]),
-        ("sha256", doc["meta"]["sample"]["sha256"]),
-        ("os", doc["meta"]["analysis"]["os"]),
-        ("format", doc["meta"]["analysis"]["format"]),
-        ("arch", doc["meta"]["analysis"]["arch"]),
-        ("path", doc["meta"]["sample"]["path"]),
+        (width("md5", 22), width(doc.meta.sample.md5, 82)),
+        ("sha1", doc.meta.sample.sha1),
+        ("sha256", doc.meta.sample.sha256),
+        ("os", doc.meta.analysis.os),
+        ("format", doc.meta.analysis.format),
+        ("arch", doc.meta.analysis.arch),
+        ("path", doc.meta.sample.path),
     ]
 
     ostream.write(tabulate.tabulate(rows, tablefmt="psql"))
     ostream.write("\n")
 
 
-def find_subrule_matches(doc):
+def find_subrule_matches(doc: rd.ResultDocument):
     """
     collect the rule names that have been matched as a subrule match.
     this way we can avoid displaying entries for things that are too specific.
     """
     matches = set([])
 
-    def rec(node):
-        if not node["success"]:
+    def rec(match: rd.Match):
+        if not match.success:
             # there's probably a bug here for rules that do `not: match: ...`
             # but we don't have any examples of this yet
             return
 
-        elif node["node"]["type"] == "statement":
-            for child in node["children"]:
+        elif isinstance(match.node, rd.StatementNode):
+            for child in match.children:
                 rec(child)
 
-        elif node["node"]["type"] == "feature":
-            if node["node"]["feature"]["type"] == "match":
-                matches.add(node["node"]["feature"]["match"])
+        elif isinstance(match.node, rd.FeatureNode) and isinstance(match.node.feature, frzf.MatchFeature):
+            matches.add(match.node.feature.match)
 
     for rule in rutils.capability_rules(doc):
-        for node in rule["matches"].values():
-            rec(node)
+        for address, match in rule.matches:
+            rec(match)
 
     return matches
 
 
-def render_capabilities(doc, ostream: StringIO):
+def render_capabilities(doc: rd.ResultDocument, ostream: StringIO):
     """
     example::
 
@@ -86,18 +87,18 @@ def render_capabilities(doc, ostream: StringIO):
 
     rows = []
     for rule in rutils.capability_rules(doc):
-        if rule["meta"]["name"] in subrule_matches:
+        if rule.meta.name in subrule_matches:
             # rules that are also matched by other rules should not get rendered by default.
             # this cuts down on the amount of output while giving approx the same detail.
             # see #224
             continue
 
-        count = len(rule["matches"])
+        count = len(rule.matches)
         if count == 1:
-            capability = rutils.bold(rule["meta"]["name"])
+            capability = rutils.bold(rule.meta.name)
         else:
-            capability = "%s (%d matches)" % (rutils.bold(rule["meta"]["name"]), count)
-        rows.append((capability, rule["meta"]["namespace"]))
+            capability = "%s (%d matches)" % (rutils.bold(rule.meta.name), count)
+        rows.append((capability, rule.meta.namespace))
 
     if rows:
         ostream.write(
@@ -108,7 +109,7 @@ def render_capabilities(doc, ostream: StringIO):
         ostream.writeln(rutils.bold("no capabilities found"))
 
 
-def render_attack(doc, ostream: StringIO):
+def render_attack(doc: rd.ResultDocument, ostream: StringIO):
     """
     example::
 
@@ -126,17 +127,14 @@ def render_attack(doc, ostream: StringIO):
     """
     tactics = collections.defaultdict(set)
     for rule in rutils.capability_rules(doc):
-        if not rule["meta"].get("att&ck"):
-            continue
-
-        for attack in rule["meta"]["att&ck"]:
-            tactics[attack["tactic"]].add((attack["technique"], attack.get("subtechnique"), attack["id"]))
+        for attack in rule.meta.attack:
+            tactics[attack.tactic].add((attack.technique, attack.subtechnique, attack.id))
 
     rows = []
     for tactic, techniques in sorted(tactics.items()):
         inner_rows = []
-        for (technique, subtechnique, id) in sorted(techniques):
-            if subtechnique is None:
+        for technique, subtechnique, id in sorted(techniques):
+            if not subtechnique:
                 inner_rows.append("%s %s" % (rutils.bold(technique), id))
             else:
                 inner_rows.append("%s::%s %s" % (rutils.bold(technique), subtechnique, id))
@@ -156,7 +154,7 @@ def render_attack(doc, ostream: StringIO):
         ostream.write("\n")
 
 
-def render_mbc(doc, ostream: StringIO):
+def render_mbc(doc: rd.ResultDocument, ostream: StringIO):
     """
     example::
 
@@ -172,17 +170,14 @@ def render_mbc(doc, ostream: StringIO):
     """
     objectives = collections.defaultdict(set)
     for rule in rutils.capability_rules(doc):
-        if not rule["meta"].get("mbc"):
-            continue
-
-        for mbc in rule["meta"]["mbc"]:
-            objectives[mbc["objective"]].add((mbc["behavior"], mbc.get("method"), mbc["id"]))
+        for mbc in rule.meta.mbc:
+            objectives[mbc.objective].add((mbc.behavior, mbc.method, mbc.id))
 
     rows = []
     for objective, behaviors in sorted(objectives.items()):
         inner_rows = []
-        for (behavior, method, id) in sorted(behaviors):
-            if method is None:
+        for behavior, method, id in sorted(behaviors):
+            if not method:
                 inner_rows.append("%s [%s]" % (rutils.bold(behavior), id))
             else:
                 inner_rows.append("%s::%s [%s]" % (rutils.bold(behavior), method, id))
@@ -200,7 +195,7 @@ def render_mbc(doc, ostream: StringIO):
         ostream.write("\n")
 
 
-def render_default(doc):
+def render_default(doc: rd.ResultDocument):
     ostream = rutils.StringIO()
 
     render_meta(doc, ostream)
@@ -215,5 +210,5 @@ def render_default(doc):
 
 
 def render(meta, rules: RuleSet, capabilities: MatchResults) -> str:
-    doc = capa.render.result_document.convert_capabilities_to_result_document(meta, rules, capabilities)
+    doc = rd.ResultDocument.from_capa(meta, rules, capabilities)
     return render_default(doc)

capa/render/json.py

@@ -1,33 +1,14 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
-import json
-
-import capa.render.result_document
+import capa.render.result_document as rd
 from capa.rules import RuleSet
 from capa.engine import MatchResults
 
 
-class CapaJsonObjectEncoder(json.JSONEncoder):
-    """JSON encoder that emits Python sets as sorted lists"""
-
-    def default(self, obj):
-        if isinstance(obj, (list, dict, int, float, bool, type(None))) or isinstance(obj, str):
-            return json.JSONEncoder.default(self, obj)
-        elif isinstance(obj, set):
-            return list(sorted(obj))
-        else:
-            # probably will TypeError
-            return json.JSONEncoder.default(self, obj)
-
-
 def render(meta, rules: RuleSet, capabilities: MatchResults) -> str:
-    return json.dumps(
-        capa.render.result_document.convert_capabilities_to_result_document(meta, rules, capabilities),
-        cls=CapaJsonObjectEncoder,
-        sort_keys=True,
-    )
+    return rd.ResultDocument.from_capa(meta, rules, capabilities).json(exclude_none=True)

capa/render/result_document.py

@@ -1,329 +1,542 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
 # Unless required by applicable law or agreed to in writing, software distributed under the License
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
-import copy
+import datetime
+from typing import Any, Dict, Tuple, Union, Optional
+
+from pydantic import Field, BaseModel
 
 import capa.rules
 import capa.engine
-import capa.render.utils
 import capa.features.common
+import capa.features.freeze as frz
+import capa.features.address
+import capa.features.freeze.features as frzf
 from capa.rules import RuleSet
 from capa.engine import MatchResults
+from capa.helpers import assert_never
 
 
-def convert_statement_to_result_document(statement):
-    """
-    "statement": {
-        "type": "or"
-    },
-
-    "statement": {
-        "max": 9223372036854775808,
-        "min": 2,
-        "type": "range"
-    },
-    """
-    statement_type = statement.name.lower()
-    result = {"type": statement_type}
-    if statement.description:
-        result["description"] = statement.description
-
-    if statement_type == "some" and statement.count == 0:
-        result["type"] = "optional"
-    elif statement_type == "some":
-        result["count"] = statement.count
-    elif statement_type == "range":
-        result["min"] = statement.min
-        result["max"] = statement.max
-        result["child"] = convert_feature_to_result_document(statement.child)
-    elif statement_type == "subscope":
-        result["subscope"] = statement.scope
-
-    return result
+class FrozenModel(BaseModel):
+    class Config:
+        frozen = True
 
 
-def convert_feature_to_result_document(feature):
-    """
-    "feature": {
-        "number": 6,
-        "type": "number"
-    },
-
-    "feature": {
-        "api": "ws2_32.WSASocket",
-        "type": "api"
-    },
-
-    "feature": {
-        "match": "create TCP socket",
-        "type": "match"
-    },
-
-    "feature": {
-        "characteristic": [
-            "loop",
-            true
-        ],
-        "type": "characteristic"
-    },
-    """
-    result = {"type": feature.name, feature.name: feature.get_value_str()}
-    if feature.description:
-        result["description"] = feature.description
-    if feature.name in ("regex", "substring"):
-        result["matches"] = feature.matches
-    return result
+class Sample(FrozenModel):
+    md5: str
+    sha1: str
+    sha256: str
+    path: str
 
 
-def convert_node_to_result_document(node):
-    """
-    "node": {
-        "type": "statement",
-        "statement": { ... }
-    },
+class BasicBlockLayout(FrozenModel):
+    address: frz.Address
 
-    "node": {
-        "type": "feature",
-        "feature": { ... }
-    },
-    """
 
-    if isinstance(node, capa.engine.Statement):
-        return {
-            "type": "statement",
-            "statement": convert_statement_to_result_document(node),
-        }
-    elif isinstance(node, capa.features.common.Feature):
-        return {
-            "type": "feature",
-            "feature": convert_feature_to_result_document(node),
-        }
+class FunctionLayout(FrozenModel):
+    address: frz.Address
+    matched_basic_blocks: Tuple[BasicBlockLayout, ...]
+
+
+class Layout(FrozenModel):
+    functions: Tuple[FunctionLayout, ...]
+
+
+class LibraryFunction(FrozenModel):
+    address: frz.Address
+    name: str
+
+
+class FunctionFeatureCount(FrozenModel):
+    address: frz.Address
+    count: int
+
+
+class FeatureCounts(FrozenModel):
+    file: int
+    functions: Tuple[FunctionFeatureCount, ...]
+
+
+class Analysis(FrozenModel):
+    format: str
+    arch: str
+    os: str
+    extractor: str
+    rules: Tuple[str, ...]
+    base_address: frz.Address
+    layout: Layout
+    feature_counts: FeatureCounts
+    library_functions: Tuple[LibraryFunction, ...]
+
+
+class Metadata(FrozenModel):
+    timestamp: datetime.datetime
+    version: str
+    argv: Optional[Tuple[str, ...]]
+    sample: Sample
+    analysis: Analysis
+
+    @classmethod
+    def from_capa(cls, meta: Any) -> "Metadata":
+        return cls(
+            timestamp=meta["timestamp"],
+            version=meta["version"],
+            argv=meta["argv"] if "argv" in meta else None,
+            sample=Sample(
+                md5=meta["sample"]["md5"],
+                sha1=meta["sample"]["sha1"],
+                sha256=meta["sample"]["sha256"],
+                path=meta["sample"]["path"],
+            ),
+            analysis=Analysis(
+                format=meta["analysis"]["format"],
+                arch=meta["analysis"]["arch"],
+                os=meta["analysis"]["os"],
+                extractor=meta["analysis"]["extractor"],
+                rules=meta["analysis"]["rules"],
+                base_address=frz.Address.from_capa(meta["analysis"]["base_address"]),
+                layout=Layout(
+                    functions=tuple(
+                        FunctionLayout(
+                            address=frz.Address.from_capa(address),
+                            matched_basic_blocks=tuple(
+                                BasicBlockLayout(address=frz.Address.from_capa(bb)) for bb in f["matched_basic_blocks"]
+                            ),
+                        )
+                        for address, f in meta["analysis"]["layout"]["functions"].items()
+                    )
+                ),
+                feature_counts=FeatureCounts(
+                    file=meta["analysis"]["feature_counts"]["file"],
+                    functions=tuple(
+                        FunctionFeatureCount(address=frz.Address.from_capa(address), count=count)
+                        for address, count in meta["analysis"]["feature_counts"]["functions"].items()
+                    ),
+                ),
+                library_functions=tuple(
+                    LibraryFunction(address=frz.Address.from_capa(address), name=name)
+                    for address, name in meta["analysis"]["library_functions"].items()
+                ),
+            ),
+        )
+
+
+class CompoundStatementType:
+    AND = "and"
+    OR = "or"
+    NOT = "not"
+    OPTIONAL = "optional"
+
+
+class StatementModel(FrozenModel):
+    ...
+
+
+class CompoundStatement(StatementModel):
+    type: str
+    description: Optional[str] = None
+
+
+class SomeStatement(StatementModel):
+    type = "some"
+    description: Optional[str] = None
+    count: int
+
+
+class RangeStatement(StatementModel):
+    type = "range"
+    description: Optional[str] = None
+    min: int
+    max: int
+    child: frz.Feature
+
+
+class SubscopeStatement(StatementModel):
+    type = "subscope"
+    description: Optional[str] = None
+    scope: capa.rules.Scope
+
+
+Statement = Union[
+    # Note! order matters, see #1161
+    RangeStatement,
+    SomeStatement,
+    SubscopeStatement,
+    CompoundStatement,
+]
+
+
+class StatementNode(FrozenModel):
+    type = "statement"
+    statement: Statement
+
+
+def statement_from_capa(node: capa.engine.Statement) -> Statement:
+    if isinstance(node, (capa.engine.And, capa.engine.Or, capa.engine.Not)):
+        return CompoundStatement(type=node.__class__.__name__.lower(), description=node.description)
+
+    elif isinstance(node, capa.engine.Some):
+        if node.count == 0:
+            return CompoundStatement(type=CompoundStatementType.OPTIONAL, description=node.description)
+
+        else:
+            return SomeStatement(
+                description=node.description,
+                count=node.count,
+            )
+
+    elif isinstance(node, capa.engine.Range):
+        return RangeStatement(
+            description=node.description,
+            min=node.min,
+            max=node.max,
+            child=frz.feature_from_capa(node.child),
+        )
+
+    elif isinstance(node, capa.engine.Subscope):
+        return SubscopeStatement(
+            description=node.description,
+            scope=capa.rules.Scope(node.scope),
+        )
+
     else:
-        raise RuntimeError("unexpected match node type")
+        raise NotImplementedError(f"statement_from_capa({type(node)}) not implemented")
 
 
-def convert_match_to_result_document(rules, capabilities, result):
+class FeatureNode(FrozenModel):
+    type = "feature"
+    feature: frz.Feature
+
+
+Node = Union[StatementNode, FeatureNode]
+
+
+def node_from_capa(node: Union[capa.engine.Statement, capa.engine.Feature]) -> Node:
+    if isinstance(node, capa.engine.Statement):
+        return StatementNode(statement=statement_from_capa(node))
+
+    elif isinstance(node, capa.engine.Feature):
+        return FeatureNode(feature=frz.feature_from_capa(node))
+
+    else:
+        assert_never(node)
+
+
+class Match(BaseModel):
     """
-    convert the given Result instance into a common, Python-native data structure.
-    this will become part of the "result document" format that can be emitted to JSON.
+    args:
+      success: did the node match?
+      node: the logic node or feature node.
+      children: any children of the logic node. not relevent for features, can be empty.
+      locations: where the feature matched. not relevant for logic nodes (except range), can be empty.
+      captures: captured values from the string/regex feature, and the locations of those values.
     """
-    doc = {
-        "success": bool(result.success),
-        "node": convert_node_to_result_document(result.statement),
-        "children": [convert_match_to_result_document(rules, capabilities, child) for child in result.children],
-    }
 
-    # logic expression, like `and`, don't have locations - their children do.
-    # so only add `locations` to feature nodes.
-    if isinstance(result.statement, capa.features.common.Feature):
-        if bool(result.success):
-            doc["locations"] = result.locations
-    elif isinstance(result.statement, capa.engine.Range):
-        if bool(result.success):
-            doc["locations"] = result.locations
+    success: bool
+    node: Node
+    children: Tuple["Match", ...]
+    locations: Tuple[frz.Address, ...]
+    captures: Dict[str, Tuple[frz.Address, ...]]
 
-    # if we have a `match` statement, then we're referencing another rule or namespace.
-    # this could an external rule (written by a human), or
-    #  rule generated to support a subscope (basic block, etc.)
-    # we still want to include the matching logic in this tree.
-    #
-    # so, we need to lookup the other rule results
-    # and then filter those down to the address used here.
-    # finally, splice that logic into this tree.
-    if (
-        doc["node"]["type"] == "feature"
-        and doc["node"]["feature"]["type"] == "match"
-        # only add subtree on success,
-        # because there won't be results for the other rule on failure.
-        and doc["success"]
-    ):
+    @classmethod
+    def from_capa(
+        cls,
+        rules: RuleSet,
+        capabilities: MatchResults,
+        result: capa.engine.Result,
+    ) -> "Match":
+        success = bool(result)
 
-        name = doc["node"]["feature"]["match"]
+        node = node_from_capa(result.statement)
+        children = [Match.from_capa(rules, capabilities, child) for child in result.children]
 
-        if name in rules:
-            # this is a rule that we're matching
-            #
-            # pull matches from the referenced rule into our tree here.
-            rule_name = doc["node"]["feature"]["match"]
+        # logic expression, like `and`, don't have locations - their children do.
+        # so only add `locations` to feature nodes.
+        locations = []
+        if isinstance(node, FeatureNode) and success:
+            locations = list(map(frz.Address.from_capa, result.locations))
+        elif isinstance(node, StatementNode) and isinstance(node.statement, RangeStatement) and success:
+            locations = list(map(frz.Address.from_capa, result.locations))
+
+        captures = {}
+        if isinstance(result.statement, (capa.features.common._MatchedSubstring, capa.features.common._MatchedRegex)):
+            captures = {
+                capture: list(map(frz.Address.from_capa, locs)) for capture, locs in result.statement.matches.items()
+            }
+
+        # if we have a `match` statement, then we're referencing another rule or namespace.
+        # this could an external rule (written by a human), or
+        #  rule generated to support a subscope (basic block, etc.)
+        # we still want to include the matching logic in this tree.
+        #
+        # so, we need to lookup the other rule results
+        # and then filter those down to the address used here.
+        # finally, splice that logic into this tree.
+        if (
+            isinstance(node, FeatureNode)
+            and isinstance(node.feature, frzf.MatchFeature)
+            # only add subtree on success,
+            # because there won't be results for the other rule on failure.
+            and success
+        ):
+            name = node.feature.match
+
+            if name in rules:
+                # this is a rule that we're matching
+                #
+                # pull matches from the referenced rule into our tree here.
+                rule_name = name
+                rule = rules[rule_name]
+                rule_matches = {address: result for (address, result) in capabilities[rule_name]}
+
+                if rule.is_subscope_rule():
+                    # for a subscope rule, fixup the node to be a scope node, rather than a match feature node.
+                    #
+                    # e.g. `contain loop/30c4c78e29bf4d54894fc74f664c62e8` -> `basic block`
+                    #
+                    # note! replace `node`
+                    node = StatementNode(
+                        statement=SubscopeStatement(
+                            scope=rule.meta["scope"],
+                        )
+                    )
+
+                for location in result.locations:
+                    children.append(Match.from_capa(rules, capabilities, rule_matches[location]))
+            else:
+                # this is a namespace that we're matching
+                #
+                # check for all rules in the namespace,
+                # seeing if they matched.
+                # if so, pull their matches into our match tree here.
+                ns_name = name
+                ns_rules = rules.rules_by_namespace[ns_name]
+
+                for rule in ns_rules:
+                    if rule.name in capabilities:
+                        # the rule matched, so splice results into our tree here.
+                        #
+                        # note, there's a shortcoming in our result document schema here:
+                        # we lose the name of the rule that matched in a namespace.
+                        # for example, if we have a statement: `match: runtime/dotnet`
+                        # and we get matches, we can say the following:
+                        #
+                        #     match: runtime/dotnet @ 0x0
+                        #       or:
+                        #         import: mscoree._CorExeMain @ 0x402000
+                        #
+                        # however, we lose the fact that it was rule
+                        #   "compiled to the .NET platform"
+                        # that contained this logic and did the match.
+                        #
+                        # we could introduce an intermediate node here.
+                        # this would be a breaking change and require updates to the renderers.
+                        # in the meantime, the above might be sufficient.
+                        rule_matches = {address: result for (address, result) in capabilities[rule.name]}
+                        for location in result.locations:
+                            # doc[locations] contains all matches for the given namespace.
+                            # for example, the feature might be `match: anti-analysis/packer`
+                            # which matches against "generic unpacker" and "UPX".
+                            # in this case, doc[locations] contains locations for *both* of thse.
+                            #
+                            # rule_matches contains the matches for the specific rule.
+                            # this is a subset of doc[locations].
+                            #
+                            # so, grab only the locations for current rule.
+                            if location in rule_matches:
+                                children.append(Match.from_capa(rules, capabilities, rule_matches[location]))
+
+        return cls(
+            success=success,
+            node=node,
+            children=children,
+            locations=locations,
+            captures=captures,
+        )
+
+
+def parse_parts_id(s: str):
+    id_ = ""
+    parts = s.split("::")
+    if len(parts) > 0:
+        last = parts.pop()
+        last, _, id_ = last.rpartition(" ")
+        id_ = id_.lstrip("[").rstrip("]")
+        parts.append(last)
+    return tuple(parts), id_
+
+
+class AttackSpec(FrozenModel):
+    """
+    given an ATT&CK spec like: `Tactic::Technique::Subtechnique [Identifier]`
+    e.g., `Execution::Command and Scripting Interpreter::Python [T1059.006]`
+
+    args:
+      tactic: like `Tactic` above, perhaps "Execution"
+      technique: like `Technique` above, perhaps "Command and Scripting Interpreter"
+      subtechnique: like `Subtechnique` above, perhaps "Python"
+      id: like `Identifier` above, perhaps "T1059.006"
+    """
+
+    parts: Tuple[str, ...]
+    tactic: str
+    technique: str
+    subtechnique: str
+    id: str
+
+    @classmethod
+    def from_str(cls, s) -> "AttackSpec":
+        tactic = ""
+        technique = ""
+        subtechnique = ""
+        parts, id_ = parse_parts_id(s)
+        if len(parts) > 0:
+            tactic = parts[0]
+        if len(parts) > 1:
+            technique = parts[1]
+        if len(parts) > 2:
+            subtechnique = parts[2]
+
+        return cls(
+            parts=parts,
+            tactic=tactic,
+            technique=technique,
+            subtechnique=subtechnique,
+            id=id_,
+        )
+
+
+class MBCSpec(FrozenModel):
+    """
+    given an MBC spec like: `Objective::Behavior::Method [Identifier]`
+    e.g., `Collection::Input Capture::Mouse Events [E1056.m01]`
+
+    args:
+      objective: like `Objective` above, perhaps "Collection"
+      behavior: like `Behavior` above, perhaps "Input Capture"
+      method: like `Method` above, perhaps "Mouse Events"
+      id: like `Identifier` above, perhaps "E1056.m01"
+    """
+
+    parts: Tuple[str, ...]
+    objective: str
+    behavior: str
+    method: str
+    id: str
+
+    @classmethod
+    def from_str(cls, s) -> "MBCSpec":
+        objective = ""
+        behavior = ""
+        method = ""
+        parts, id_ = parse_parts_id(s)
+        if len(parts) > 0:
+            objective = parts[0]
+        if len(parts) > 1:
+            behavior = parts[1]
+        if len(parts) > 2:
+            method = parts[2]
+
+        return cls(
+            parts=parts,
+            objective=objective,
+            behavior=behavior,
+            method=method,
+            id=id_,
+        )
+
+
+class MaecMetadata(FrozenModel):
+    analysis_conclusion: Optional[str] = Field(None, alias="analysis-conclusion")
+    analysis_conclusion_ov: Optional[str] = Field(None, alias="analysis-conclusion-ov")
+    malware_family: Optional[str] = Field(None, alias="malware-family")
+    malware_category: Optional[str] = Field(None, alias="malware-category")
+    malware_category_ov: Optional[str] = Field(None, alias="malware-category-ov")
+
+    class Config:
+        frozen = True
+        allow_population_by_field_name = True
+
+
+class RuleMetadata(FrozenModel):
+    name: str
+    namespace: Optional[str]
+    authors: Tuple[str, ...]
+    scope: capa.rules.Scope
+    attack: Tuple[AttackSpec, ...] = Field(alias="att&ck")
+    mbc: Tuple[MBCSpec, ...]
+    references: Tuple[str, ...]
+    examples: Tuple[str, ...]
+    description: str
+
+    lib: bool = Field(False, alias="lib")
+    is_subscope_rule: bool = Field(False, alias="capa/subscope")
+    maec: MaecMetadata
+
+    @classmethod
+    def from_capa(cls, rule: capa.rules.Rule) -> "RuleMetadata":
+        return cls(
+            name=rule.meta.get("name"),
+            namespace=rule.meta.get("namespace"),
+            authors=rule.meta.get("authors"),
+            scope=capa.rules.Scope(rule.meta.get("scope")),
+            attack=list(map(AttackSpec.from_str, rule.meta.get("att&ck", []))),
+            mbc=list(map(MBCSpec.from_str, rule.meta.get("mbc", []))),
+            references=rule.meta.get("references", []),
+            examples=rule.meta.get("examples", []),
+            description=rule.meta.get("description", ""),
+            lib=rule.meta.get("lib", False),
+            capa_subscope=rule.meta.get("capa/subscope", False),
+            maec=MaecMetadata(
+                analysis_conclusion=rule.meta.get("maec/analysis-conclusion"),
+                analysis_conclusion_ov=rule.meta.get("maec/analysis-conclusion-ov"),
+                malware_family=rule.meta.get("maec/malware-family"),
+                malware_category=rule.meta.get("maec/malware-category"),
+                malware_category_ov=rule.meta.get("maec/malware-category-ov"),
+            ),
+        )
+
+    class Config:
+        frozen = True
+        allow_population_by_field_name = True
+
+
+class RuleMatches(BaseModel):
+    """
+    args:
+        meta: the metadata from the rule
+        source: the raw rule text
+    """
+
+    meta: RuleMetadata
+    source: str
+    matches: Tuple[Tuple[frz.Address, Match], ...]
+
+
+class ResultDocument(BaseModel):
+    meta: Metadata
+    rules: Dict[str, RuleMatches]
+
+    @classmethod
+    def from_capa(cls, meta, rules: RuleSet, capabilities: MatchResults) -> "ResultDocument":
+        rule_matches: Dict[str, RuleMatches] = {}
+        for rule_name, matches in capabilities.items():
             rule = rules[rule_name]
-            rule_matches = {address: result for (address, result) in capabilities[rule_name]}
 
             if rule.meta.get("capa/subscope-rule"):
-                # for a subscope rule, fixup the node to be a scope node, rather than a match feature node.
-                #
-                # e.g. `contain loop/30c4c78e29bf4d54894fc74f664c62e8` -> `basic block`
-                scope = rule.meta["scope"]
-                doc["node"] = {
-                    "type": "statement",
-                    "statement": {
-                        "type": "subscope",
-                        "subscope": scope,
-                    },
-                }
+                continue
 
-            for location in doc["locations"]:
-                doc["children"].append(convert_match_to_result_document(rules, capabilities, rule_matches[location]))
-        else:
-            # this is a namespace that we're matching
-            #
-            # check for all rules in the namespace,
-            # seeing if they matched.
-            # if so, pull their matches into our match tree here.
-            ns_name = doc["node"]["feature"]["match"]
-            ns_rules = rules.rules_by_namespace[ns_name]
+            rule_matches[rule_name] = RuleMatches(
+                meta=RuleMetadata.from_capa(rule),
+                source=rule.definition,
+                matches=tuple(
+                    (frz.Address.from_capa(addr), Match.from_capa(rules, capabilities, match))
+                    for addr, match in matches
+                ),
+            )
 
-            for rule in ns_rules:
-                if rule.name in capabilities:
-                    # the rule matched, so splice results into our tree here.
-                    #
-                    # note, there's a shortcoming in our result document schema here:
-                    # we lose the name of the rule that matched in a namespace.
-                    # for example, if we have a statement: `match: runtime/dotnet`
-                    # and we get matches, we can say the following:
-                    #
-                    #     match: runtime/dotnet @ 0x0
-                    #       or:
-                    #         import: mscoree._CorExeMain @ 0x402000
-                    #
-                    # however, we lose the fact that it was rule
-                    #   "compiled to the .NET platform"
-                    # that contained this logic and did the match.
-                    #
-                    # we could introduce an intermediate node here.
-                    # this would be a breaking change and require updates to the renderers.
-                    # in the meantime, the above might be sufficient.
-                    rule_matches = {address: result for (address, result) in capabilities[rule.name]}
-                    for location in doc["locations"]:
-                        # doc[locations] contains all matches for the given namespace.
-                        # for example, the feature might be `match: anti-analysis/packer`
-                        # which matches against "generic unpacker" and "UPX".
-                        # in this case, doc[locations] contains locations for *both* of thse.
-                        #
-                        # rule_matches contains the matches for the specific rule.
-                        # this is a subset of doc[locations].
-                        #
-                        # so, grab only the locations for current rule.
-                        if location in rule_matches:
-                            doc["children"].append(
-                                convert_match_to_result_document(rules, capabilities, rule_matches[location])
-                            )
-
-    return doc
-
-
-def convert_meta_to_result_document(meta):
-    # make a copy so that we don't modify the given parameter
-    meta = copy.deepcopy(meta)
-
-    attacks = meta.get("att&ck", [])
-    meta["att&ck"] = [parse_canonical_attack(attack) for attack in attacks]
-    mbcs = meta.get("mbc", [])
-    meta["mbc"] = [parse_canonical_mbc(mbc) for mbc in mbcs]
-    return meta
-
-
-def parse_canonical_attack(attack: str):
-    """
-    parse capa's canonical ATT&CK representation: `Tactic::Technique::Subtechnique [Identifier]`
-    """
-    tactic = ""
-    technique = ""
-    subtechnique = ""
-    parts, id = capa.render.utils.parse_parts_id(attack)
-    if len(parts) > 0:
-        tactic = parts[0]
-    if len(parts) > 1:
-        technique = parts[1]
-    if len(parts) > 2:
-        subtechnique = parts[2]
-
-    return {
-        "parts": parts,
-        "id": id,
-        "tactic": tactic,
-        "technique": technique,
-        "subtechnique": subtechnique,
-    }
-
-
-def parse_canonical_mbc(mbc: str):
-    """
-    parse capa's canonical MBC representation: `Objective::Behavior::Method [Identifier]`
-    """
-    objective = ""
-    behavior = ""
-    method = ""
-    parts, id = capa.render.utils.parse_parts_id(mbc)
-    if len(parts) > 0:
-        objective = parts[0]
-    if len(parts) > 1:
-        behavior = parts[1]
-    if len(parts) > 2:
-        method = parts[2]
-
-    return {
-        "parts": parts,
-        "id": id,
-        "objective": objective,
-        "behavior": behavior,
-        "method": method,
-    }
-
-
-def convert_capabilities_to_result_document(meta, rules: RuleSet, capabilities: MatchResults):
-    """
-    convert the given rule set and capabilities result to a common, Python-native data structure.
-    this format can be directly emitted to JSON, or passed to the other `capa.render.*.render()` routines
-     to render as text.
-
-    see examples of substructures in above routines.
-
-    schema:
-
-    ```json
-    {
-      "meta": {...},
-      "rules: {
-        $rule-name: {
-          "meta": {...copied from rule.meta...},
-          "matches: {
-            $address: {...match details...},
-            ...
-          }
-        },
-        ...
-      }
-    }
-    ```
-
-    Args:
-      meta (Dict[str, Any]):
-      rules (RuleSet):
-      capabilities (Dict[str, List[Tuple[int, Result]]]):
-    """
-    doc = {
-        "meta": meta,
-        "rules": {},
-    }
-
-    for rule_name, matches in capabilities.items():
-        rule = rules[rule_name]
-
-        if rule.meta.get("capa/subscope-rule"):
-            continue
-
-        rule_meta = convert_meta_to_result_document(rule.meta)
-
-        doc["rules"][rule_name] = {
-            "meta": rule_meta,
-            "source": rule.definition,
-            "matches": {
-                addr: convert_match_to_result_document(rules, capabilities, match) for (addr, match) in matches
-            },
-        }
-
-    return doc
+        return ResultDocument(meta=Metadata.from_capa(meta), rules=rule_matches)

capa/render/utils.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -7,9 +7,12 @@
 # See the License for the specific language governing permissions and limitations under the License.
 
 import io
+from typing import Union, Iterator
 
 import termcolor
 
+import capa.render.result_document as rd
+
 
 def bold(s: str) -> str:
     """draw attention to the given string"""
@@ -21,48 +24,33 @@ def bold2(s: str) -> str:
     return termcolor.colored(s, "green")
 
 
-def hex(n: int) -> str:
-    """render the given number using upper case hex, like: 0x123ABC"""
-    if n < 0:
-        return "-0x%X" % (-n)
-    else:
-        return "0x%X" % n
+def warn(s: str) -> str:
+    return termcolor.colored(s, "yellow")
 
 
-def parse_parts_id(s: str):
-    id = ""
-    parts = s.split("::")
-    if len(parts) > 0:
-        last = parts.pop()
-        last, _, id = last.rpartition(" ")
-        id = id.lstrip("[").rstrip("]")
-        parts.append(last)
-    return parts, id
-
-
-def format_parts_id(data):
+def format_parts_id(data: Union[rd.AttackSpec, rd.MBCSpec]):
     """
     format canonical representation of ATT&CK/MBC parts and ID
     """
-    return "%s [%s]" % ("::".join(data["parts"]), data["id"])
+    return "%s [%s]" % ("::".join(data.parts), data.id)
 
 
-def capability_rules(doc):
+def capability_rules(doc: rd.ResultDocument) -> Iterator[rd.RuleMatches]:
     """enumerate the rules in (namespace, name) order that are 'capability' rules (not lib/subscope/disposition/etc)."""
-    for (_, _, rule) in sorted(
-        map(lambda rule: (rule["meta"].get("namespace", ""), rule["meta"]["name"], rule), doc["rules"].values())
-    ):
-        if rule["meta"].get("lib"):
+    for _, _, rule in sorted(map(lambda rule: (rule.meta.namespace or "", rule.meta.name, rule), doc.rules.values())):
+        if rule.meta.lib:
             continue
-        if rule["meta"].get("capa/subscope"):
+        if rule.meta.is_subscope_rule:
             continue
-        if rule["meta"].get("maec/analysis-conclusion"):
+        if rule.meta.maec.analysis_conclusion:
             continue
-        if rule["meta"].get("maec/analysis-conclusion-ov"):
+        if rule.meta.maec.analysis_conclusion_ov:
             continue
-        if rule["meta"].get("maec/malware-category"):
+        if rule.meta.maec.malware_family:
             continue
-        if rule["meta"].get("maec/malware-category-ov"):
+        if rule.meta.maec.malware_category:
+            continue
+        if rule.meta.maec.malware_category_ov:
             continue
 
         yield rule

capa/render/verbose.py

@@ -3,7 +3,7 @@ example::
 
     send data
     namespace    communication
-    author       william.ballenthin@fireeye.com
+    author       william.ballenthin@mandiant.com
     description  all known techniques for sending data to a potential C2 server
     scope        function
     examples     BFB9B5391A13D0AFD787E87AB90F14F5:0x13145D60
@@ -14,7 +14,7 @@ example::
                  0x10003415
                  0x10003797
 
-Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 Licensed under the Apache License, Version 2.0 (the "License");
  you may not use this file except in compliance with the License.
 You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -22,16 +22,38 @@ Unless required by applicable law or agreed to in writing, software distributed
  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and limitations under the License.
 """
+import enum
+
 import tabulate
 
 import capa.rules
+import capa.helpers
 import capa.render.utils as rutils
-import capa.render.result_document
+import capa.features.freeze as frz
+import capa.render.result_document as rd
 from capa.rules import RuleSet
 from capa.engine import MatchResults
 
 
-def render_meta(ostream, doc):
+def format_address(address: frz.Address) -> str:
+    if address.type == frz.AddressType.ABSOLUTE:
+        return capa.helpers.hex(address.value)
+    elif address.type == frz.AddressType.RELATIVE:
+        return f"base address+{capa.helpers.hex(address.value)}"
+    elif address.type == frz.AddressType.FILE:
+        return f"file+{capa.helpers.hex(address.value)}"
+    elif address.type == frz.AddressType.DN_TOKEN:
+        return f"token({capa.helpers.hex(address.value)})"
+    elif address.type == frz.AddressType.DN_TOKEN_OFFSET:
+        token, offset = address.value
+        return f"token({capa.helpers.hex(token)})+{capa.helpers.hex(offset)}"
+    elif address.type == frz.AddressType.NO_ADDRESS:
+        return "global"
+    else:
+        raise ValueError("unexpected address type")
+
+
+def render_meta(ostream, doc: rd.ResultDocument):
     """
     like:
 
@@ -51,30 +73,31 @@ def render_meta(ostream, doc):
         total feature count  1918
     """
     rows = [
-        ("md5", doc["meta"]["sample"]["md5"]),
-        ("sha1", doc["meta"]["sample"]["sha1"]),
-        ("sha256", doc["meta"]["sample"]["sha256"]),
-        ("path", doc["meta"]["sample"]["path"]),
-        ("timestamp", doc["meta"]["timestamp"]),
-        ("capa version", doc["meta"]["version"]),
-        ("os", doc["meta"]["analysis"]["os"]),
-        ("format", doc["meta"]["analysis"]["format"]),
-        ("arch", doc["meta"]["analysis"]["arch"]),
-        ("extractor", doc["meta"]["analysis"]["extractor"]),
-        ("base address", hex(doc["meta"]["analysis"]["base_address"])),
-        ("rules", doc["meta"]["analysis"]["rules"]),
-        ("function count", len(doc["meta"]["analysis"]["feature_counts"]["functions"])),
-        ("library function count", len(doc["meta"]["analysis"]["library_functions"])),
+        ("md5", doc.meta.sample.md5),
+        ("sha1", doc.meta.sample.sha1),
+        ("sha256", doc.meta.sample.sha256),
+        ("path", doc.meta.sample.path),
+        ("timestamp", doc.meta.timestamp),
+        ("capa version", doc.meta.version),
+        ("os", doc.meta.analysis.os),
+        ("format", doc.meta.analysis.format),
+        ("arch", doc.meta.analysis.arch),
+        ("extractor", doc.meta.analysis.extractor),
+        ("base address", format_address(doc.meta.analysis.base_address)),
+        ("rules", "\n".join(doc.meta.analysis.rules)),
+        ("function count", len(doc.meta.analysis.feature_counts.functions)),
+        ("library function count", len(doc.meta.analysis.library_functions)),
         (
             "total feature count",
-            doc["meta"]["analysis"]["feature_counts"]["file"]
-            + sum(doc["meta"]["analysis"]["feature_counts"]["functions"].values()),
+            doc.meta.analysis.feature_counts.file
+            + sum(map(lambda f: f.count, doc.meta.analysis.feature_counts.functions)),
         ),
     ]
+
     ostream.writeln(tabulate.tabulate(rows, tablefmt="plain"))
 
 
-def render_rules(ostream, doc):
+def render_rules(ostream, doc: rd.ResultDocument):
     """
     like:
 
@@ -87,28 +110,32 @@ def render_rules(ostream, doc):
     """
     had_match = False
     for rule in rutils.capability_rules(doc):
-        count = len(rule["matches"])
+        count = len(rule.matches)
         if count == 1:
-            capability = rutils.bold(rule["meta"]["name"])
+            capability = rutils.bold(rule.meta.name)
         else:
-            capability = "%s (%d matches)" % (rutils.bold(rule["meta"]["name"]), count)
+            capability = "%s (%d matches)" % (rutils.bold(rule.meta.name), count)
 
         ostream.writeln(capability)
         had_match = True
 
         rows = []
         for key in ("namespace", "description", "scope"):
-            if key == "name" or key not in rule["meta"]:
+            v = getattr(rule.meta, key)
+            if not v:
                 continue
 
-            v = rule["meta"][key]
             if isinstance(v, list) and len(v) == 1:
                 v = v[0]
+
+            if isinstance(v, enum.Enum):
+                v = v.value
+
             rows.append((key, v))
 
-        if rule["meta"]["scope"] != capa.rules.FILE_SCOPE:
-            locations = doc["rules"][rule["meta"]["name"]]["matches"].keys()
-            rows.append(("matches", "\n".join(map(rutils.hex, locations))))
+        if rule.meta.scope != capa.rules.FILE_SCOPE:
+            locations = list(map(lambda m: m[0], doc.rules[rule.meta.name].matches))
+            rows.append(("matches", "\n".join(map(format_address, locations))))
 
         ostream.writeln(tabulate.tabulate(rows, tablefmt="plain"))
         ostream.write("\n")
@@ -117,7 +144,7 @@ def render_rules(ostream, doc):
         ostream.writeln(rutils.bold("no capabilities found"))
 
 
-def render_verbose(doc):
+def render_verbose(doc: rd.ResultDocument):
     ostream = rutils.StringIO()
 
     render_meta(ostream, doc)
@@ -130,5 +157,4 @@ def render_verbose(doc):
 
 
 def render(meta, rules: RuleSet, capabilities: MatchResults) -> str:
-    doc = capa.render.result_document.convert_capabilities_to_result_document(meta, rules, capabilities)
-    return render_verbose(doc)
+    return render_verbose(rd.ResultDocument.from_capa(meta, rules, capabilities))

capa/render/vverbose.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -6,115 +6,177 @@
 #  is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and limitations under the License.
 
+from typing import Dict, Iterable
+
 import tabulate
 
 import capa.rules
+import capa.helpers
 import capa.render.utils as rutils
 import capa.render.verbose
 import capa.features.common
-import capa.render.result_document
+import capa.features.freeze as frz
+import capa.features.address
+import capa.render.result_document as rd
+import capa.features.freeze.features as frzf
 from capa.rules import RuleSet
 from capa.engine import MatchResults
 
 
-def render_locations(ostream, match):
+def render_locations(ostream, locations: Iterable[frz.Address]):
+    import capa.render.verbose as v
+
     # its possible to have an empty locations array here,
     # such as when we're in MODE_FAILURE and showing the logic
     # under a `not` statement (which will have no matched locations).
-    locations = list(sorted(match.get("locations", [])))
+    locations = list(sorted(locations))
+
+    if len(locations) == 0:
+        return
+
+    ostream.write(" @ ")
+
     if len(locations) == 1:
-        ostream.write(" @ ")
-        ostream.write(rutils.hex(locations[0]))
+        ostream.write(v.format_address(locations[0]))
+
+    elif len(locations) > 4:
+        # don't display too many locations, because it becomes very noisy.
+        # probably only the first handful of locations will be useful for inspection.
+        ostream.write(", ".join(map(v.format_address, locations[0:4])))
+        ostream.write(", and %d more..." % (len(locations) - 4))
+
     elif len(locations) > 1:
-        ostream.write(" @ ")
-        if len(locations) > 4:
-            # don't display too many locations, because it becomes very noisy.
-            # probably only the first handful of locations will be useful for inspection.
-            ostream.write(", ".join(map(rutils.hex, locations[0:4])))
-            ostream.write(", and %d more..." % (len(locations) - 4))
-        else:
-            ostream.write(", ".join(map(rutils.hex, locations)))
+        ostream.write(", ".join(map(v.format_address, locations)))
+
+    else:
+        raise RuntimeError("unreachable")
 
 
-def render_statement(ostream, match, statement, indent=0):
+def render_statement(ostream, match: rd.Match, statement: rd.Statement, indent=0):
     ostream.write("  " * indent)
-    if statement["type"] in ("and", "or", "optional", "not", "subscope"):
-        ostream.write(statement["type"])
+
+    if isinstance(statement, rd.SubscopeStatement):
+        # emit `basic block:`
+        # rather than `subscope:`
+        ostream.write(statement.scope)
+
         ostream.write(":")
-        if statement.get("description"):
-            ostream.write(" = %s" % statement["description"])
+        if statement.description:
+            ostream.write(" = %s" % statement.description)
         ostream.writeln("")
-    elif statement["type"] == "some":
-        ostream.write("%d or more:" % (statement["count"]))
-        if statement.get("description"):
-            ostream.write(" = %s" % statement["description"])
+
+    elif isinstance(statement, (rd.CompoundStatement)):
+        # emit `and:`  `or:`  `optional:`  `not:`
+        ostream.write(statement.type)
+
+        ostream.write(":")
+        if statement.description:
+            ostream.write(" = %s" % statement.description)
         ostream.writeln("")
-    elif statement["type"] == "range":
+
+    elif isinstance(statement, rd.SomeStatement):
+        ostream.write("%d or more:" % (statement.count))
+
+        if statement.description:
+            ostream.write(" = %s" % statement.description)
+        ostream.writeln("")
+
+    elif isinstance(statement, rd.RangeStatement):
         # `range` is a weird node, its almost a hybrid of statement+feature.
         # it is a specific feature repeated multiple times.
         # there's no additional logic in the feature part, just the existence of a feature.
         # so, we have to inline some of the feature rendering here.
 
-        child = statement["child"]
+        child = statement.child
+        value = child.dict(by_alias=True).get(child.type)
+
+        if value:
+            if isinstance(child, frzf.StringFeature):
+                value = '"%s"' % capa.features.common.escape_string(value)
 
-        if child[child["type"]]:
-            if child["type"] == "string":
-                value = '"%s"' % capa.features.common.escape_string(child[child["type"]])
-            else:
-                value = child[child["type"]]
             value = rutils.bold2(value)
-            if child.get("description"):
-                ostream.write("count(%s(%s = %s)): " % (child["type"], value, child["description"]))
+
+            if child.description:
+                ostream.write("count(%s(%s = %s)): " % (child.type, value, child.description))
             else:
-                ostream.write("count(%s(%s)): " % (child["type"], value))
+                ostream.write("count(%s(%s)): " % (child.type, value))
         else:
-            ostream.write("count(%s): " % child["type"])
+            ostream.write("count(%s): " % child.type)
 
-        if statement["max"] == statement["min"]:
-            ostream.write("%d" % (statement["min"]))
-        elif statement["min"] == 0:
-            ostream.write("%d or fewer" % (statement["max"]))
-        elif statement["max"] == (1 << 64 - 1):
-            ostream.write("%d or more" % (statement["min"]))
+        if statement.max == statement.min:
+            ostream.write("%d" % (statement.min))
+        elif statement.min == 0:
+            ostream.write("%d or fewer" % (statement.max))
+        elif statement.max == (1 << 64 - 1):
+            ostream.write("%d or more" % (statement.min))
         else:
-            ostream.write("between %d and %d" % (statement["min"], statement["max"]))
+            ostream.write("between %d and %d" % (statement.min, statement.max))
 
-        if statement.get("description"):
-            ostream.write(" = %s" % statement["description"])
-        render_locations(ostream, match)
+        if statement.description:
+            ostream.write(" = %s" % statement.description)
+        render_locations(ostream, match.locations)
         ostream.writeln("")
+
     else:
         raise RuntimeError("unexpected match statement type: " + str(statement))
 
 
-def render_string_value(s):
+def render_string_value(s: str) -> str:
     return '"%s"' % capa.features.common.escape_string(s)
 
 
-def render_feature(ostream, match, feature, indent=0):
+def render_feature(ostream, match: rd.Match, feature: frzf.Feature, indent=0):
     ostream.write("  " * indent)
 
-    key = feature["type"]
-    value = feature[feature["type"]]
+    key = feature.type
+    if isinstance(feature, frzf.BasicBlockFeature):
+        # i don't think it makes sense to have standalone basic block features.
+        # we don't parse them from rules, only things like: `count(basic block) > 1`
+        raise ValueError("cannot render basic block feature directly")
+    elif isinstance(feature, frzf.ImportFeature):
+        # fixup access to Python reserved name
+        value = feature.import_
+    elif isinstance(feature, frzf.ClassFeature):
+        value = feature.class_
+    else:
+        # convert attributes to dictionary using aliased names, if applicable
+        value = feature.dict(by_alias=True).get(key, None)
 
-    if key not in ("regex", "substring"):
+    if value is None:
+        raise ValueError("%s contains None" % key)
+
+    if not isinstance(feature, (frzf.RegexFeature, frzf.SubstringFeature)):
         # like:
         #   number: 10 = SOME_CONSTANT @ 0x401000
-        if key == "string":
+        if isinstance(feature, frzf.StringFeature):
             value = render_string_value(value)
 
-        ostream.write(key)
-        ostream.write(": ")
+        elif isinstance(
+            feature, (frzf.NumberFeature, frzf.OffsetFeature, frzf.OperandNumberFeature, frzf.OperandOffsetFeature)
+        ):
+            assert isinstance(value, int)
+            value = capa.helpers.hex(value)
+
+        if isinstance(feature, frzf.PropertyFeature) and feature.access is not None:
+            key = f"property/{feature.access}"
+
+        elif isinstance(feature, frzf.OperandNumberFeature):
+            key = f"operand[{feature.index}].number"
+
+        elif isinstance(feature, frzf.OperandOffsetFeature):
+            key = f"operand[{feature.index}].offset"
+
+        ostream.write(f"{key}: ")
 
         if value:
             ostream.write(rutils.bold2(value))
 
-            if "description" in feature:
+            if feature.description:
                 ostream.write(capa.rules.DESCRIPTION_SEPARATOR)
-                ostream.write(feature["description"])
+                ostream.write(feature.description)
 
-        if key not in ("os", "arch"):
-            render_locations(ostream, match)
+        if not isinstance(feature, (frzf.OSFeature, frzf.ArchFeature, frzf.FormatFeature)):
+            render_locations(ostream, match.locations)
         ostream.write("\n")
     else:
         # like:
@@ -126,19 +188,19 @@ def render_feature(ostream, match, feature, indent=0):
         ostream.write(value)
         ostream.write("\n")
 
-        for match, locations in sorted(feature["matches"].items(), key=lambda p: p[0]):
+        for capture, locations in sorted(match.captures.items()):
             ostream.write("  " * (indent + 1))
             ostream.write("- ")
-            ostream.write(rutils.bold2(render_string_value(match)))
-            render_locations(ostream, {"locations": locations})
+            ostream.write(rutils.bold2(render_string_value(capture)))
+            render_locations(ostream, locations)
             ostream.write("\n")
 
 
-def render_node(ostream, match, node, indent=0):
-    if node["type"] == "statement":
-        render_statement(ostream, match, node["statement"], indent=indent)
-    elif node["type"] == "feature":
-        render_feature(ostream, match, node["feature"], indent=indent)
+def render_node(ostream, match: rd.Match, node: rd.Node, indent=0):
+    if isinstance(node, rd.StatementNode):
+        render_statement(ostream, match, node.statement, indent=indent)
+    elif isinstance(node, rd.FeatureNode):
+        render_feature(ostream, match, node.feature, indent=indent)
     else:
         raise RuntimeError("unexpected node type: " + str(node))
 
@@ -151,111 +213,166 @@ MODE_SUCCESS = "success"
 MODE_FAILURE = "failure"
 
 
-def render_match(ostream, match, indent=0, mode=MODE_SUCCESS):
+def render_match(ostream, match: rd.Match, indent=0, mode=MODE_SUCCESS):
     child_mode = mode
     if mode == MODE_SUCCESS:
         # display only nodes that evaluated successfully.
-        if not match["success"]:
+        if not match.success:
             return
+
         # optional statement with no successful children is empty
-        if match["node"].get("statement", {}).get("type") == "optional" and not any(
-            map(lambda m: m["success"], match["children"])
-        ):
-            return
+        if isinstance(match.node, rd.StatementNode) and match.node.statement.type == rd.CompoundStatementType.OPTIONAL:
+            if not any(map(lambda m: m.success, match.children)):
+                return
+
         # not statement, so invert the child mode to show failed evaluations
-        if match["node"].get("statement", {}).get("type") == "not":
+        if isinstance(match.node, rd.StatementNode) and match.node.statement.type == rd.CompoundStatementType.NOT:
             child_mode = MODE_FAILURE
+
     elif mode == MODE_FAILURE:
         # display only nodes that did not evaluate to True
-        if match["success"]:
+        if match.success:
             return
+
         # optional statement with successful children is not relevant
-        if match["node"].get("statement", {}).get("type") == "optional" and any(
-            map(lambda m: m["success"], match["children"])
-        ):
-            return
+        if isinstance(match.node, rd.StatementNode) and match.node.statement.type == rd.CompoundStatementType.OPTIONAL:
+            if any(map(lambda m: m.success, match.children)):
+                return
+
         # not statement, so invert the child mode to show successful evaluations
-        if match["node"].get("statement", {}).get("type") == "not":
+        if isinstance(match.node, rd.StatementNode) and match.node.statement.type == rd.CompoundStatementType.NOT:
             child_mode = MODE_SUCCESS
     else:
         raise RuntimeError("unexpected mode: " + mode)
 
-    render_node(ostream, match, match["node"], indent=indent)
+    render_node(ostream, match, match.node, indent=indent)
 
-    for child in match["children"]:
+    for child in match.children:
         render_match(ostream, child, indent=indent + 1, mode=child_mode)
 
 
-def render_rules(ostream, doc):
+def render_rules(ostream, doc: rd.ResultDocument):
     """
     like:
 
         ## rules
         check for OutputDebugString error
         namespace  anti-analysis/anti-debugging/debugger-detection
-        author     michael.hunhoff@fireeye.com
+        author     michael.hunhoff@mandiant.com
         scope      function
         mbc        Anti-Behavioral Analysis::Detect Debugger::OutputDebugString
-        examples   Practical Malware Analysis Lab 16-02.exe_:0x401020
         function @ 0x10004706
           and:
             api: kernel32.SetLastError @ 0x100047C2
             api: kernel32.GetLastError @ 0x10004A87
             api: kernel32.OutputDebugString @ 0x10004767, 0x10004787, 0x10004816, 0x10004895
     """
+    functions_by_bb: Dict[capa.features.address.Address, capa.features.address.Address] = {}
+    for finfo in doc.meta.analysis.layout.functions:
+        faddress = finfo.address.to_capa()
+
+        for bb in finfo.matched_basic_blocks:
+            bbaddress = bb.address.to_capa()
+            functions_by_bb[bbaddress] = faddress
+
     had_match = False
-    for rule in rutils.capability_rules(doc):
-        count = len(rule["matches"])
+
+    for _, _, rule in sorted(map(lambda rule: (rule.meta.namespace or "", rule.meta.name, rule), doc.rules.values())):
+        # default scope hides things like lib rules, malware-category rules, etc.
+        # but in vverbose mode, we really want to show everything.
+        #
+        # still ignore subscope rules because they're stitched into the final document.
+        if rule.meta.is_subscope_rule:
+            continue
+
+        lib_info = ""
+        count = len(rule.matches)
         if count == 1:
-            capability = rutils.bold(rule["meta"]["name"])
+            if rule.meta.lib:
+                lib_info = " (library rule)"
+            capability = "%s%s" % (rutils.bold(rule.meta.name), lib_info)
         else:
-            capability = "%s (%d matches)" % (rutils.bold(rule["meta"]["name"]), count)
+            if rule.meta.lib:
+                lib_info = ", only showing first match of library rule"
+            capability = "%s (%d matches%s)" % (rutils.bold(rule.meta.name), count, lib_info)
 
         ostream.writeln(capability)
         had_match = True
 
         rows = []
-        for key in capa.rules.META_KEYS:
-            if key == "name" or key not in rule["meta"]:
-                continue
+        if not rule.meta.lib:
+            # library rules should not have a namespace
+            rows.append(("namespace", rule.meta.namespace))
 
-            v = rule["meta"][key]
-            if not v:
-                continue
+        if rule.meta.maec.analysis_conclusion or rule.meta.maec.analysis_conclusion_ov:
+            rows.append(
+                (
+                    "maec/analysis-conclusion",
+                    rule.meta.maec.analysis_conclusion or rule.meta.maec.analysis_conclusion_ov,
+                )
+            )
 
-            if key in ("att&ck", "mbc"):
-                v = [rutils.format_parts_id(vv) for vv in v]
+        if rule.meta.maec.malware_family:
+            rows.append(("maec/malware-family", rule.meta.maec.malware_family))
 
-            if isinstance(v, list) and len(v) == 1:
-                v = v[0]
-            elif isinstance(v, list) and len(v) > 1:
-                v = ", ".join(v)
-            rows.append((key, v))
+        if rule.meta.maec.malware_category or rule.meta.maec.malware_category_ov:
+            rows.append(
+                ("maec/malware-category", rule.meta.maec.malware_category or rule.meta.maec.malware_category_ov)
+            )
+
+        rows.append(("author", ", ".join(rule.meta.authors)))
+
+        rows.append(("scope", rule.meta.scope.value))
+
+        if rule.meta.attack:
+            rows.append(("att&ck", ", ".join([rutils.format_parts_id(v) for v in rule.meta.attack])))
+
+        if rule.meta.mbc:
+            rows.append(("mbc", ", ".join([rutils.format_parts_id(v) for v in rule.meta.mbc])))
+
+        if rule.meta.references:
+            rows.append(("references", ", ".join(rule.meta.references)))
+
+        if rule.meta.description:
+            rows.append(("description", rule.meta.description))
 
         ostream.writeln(tabulate.tabulate(rows, tablefmt="plain"))
 
-        if rule["meta"]["scope"] == capa.rules.FILE_SCOPE:
-            matches = list(doc["rules"][rule["meta"]["name"]]["matches"].values())
+        if rule.meta.scope == capa.rules.FILE_SCOPE:
+            matches = doc.rules[rule.meta.name].matches
             if len(matches) != 1:
                 # i think there should only ever be one match per file-scope rule,
                 # because we do the file-scope evaluation a single time.
                 # but i'm not 100% sure if this is/will always be true.
                 # so, lets be explicit about our assumptions and raise an exception if they fail.
                 raise RuntimeError("unexpected file scope match count: %d" % (len(matches)))
-            render_match(ostream, matches[0], indent=0)
+            first_address, first_match = matches[0]
+            render_match(ostream, first_match, indent=0)
         else:
-            for location, match in sorted(doc["rules"][rule["meta"]["name"]]["matches"].items()):
-                ostream.write(rule["meta"]["scope"])
+            for location, match in sorted(doc.rules[rule.meta.name].matches):
+                ostream.write(rule.meta.scope)
                 ostream.write(" @ ")
-                ostream.writeln(rutils.hex(location))
+                ostream.write(capa.render.verbose.format_address(location))
+
+                if rule.meta.scope == capa.rules.BASIC_BLOCK_SCOPE:
+                    ostream.write(
+                        " in function "
+                        + capa.render.verbose.format_address(frz.Address.from_capa(functions_by_bb[location.to_capa()]))
+                    )
+
+                ostream.write("\n")
                 render_match(ostream, match, indent=1)
+                if rule.meta.lib:
+                    # only show first match
+                    break
+
         ostream.write("\n")
 
     if not had_match:
         ostream.writeln(rutils.bold("no capabilities found"))
 
 
-def render_vverbose(doc):
+def render_vverbose(doc: rd.ResultDocument):
     ostream = rutils.StringIO()
 
     capa.render.verbose.render_meta(ostream, doc)
@@ -268,5 +385,4 @@ def render_vverbose(doc):
 
 
 def render(meta, rules: RuleSet, capabilities: MatchResults) -> str:
-    doc = capa.render.result_document.convert_capabilities_to_result_document(meta, rules, capabilities)
-    return render_vverbose(doc)
+    return render_vverbose(rd.ResultDocument.from_capa(meta, rules, capabilities))

capa/rules/__init__.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 # Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at: [package root]/LICENSE.txt
@@ -12,8 +12,10 @@ import uuid
 import codecs
 import logging
 import binascii
-import functools
 import collections
+from enum import Enum
+
+from capa.helpers import assert_never
 
 try:
     from functools import lru_cache
@@ -22,19 +24,24 @@ except ImportError:
     # https://github.com/python/mypy/issues/1153
     from backports.functools_lru_cache import lru_cache  # type: ignore
 
-from typing import Any, Dict, List, Union, Iterator
+from typing import Any, Set, Dict, List, Tuple, Union, Iterator
 
 import yaml
+import pydantic
 import ruamel.yaml
+import yaml.parser
 
+import capa.perf
 import capa.engine as ceng
 import capa.features
+import capa.optimizer
 import capa.features.file
 import capa.features.insn
 import capa.features.common
 import capa.features.basicblock
 from capa.engine import Statement, FeatureSet
 from capa.features.common import MAX_BYTES_FEATURE_SIZE, Feature
+from capa.features.address import Address
 
 logger = logging.getLogger(__name__)
 
@@ -43,12 +50,12 @@ logger = logging.getLogger(__name__)
 META_KEYS = (
     "name",
     "namespace",
-    "rule-category",
     "maec/analysis-conclusion",
     "maec/analysis-conclusion-ov",
+    "maec/malware-family",
     "maec/malware-category",
     "maec/malware-category-ov",
-    "author",
+    "authors",
     "description",
     "lib",
     "scope",
@@ -64,12 +71,29 @@ META_KEYS = (
 HIDDEN_META_KEYS = ("capa/nursery", "capa/path")
 
 
-FILE_SCOPE = "file"
-FUNCTION_SCOPE = "function"
-BASIC_BLOCK_SCOPE = "basic block"
+class Scope(str, Enum):
+    FILE = "file"
+    FUNCTION = "function"
+    BASIC_BLOCK = "basic block"
+    INSTRUCTION = "instruction"
 
 
-SUPPORTED_FEATURES = {
+FILE_SCOPE = Scope.FILE.value
+FUNCTION_SCOPE = Scope.FUNCTION.value
+BASIC_BLOCK_SCOPE = Scope.BASIC_BLOCK.value
+INSTRUCTION_SCOPE = Scope.INSTRUCTION.value
+# used only to specify supported features per scope.
+# not used to validate rules.
+GLOBAL_SCOPE = "global"
+
+
+SUPPORTED_FEATURES: Dict[str, Set] = {
+    GLOBAL_SCOPE: {
+        # these will be added to other scopes, see below.
+        capa.features.common.OS,
+        capa.features.common.Arch,
+        capa.features.common.Format,
+    },
     FILE_SCOPE: {
         capa.features.common.MatchedRule,
         capa.features.file.Export,
@@ -78,48 +102,64 @@ SUPPORTED_FEATURES = {
         capa.features.file.FunctionName,
         capa.features.common.Characteristic("embedded pe"),
         capa.features.common.String,
-        capa.features.common.Format,
-        capa.features.common.OS,
-        capa.features.common.Arch,
+        capa.features.common.Class,
+        capa.features.common.Namespace,
+        capa.features.common.Characteristic("mixed mode"),
     },
     FUNCTION_SCOPE: {
-        # plus basic block scope features, see below
+        capa.features.common.MatchedRule,
         capa.features.basicblock.BasicBlock,
         capa.features.common.Characteristic("calls from"),
         capa.features.common.Characteristic("calls to"),
         capa.features.common.Characteristic("loop"),
         capa.features.common.Characteristic("recursive call"),
-        capa.features.common.OS,
-        capa.features.common.Arch,
+        # plus basic block scope features, see below
     },
     BASIC_BLOCK_SCOPE: {
+        capa.features.common.MatchedRule,
+        capa.features.common.Characteristic("tight loop"),
+        capa.features.common.Characteristic("stack string"),
+        # plus instruction scope features, see below
+    },
+    INSTRUCTION_SCOPE: {
         capa.features.common.MatchedRule,
         capa.features.insn.API,
+        capa.features.insn.Property,
         capa.features.insn.Number,
         capa.features.common.String,
         capa.features.common.Bytes,
         capa.features.insn.Offset,
         capa.features.insn.Mnemonic,
+        capa.features.insn.OperandNumber,
+        capa.features.insn.OperandOffset,
         capa.features.common.Characteristic("nzxor"),
         capa.features.common.Characteristic("peb access"),
         capa.features.common.Characteristic("fs access"),
         capa.features.common.Characteristic("gs access"),
-        capa.features.common.Characteristic("cross section flow"),
-        capa.features.common.Characteristic("tight loop"),
-        capa.features.common.Characteristic("stack string"),
         capa.features.common.Characteristic("indirect call"),
-        capa.features.common.OS,
-        capa.features.common.Arch,
+        capa.features.common.Characteristic("call $+5"),
+        capa.features.common.Characteristic("cross section flow"),
+        capa.features.common.Characteristic("unmanaged call"),
+        capa.features.common.Class,
+        capa.features.common.Namespace,
     },
 }
 
+# global scope features are available in all other scopes
+SUPPORTED_FEATURES[INSTRUCTION_SCOPE].update(SUPPORTED_FEATURES[GLOBAL_SCOPE])
+SUPPORTED_FEATURES[BASIC_BLOCK_SCOPE].update(SUPPORTED_FEATURES[GLOBAL_SCOPE])
+SUPPORTED_FEATURES[FUNCTION_SCOPE].update(SUPPORTED_FEATURES[GLOBAL_SCOPE])
+SUPPORTED_FEATURES[FILE_SCOPE].update(SUPPORTED_FEATURES[GLOBAL_SCOPE])
+
+# all instruction scope features are also basic block features
+SUPPORTED_FEATURES[BASIC_BLOCK_SCOPE].update(SUPPORTED_FEATURES[INSTRUCTION_SCOPE])
 # all basic block scope features are also function scope features
 SUPPORTED_FEATURES[FUNCTION_SCOPE].update(SUPPORTED_FEATURES[BASIC_BLOCK_SCOPE])
 
 
 class InvalidRule(ValueError):
     def __init__(self, msg):
-        super(InvalidRule, self).__init__()
+        super().__init__()
         self.msg = msg
 
     def __str__(self):
@@ -131,7 +171,7 @@ class InvalidRule(ValueError):
 
 class InvalidRuleWithPath(InvalidRule):
     def __init__(self, path, msg):
-        super(InvalidRuleWithPath, self).__init__(msg)
+        super().__init__(msg)
         self.path = path
         self.msg = msg
         self.__cause__ = None
@@ -142,7 +182,7 @@ class InvalidRuleWithPath(InvalidRule):
 
 class InvalidRuleSet(ValueError):
     def __init__(self, msg):
-        super(InvalidRuleSet, self).__init__()
+        super().__init__()
         self.msg = msg
 
     def __str__(self):
@@ -194,23 +234,23 @@ def parse_range(s: str):
     min_spec = min_spec.strip()
     max_spec = max_spec.strip()
 
-    min = None
+    min_ = None
     if min_spec:
-        min = parse_int(min_spec)
-        if min < 0:
+        min_ = parse_int(min_spec)
+        if min_ < 0:
             raise InvalidRule("range min less than zero")
 
-    max = None
+    max_ = None
     if max_spec:
-        max = parse_int(max_spec)
-        if max < 0:
+        max_ = parse_int(max_spec)
+        if max_ < 0:
             raise InvalidRule("range max less than zero")
 
-    if min is not None and max is not None:
-        if max < min:
+    if min_ is not None and max_ is not None:
+        if max_ < min_:
             raise InvalidRule("range max less than min")
 
-    return min, max
+    return min_, max_
 
 
 def parse_feature(key: str):
@@ -225,20 +265,8 @@ def parse_feature(key: str):
         return capa.features.common.Bytes
     elif key == "number":
         return capa.features.insn.Number
-    elif key.startswith("number/"):
-        bitness = key.partition("/")[2]
-        # the other handlers here return constructors for features,
-        # and we want to as well,
-        # however, we need to preconfigure one of the arguments (`bitness`).
-        # so, instead we return a partially-applied function that
-        #  provides `bitness` to the feature constructor.
-        # it forwards any other arguments provided to the closure along to the constructor.
-        return functools.partial(capa.features.insn.Number, bitness=bitness)
     elif key == "offset":
         return capa.features.insn.Offset
-    elif key.startswith("offset/"):
-        bitness = key.partition("/")[2]
-        return functools.partial(capa.features.insn.Offset, bitness=bitness)
     elif key == "mnemonic":
         return capa.features.insn.Mnemonic
     elif key == "basic blocks":
@@ -260,8 +288,13 @@ def parse_feature(key: str):
     elif key == "format":
         return capa.features.common.Format
     elif key == "arch":
-
         return capa.features.common.Arch
+    elif key == "class":
+        return capa.features.common.Class
+    elif key == "namespace":
+        return capa.features.common.Namespace
+    elif key == "property":
+        return capa.features.insn.Property
     else:
         raise InvalidRule("unexpected statement: %s" % key)
 
@@ -328,7 +361,14 @@ def parse_description(s: Union[str, int, bytes], value_type: str, description=No
             # the string "10" that needs to become the number 10.
             if value_type == "bytes":
                 value = parse_bytes(value)
-            elif value_type in ("number", "offset") or value_type.startswith(("number/", "offset/")):
+            elif (
+                value_type in ("number", "offset")
+                or value_type.startswith(("number/", "offset/"))
+                or (
+                    value_type.startswith("operand[")
+                    and (value_type.endswith("].number") or value_type.endswith("].offset"))
+                )
+            ):
                 try:
                     value = parse_int(value)
                 except ValueError:
@@ -406,7 +446,7 @@ def build_statements(d, scope: str):
         if len(d[key]) != 1:
             raise InvalidRule("subscope must have exactly one child statement")
 
-        return ceng.Subscope(FUNCTION_SCOPE, build_statements(d[key][0], FUNCTION_SCOPE))
+        return ceng.Subscope(FUNCTION_SCOPE, build_statements(d[key][0], FUNCTION_SCOPE), description=description)
 
     elif key == "basic block":
         if scope != FUNCTION_SCOPE:
@@ -415,7 +455,30 @@ def build_statements(d, scope: str):
         if len(d[key]) != 1:
             raise InvalidRule("subscope must have exactly one child statement")
 
-        return ceng.Subscope(BASIC_BLOCK_SCOPE, build_statements(d[key][0], BASIC_BLOCK_SCOPE))
+        return ceng.Subscope(BASIC_BLOCK_SCOPE, build_statements(d[key][0], BASIC_BLOCK_SCOPE), description=description)
+
+    elif key == "instruction":
+        if scope not in (FUNCTION_SCOPE, BASIC_BLOCK_SCOPE):
+            raise InvalidRule("instruction subscope supported only for function and basic block scope")
+
+        if len(d[key]) == 1:
+            statements = build_statements(d[key][0], INSTRUCTION_SCOPE)
+        else:
+            # for instruction subscopes, we support a shorthand in which the top level AND is implied.
+            # the following are equivalent:
+            #
+            #     - instruction:
+            #       - and:
+            #         - arch: i386
+            #         - mnemonic: cmp
+            #
+            #     - instruction:
+            #       - arch: i386
+            #       - mnemonic: cmp
+            #
+            statements = ceng.And([build_statements(dd, INSTRUCTION_SCOPE) for dd in d[key]])
+
+        return ceng.Subscope(INSTRUCTION_SCOPE, statements, description=description)
 
     elif key.startswith("count(") and key.endswith(")"):
         # e.g.:
@@ -472,19 +535,66 @@ def build_statements(d, scope: str):
             raise InvalidRule("unexpected range: %s" % (count))
     elif key == "string" and not isinstance(d[key], str):
         raise InvalidRule("ambiguous string value %s, must be defined as explicit string" % d[key])
+
+    elif key.startswith("operand[") and key.endswith("].number"):
+        index = key[len("operand[") : -len("].number")]
+        try:
+            index = int(index)
+        except ValueError as e:
+            raise InvalidRule("operand index must be an integer") from e
+
+        value, description = parse_description(d[key], key, d.get("description"))
+        assert isinstance(value, int)
+        try:
+            feature = capa.features.insn.OperandNumber(index, value, description=description)
+        except ValueError as e:
+            raise InvalidRule(str(e)) from e
+        ensure_feature_valid_for_scope(scope, feature)
+        return feature
+
+    elif key.startswith("operand[") and key.endswith("].offset"):
+        index = key[len("operand[") : -len("].offset")]
+        try:
+            index = int(index)
+        except ValueError as e:
+            raise InvalidRule("operand index must be an integer") from e
+
+        value, description = parse_description(d[key], key, d.get("description"))
+        assert isinstance(value, int)
+        try:
+            feature = capa.features.insn.OperandOffset(index, value, description=description)
+        except ValueError as e:
+            raise InvalidRule(str(e)) from e
+        ensure_feature_valid_for_scope(scope, feature)
+        return feature
+
     elif (
         (key == "os" and d[key] not in capa.features.common.VALID_OS)
         or (key == "format" and d[key] not in capa.features.common.VALID_FORMAT)
         or (key == "arch" and d[key] not in capa.features.common.VALID_ARCH)
     ):
         raise InvalidRule("unexpected %s value %s" % (key, d[key]))
+
+    elif key.startswith("property/"):
+        access = key[len("property/") :]
+        if access not in capa.features.common.VALID_FEATURE_ACCESS:
+            raise InvalidRule("unexpected %s access %s" % (key, access))
+
+        value, description = parse_description(d[key], key, d.get("description"))
+        try:
+            feature = capa.features.insn.Property(value, access=access, description=description)
+        except ValueError as e:
+            raise InvalidRule(str(e)) from e
+        ensure_feature_valid_for_scope(scope, feature)
+        return feature
+
     else:
         Feature = parse_feature(key)
         value, description = parse_description(d[key], key, d.get("description"))
         try:
             feature = Feature(value, description=description)
         except ValueError as e:
-            raise InvalidRule(str(e))
+            raise InvalidRule(str(e)) from e
         ensure_feature_valid_for_scope(scope, feature)
         return feature
 
@@ -499,7 +609,7 @@ def second(s: List[Any]) -> Any:
 
 class Rule:
     def __init__(self, name: str, scope: str, statement: Statement, meta, definition=""):
-        super(Rule, self).__init__()
+        super().__init__()
         self.name = name
         self.scope = scope
         self.statement = statement
@@ -525,7 +635,7 @@ class Rule:
         Returns:
           List[str]: names of rules upon which this rule depends.
         """
-        deps = set([])
+        deps: Set[str] = set([])
 
         def rec(statement):
             if isinstance(statement, capa.features.common.MatchedRule):
@@ -542,6 +652,7 @@ class Rule:
                     deps.update(map(lambda r: r.name, namespaces[statement.value]))
                 else:
                     # not a namespace, assume its a rule name.
+                    assert isinstance(statement.value, str)
                     deps.add(statement.value)
 
             elif isinstance(statement, ceng.Statement):
@@ -557,7 +668,11 @@ class Rule:
     def _extract_subscope_rules_rec(self, statement):
         if isinstance(statement, ceng.Statement):
             # for each child that is a subscope,
-            for subscope in filter(lambda statement: isinstance(statement, ceng.Subscope), statement.get_children()):
+            for child in statement.get_children():
+                if not isinstance(child, ceng.Subscope):
+                    continue
+
+                subscope = child
 
                 # create a new rule from it.
                 # the name is a randomly generated, hopefully unique value.
@@ -595,6 +710,9 @@ class Rule:
                 for new_rule in self._extract_subscope_rules_rec(child):
                     yield new_rule
 
+    def is_subscope_rule(self):
+        return bool(self.meta.get("capa/subscope-rule", False))
+
     def extract_subscope_rules(self):
         """
         scan through the statements of this rule,
@@ -619,11 +737,13 @@ class Rule:
         for new_rule in self._extract_subscope_rules_rec(self.statement):
             yield new_rule
 
-    def evaluate(self, features: FeatureSet):
-        return self.statement.evaluate(features)
+    def evaluate(self, features: FeatureSet, short_circuit=True):
+        capa.perf.counters["evaluate.feature"] += 1
+        capa.perf.counters["evaluate.feature.rule"] += 1
+        return self.statement.evaluate(features, short_circuit=short_circuit)
 
     @classmethod
-    def from_dict(cls, d, definition):
+    def from_dict(cls, d: Dict[str, Any], definition: str) -> "Rule":
         meta = d["rule"]["meta"]
         name = meta["name"]
         # if scope is not specified, default to function scope.
@@ -657,14 +777,12 @@ class Rule:
             # prefer to use CLoader to be fast, see #306
             # on Linux, make sure you install libyaml-dev or similar
             # on Windows, get WHLs from pyyaml.org/pypi
-            loader = yaml.CLoader
             logger.debug("using libyaml CLoader.")
+            return yaml.CLoader
         except:
-            loader = yaml.Loader
             logger.debug("unable to import libyaml CLoader, falling back to Python yaml parser.")
             logger.debug("this will be slower to load rules.")
-
-        return loader
+            return yaml.Loader
 
     @staticmethod
     def _get_ruamel_yaml_parser():
@@ -676,8 +794,9 @@ class Rule:
         # use block mode, not inline json-like mode
         y.default_flow_style = False
 
-        # leave quotes unchanged
-        y.preserve_quotes = True
+        # leave quotes unchanged.
+        # manually verified this property exists, even if mypy complains.
+        y.preserve_quotes = True  # type: ignore
 
         # indent lists by two spaces below their parent
         #
@@ -688,12 +807,13 @@ class Rule:
         y.indent(sequence=2, offset=2)
 
         # avoid word wrapping
-        y.width = 4096
+        # manually verified this property exists, even if mypy complains.
+        y.width = 4096  # type: ignore
 
         return y
 
     @classmethod
-    def from_yaml(cls, s, use_ruamel=False):
+    def from_yaml(cls, s: str, use_ruamel=False) -> "Rule":
         if use_ruamel:
             # ruamel enables nice formatting and doc roundtripping with comments
             doc = cls._get_ruamel_yaml_parser().load(s)
@@ -703,14 +823,24 @@ class Rule:
         return cls.from_dict(doc, s)
 
     @classmethod
-    def from_yaml_file(cls, path, use_ruamel=False):
+    def from_yaml_file(cls, path, use_ruamel=False) -> "Rule":
         with open(path, "rb") as f:
             try:
-                return cls.from_yaml(f.read().decode("utf-8"), use_ruamel=use_ruamel)
-            except InvalidRule as e:
-                raise InvalidRuleWithPath(path, str(e))
+                rule = cls.from_yaml(f.read().decode("utf-8"), use_ruamel=use_ruamel)
+                # import here to avoid circular dependency
+                from capa.render.result_document import RuleMetadata
 
-    def to_yaml(self):
+                # validate meta data fields
+                _ = RuleMetadata.from_capa(rule)
+                return rule
+            except InvalidRule as e:
+                raise InvalidRuleWithPath(path, str(e)) from e
+            except pydantic.ValidationError as e:
+                raise InvalidRuleWithPath(path, str(e)) from e
+            except yaml.parser.ParserError as e:
+                raise InvalidRuleWithPath(path, str(e)) from e
+
+    def to_yaml(self) -> str:
         # reformat the yaml document with a common style.
         # this includes:
         #  - ordering the meta elements
@@ -947,10 +1077,18 @@ class RuleSet:
     """
 
     def __init__(self, rules: List[Rule]):
-        super(RuleSet, self).__init__()
+        super().__init__()
 
         ensure_rules_are_unique(rules)
 
+        # in the next step we extract subscope rules,
+        # which may inflate the number of rules tracked in this ruleset.
+        # so record number of rules initially provided to this ruleset.
+        #
+        # this number is really only meaningful to the user,
+        # who may compare it against the number of files on their file system.
+        self.source_rule_count = len(rules)
+
         rules = self._extract_subscope_rules(rules)
 
         ensure_rule_dependencies_are_met(rules)
@@ -958,12 +1096,27 @@ class RuleSet:
         if len(rules) == 0:
             raise InvalidRuleSet("no rules selected")
 
+        rules = capa.optimizer.optimize_rules(rules)
+
         self.file_rules = self._get_rules_for_scope(rules, FILE_SCOPE)
         self.function_rules = self._get_rules_for_scope(rules, FUNCTION_SCOPE)
         self.basic_block_rules = self._get_rules_for_scope(rules, BASIC_BLOCK_SCOPE)
+        self.instruction_rules = self._get_rules_for_scope(rules, INSTRUCTION_SCOPE)
         self.rules = {rule.name: rule for rule in rules}
         self.rules_by_namespace = index_rules_by_namespace(rules)
 
+        # unstable
+        (self._easy_file_rules_by_feature, self._hard_file_rules) = self._index_rules_by_feature(self.file_rules)
+        (self._easy_function_rules_by_feature, self._hard_function_rules) = self._index_rules_by_feature(
+            self.function_rules
+        )
+        (self._easy_basic_block_rules_by_feature, self._hard_basic_block_rules) = self._index_rules_by_feature(
+            self.basic_block_rules
+        )
+        (self._easy_instruction_rules_by_feature, self._hard_instruction_rules) = self._index_rules_by_feature(
+            self.instruction_rules
+        )
+
     def __len__(self):
         return len(self.rules)
 
@@ -974,7 +1127,157 @@ class RuleSet:
         return rulename in self.rules
 
     @staticmethod
-    def _get_rules_for_scope(rules, scope):
+    def _index_rules_by_feature(rules) -> Tuple[Dict[Feature, Set[str]], List[str]]:
+        """
+        split the given rules into two structures:
+          - "easy rules" are indexed by feature,
+            such that you can quickly find the rules that contain a given feature.
+          - "hard rules" are those that contain substring/regex/bytes features or match statements.
+            these continue to be ordered topologically.
+
+        a rule evaluator can use the "easy rule" index to restrict the
+        candidate rules that might match a given set of features.
+
+        at this time, a rule evaluator can't do anything special with
+        the "hard rules". it must still do a full top-down match of each
+        rule, in topological order.
+
+        this does not index global features, because these are not selective, and
+        won't be used as the sole feature used to match.
+        """
+
+        # we'll do a couple phases:
+        #
+        #  1. recursively visit all nodes in all rules,
+        #    a. indexing all features
+        #    b. recording the types of features found per rule
+        #  2. compute the easy and hard rule sets
+        #  3. remove hard rules from the rules-by-feature index
+        #  4. construct the topologically ordered list of hard rules
+        rules_with_easy_features: Set[str] = set()
+        rules_with_hard_features: Set[str] = set()
+        rules_by_feature: Dict[Feature, Set[str]] = collections.defaultdict(set)
+
+        def rec(rule_name: str, node: Union[Feature, Statement]):
+            """
+            walk through a rule's logic tree, indexing the easy and hard rules,
+            and the features referenced by easy rules.
+            """
+            if isinstance(
+                node,
+                (
+                    # these are the "hard features"
+                    # substring: scanning feature
+                    capa.features.common.Substring,
+                    # regex: scanning feature
+                    capa.features.common.Regex,
+                    # bytes: scanning feature
+                    capa.features.common.Bytes,
+                    # match: dependency on another rule,
+                    # which we have to evaluate first,
+                    # and is therefore tricky.
+                    capa.features.common.MatchedRule,
+                ),
+            ):
+                # hard feature: requires scan or match lookup
+                rules_with_hard_features.add(rule_name)
+            elif isinstance(node, capa.features.common.Feature):
+                if capa.features.common.is_global_feature(node):
+                    # we don't want to index global features
+                    # because they're not very selective.
+                    #
+                    # they're global, so if they match at one location in a file,
+                    # they'll match at every location in a file.
+                    # so thats not helpful to decide how to downselect.
+                    #
+                    # and, a global rule will never be the sole selector in a rule.
+                    # TODO: probably want a lint for this.
+                    pass
+                else:
+                    # easy feature: hash lookup
+                    rules_with_easy_features.add(rule_name)
+                    rules_by_feature[node].add(rule_name)
+            elif isinstance(node, (ceng.Not)):
+                # `not:` statements are tricky to deal with.
+                #
+                # first, features found under a `not:` should not be indexed,
+                # because they're not wanted to be found.
+                # second, `not:` can be nested under another `not:`, or two, etc.
+                # third, `not:` at the root or directly under an `or:`
+                # means the rule will match against *anything* not specified there,
+                # which is a difficult set of things to compute and index.
+                #
+                # so, if a rule has a `not:` statement, its hard.
+                # as of writing, this is an uncommon statement, with only 6 instances in 740 rules.
+                rules_with_hard_features.add(rule_name)
+            elif isinstance(node, (ceng.Some)) and node.count == 0:
+                # `optional:` and `0 or more:` are tricky to deal with.
+                #
+                # when a subtree is optional, it may match, but not matching
+                # doesn't have any impact either.
+                # now, our rule authors *should* not put this under `or:`
+                # and this is checked by the linter,
+                # but this could still happen (e.g. private rule set without linting)
+                # and would be hard to trace down.
+                #
+                # so better to be safe than sorry and consider this a hard case.
+                rules_with_hard_features.add(rule_name)
+            elif isinstance(node, (ceng.Range)) and node.min == 0:
+                # `count(foo): 0 or more` are tricky to deal with.
+                # because the min is 0,
+                # this subtree *can* match just about any feature
+                # (except the given one)
+                # which is a difficult set of things to compute and index.
+                rules_with_hard_features.add(rule_name)
+            elif isinstance(node, (ceng.Range)):
+                rec(rule_name, node.child)
+            elif isinstance(node, (ceng.And, ceng.Or, ceng.Some)):
+                for child in node.children:
+                    rec(rule_name, child)
+            elif isinstance(node, ceng.Statement):
+                # unhandled type of statement.
+                # this should only happen if a new subtype of `Statement`
+                # has since been added to capa.
+                #
+                # ideally, we'd like to use mypy for exhaustiveness checking
+                # for all the subtypes of `Statement`.
+                # but, as far as i can tell, mypy does not support this type
+                # of checking.
+                #
+                # in a way, this makes some intuitive sense:
+                # the set of subtypes of type A is unbounded,
+                # because any user might come along and create a new subtype B,
+                # so mypy can't reason about this set of types.
+                assert False, f"Unhandled value: {node} ({type(node).__name__})"
+            else:
+                # programming error
+                assert_never(node)
+
+        for rule in rules:
+            rule_name = rule.meta["name"]
+            root = rule.statement
+            rec(rule_name, root)
+
+        # if a rule has a hard feature,
+        # dont consider it easy, and therefore,
+        # don't index any of its features.
+        #
+        # otherwise, its an easy rule, and index its features
+        for rules_with_feature in rules_by_feature.values():
+            rules_with_feature.difference_update(rules_with_hard_features)
+        easy_rules_by_feature = rules_by_feature
+
+        # `rules` is already topologically ordered,
+        # so extract our hard set into the topological ordering.
+        hard_rules = []
+        for rule in rules:
+            if rule.meta["name"] in rules_with_hard_features:
+                hard_rules.append(rule.meta["name"])
+
+        return (easy_rules_by_feature, hard_rules)
+
+    @staticmethod
+    def _get_rules_for_scope(rules, scope) -> List[Rule]:
         """
         given a collection of rules, collect the rules that are needed at the given scope.
         these rules are ordered topologically.
@@ -982,21 +1285,21 @@ class RuleSet:
         don't include auto-generated "subscope" rules.
         we want to include general "lib" rules here - even if they are not dependencies of other rules, see #398
         """
-        scope_rules = set([])
+        scope_rules: Set[Rule] = set([])
 
         # we need to process all rules, not just rules with the given scope.
         # this is because rules with a higher scope, e.g. file scope, may have subscope rules
         #  at lower scope, e.g. function scope.
         # so, we find all dependencies of all rules, and later will filter them down.
         for rule in rules:
-            if rule.meta.get("capa/subscope-rule", False):
+            if rule.is_subscope_rule():
                 continue
 
             scope_rules.update(get_rules_and_dependencies(rules, rule.name))
         return get_rules_with_scope(topologically_order_rules(list(scope_rules)), scope)
 
     @staticmethod
-    def _extract_subscope_rules(rules):
+    def _extract_subscope_rules(rules) -> List[Rule]:
         """
         process the given sequence of rules.
         for each one, extract any embedded subscope rules into their own rule.
@@ -1032,6 +1335,78 @@ class RuleSet:
             for k, v in rule.meta.items():
                 if isinstance(v, str) and tag in v:
                     logger.debug('using rule "%s" and dependencies, found tag in meta.%s: %s', rule.name, k, v)
-                    rules_filtered.update(set(capa.rules.get_rules_and_dependencies(rules, rule.name)))
+                    rules_filtered.update(set(get_rules_and_dependencies(rules, rule.name)))
                     break
+                if isinstance(v, list):
+                    for vv in v:
+                        if tag in vv:
+                            logger.debug('using rule "%s" and dependencies, found tag in meta.%s: %s', rule.name, k, vv)
+                            rules_filtered.update(set(get_rules_and_dependencies(rules, rule.name)))
+                            break
         return RuleSet(list(rules_filtered))
+
+    def match(self, scope: Scope, features: FeatureSet, addr: Address) -> Tuple[FeatureSet, ceng.MatchResults]:
+        """
+        match rules from this ruleset at the given scope against the given features.
+
+        this routine should act just like `capa.engine.match`,
+        except that it may be more performant.
+        """
+        easy_rules_by_feature = {}
+        if scope is Scope.FILE:
+            easy_rules_by_feature = self._easy_file_rules_by_feature
+            hard_rule_names = self._hard_file_rules
+        elif scope is Scope.FUNCTION:
+            easy_rules_by_feature = self._easy_function_rules_by_feature
+            hard_rule_names = self._hard_function_rules
+        elif scope is Scope.BASIC_BLOCK:
+            easy_rules_by_feature = self._easy_basic_block_rules_by_feature
+            hard_rule_names = self._hard_basic_block_rules
+        elif scope is Scope.INSTRUCTION:
+            easy_rules_by_feature = self._easy_instruction_rules_by_feature
+            hard_rule_names = self._hard_instruction_rules
+        else:
+            assert_never(scope)
+
+        candidate_rule_names = set()
+        for feature in features:
+            easy_rule_names = easy_rules_by_feature.get(feature)
+            if easy_rule_names:
+                candidate_rule_names.update(easy_rule_names)
+
+        # first, match against the set of rules that have at least one
+        # feature shared with our feature set.
+        candidate_rules = [self.rules[name] for name in candidate_rule_names]
+        features2, easy_matches = ceng.match(candidate_rules, features, addr)
+
+        # note that we've stored the updated feature set in `features2`.
+        # this contains a superset of the features in `features`;
+        # it contains additional features for any easy rule matches.
+        # we'll pass this feature set to hard rule matching, since one
+        # of those rules might rely on an easy rule match.
+        #
+        # the updated feature set from hard matching will go into `features3`.
+        # this is a superset of `features2` is a superset of `features`.
+        # ultimately, this is what we'll return to the caller.
+        #
+        # in each case, we could have assigned the updated feature set back to `features`,
+        # but this is slightly more explicit how we're tracking the data.
+
+        # now, match against (topologically ordered) list of rules
+        # that we can't really make any guesses about.
+        # these are rules with hard features, like substring/regex/bytes and match statements.
+        hard_rules = [self.rules[name] for name in hard_rule_names]
+        features3, hard_matches = ceng.match(hard_rules, features2, addr)
+
+        # note that above, we probably are skipping matching a bunch of
+        # rules that definitely would never hit.
+        # specifically, "easy rules" that don't share any features with
+        # feature set.
+
+        # MatchResults doesn't technically have an .update() method
+        # but a dict does.
+        matches = {}  # type: ignore
+        matches.update(easy_matches)
+        matches.update(hard_matches)
+
+        return (features3, matches)

capa/rules/cache.py

@@ -0,0 +1,155 @@
+import sys
+import zlib
+import pickle
+import hashlib
+import logging
+import os.path
+from typing import List, Optional
+from dataclasses import dataclass
+
+import capa.rules
+import capa.helpers
+import capa.version
+
+logger = logging.getLogger(__name__)
+
+
+# TypeAlias. note: using `foo: TypeAlias = bar` is Python 3.10+
+CacheIdentifier = str
+
+
+def compute_cache_identifier(rule_content: List[bytes]) -> CacheIdentifier:
+    hash = hashlib.sha256()
+
+    # note that this changes with each release,
+    # so cache identifiers will never collide across releases.
+    version = capa.version.__version__
+
+    hash.update(version.encode("utf-8"))
+    hash.update(b"\x00")
+
+    rule_hashes = list(sorted([hashlib.sha256(buf).hexdigest() for buf in rule_content]))
+    for rule_hash in rule_hashes:
+        hash.update(rule_hash.encode("ascii"))
+        hash.update(b"\x00")
+
+    return hash.hexdigest()
+
+
+def get_default_cache_directory() -> str:
+    # ref: https://github.com/mandiant/capa/issues/1212#issuecomment-1361259813
+    #
+    # Linux:   $XDG_CACHE_HOME/capa/
+    # Windows: %LOCALAPPDATA%\flare\capa\cache
+    # MacOS:   ~/Library/Caches/capa
+
+    # ref: https://stackoverflow.com/a/8220141/87207
+    if sys.platform == "linux" or sys.platform == "linux2":
+        directory = os.environ.get("XDG_CACHE_HOME", os.path.join(os.environ["HOME"], ".cache", "capa"))
+    elif sys.platform == "darwin":
+        directory = os.path.join(os.environ["HOME"], "Library", "Caches", "capa")
+    elif sys.platform == "win32":
+        directory = os.path.join(os.environ["LOCALAPPDATA"], "flare", "capa", "cache")
+    else:
+        raise NotImplementedError(f"unsupported platform: {sys.platform}")
+
+    os.makedirs(directory, exist_ok=True)
+
+    return directory
+
+
+def get_cache_path(cache_dir: str, id: CacheIdentifier) -> str:
+    filename = "capa-" + id[:8] + ".cache"
+    return os.path.join(cache_dir, filename)
+
+
+MAGIC = b"capa"
+VERSION = b"\x00\x00\x00\x01"
+
+
+@dataclass
+class RuleCache:
+    id: CacheIdentifier
+    ruleset: capa.rules.RuleSet
+
+    def dump(self):
+        return MAGIC + VERSION + self.id.encode("ascii") + zlib.compress(pickle.dumps(self))
+
+    @staticmethod
+    def load(data):
+        assert data.startswith(MAGIC + VERSION)
+
+        id = data[0x8:0x48].decode("ascii")
+        cache = pickle.loads(zlib.decompress(data[0x48:]))
+
+        assert isinstance(cache, RuleCache)
+        assert cache.id == id
+
+        return cache
+
+
+def get_ruleset_content(ruleset: capa.rules.RuleSet) -> List[bytes]:
+    rule_contents = []
+    for rule in ruleset.rules.values():
+        if rule.is_subscope_rule():
+            continue
+        rule_contents.append(rule.definition.encode("utf-8"))
+    return rule_contents
+
+
+def compute_ruleset_cache_identifier(ruleset: capa.rules.RuleSet) -> CacheIdentifier:
+    rule_contents = get_ruleset_content(ruleset)
+    return compute_cache_identifier(rule_contents)
+
+
+def cache_ruleset(cache_dir: str, ruleset: capa.rules.RuleSet):
+    """
+    cache the given ruleset to disk, using the given cache directory.
+    this can subsequently be reloaded via `load_cached_ruleset`,
+    assuming the capa version and rule content does not change.
+
+    callers should use this function to avoid the performance overhead
+    of validating rules on each run.
+    """
+    id = compute_ruleset_cache_identifier(ruleset)
+    path = get_cache_path(cache_dir, id)
+    if os.path.exists(path):
+        logger.debug("rule set already cached to %s", path)
+        return
+
+    cache = RuleCache(id, ruleset)
+    with open(path, "wb") as f:
+        f.write(cache.dump())
+
+    logger.debug("rule set cached to %s", path)
+    return
+
+
+def load_cached_ruleset(cache_dir: str, rule_contents: List[bytes]) -> Optional[capa.rules.RuleSet]:
+    """
+    load a cached ruleset from disk, using the given cache directory.
+    the raw rule contents are required here to prove that the rules haven't changed
+    and to avoid stale cache entries.
+
+    callers should use this function to avoid the performance overhead
+    of validating rules on each run.
+    """
+    id = compute_cache_identifier(rule_contents)
+    path = get_cache_path(cache_dir, id)
+    if not os.path.exists(path):
+        logger.debug("rule set cache does not exist: %s", path)
+        return None
+
+    logger.debug("loading rule set from cache: %s", path)
+    with open(path, "rb") as f:
+        buf = f.read()
+
+    try:
+        cache = RuleCache.load(buf)
+    except AssertionError:
+        logger.debug("rule set cache is invalid: %s", path)
+        # delete the cache that seems to be invalid.
+        os.remove(path)
+        return None
+    else:
+        return cache.ruleset

capa/version.py

@@ -1 +1,5 @@
-__version__ = "3.0.2"
+__version__ = "5.0.0"
+
+
+def get_major_version():
+    return int(__version__.partition(".")[0])

doc/installation.md

@@ -1,18 +1,16 @@
 # Installation
-You can install capa in a few different ways. First, if you simply want to use capa, just download the [standalone binary](https://github.com/fireeye/capa/releases). If you want to use capa as a Python library, you can install the package directly from GitHub using `pip`. If you'd like to contribute patches or features to capa, you can work with a local copy of the source code.
+You can install capa in a few different ways. First, if you simply want to use capa, just download the [standalone binary](https://github.com/mandiant/capa/releases). If you want to use capa as a Python library, you can install the package directly from GitHub using `pip`. If you'd like to contribute patches or features to capa, you can work with a local copy of the source code.
 
 ## Method 1: Standalone installation
-If you simply want to use capa, use the standalone binaries we host on GitHub: https://github.com/fireeye/capa/releases. These binary executable files contain all the source code, Python interpreter, and associated resources needed to make capa run. This means you can run it without any installation! Just invoke the file using your terminal shell to see the help documentation.
+If you simply want to use capa, use the standalone binaries we host on GitHub: https://github.com/mandiant/capa/releases. These binary executable files contain all the source code, Python interpreter, and associated resources needed to make capa run. This means you can run it without any installation! Just invoke the file using your terminal shell to see the help documentation.
 
 We use PyInstaller to create these packages.
 
-The capa [README](../README.md#download) also links to nightly builds of standalone binaries from the latest development branch.
-
 ### Linux Standalone installation
 
 The Linux Standalone binary has been built using GLIB 2.26.
-Consequently it works when using GLIB >= 2.26.
-This requirement is satisfied by default in most newer distribution such as Ubuntu >= 18, Debian >= 10, openSUSE >= 15.1 and CentOS >= 8.
+Consequently, it works when using GLIB >= 2.26.
+This requirement is satisfied by default in newer distribution such as Ubuntu >= 18, Debian >= 10, openSUSE >= 15.1 and CentOS >= 8.
 But the binary may not work in older distributions.
 
 ### MacOS Standalone installation
@@ -24,18 +22,27 @@ By default, on MacOS Catalina or greater, Gatekeeper will block execution of the
 ## Method 2: Using capa as a Python library
 To install capa as a Python library use `pip` to fetch the `flare-capa` module.
 
-#### *Note*:
-This method is appropriate for integrating capa in an existing project.
-This technique doesn't pull the default rule set, so you should check it out separately from [capa-rules](https://github.com/fireeye/capa-rules/) and pass the directory to the entrypoint using `-r` or set the rules path in the IDA Pro plugin.
-This technique also doesn't set up the default library identification [signatures](https://github.com/fireeye/capa/tree/master/sigs). You can pass the signature directory using the `-s` argument.
-For example, to run capa with both a rule path and a signature path:
-
-    capa   -r /path/to/capa-rules   -s /path/to/capa-sigs  suspicious.exe
-Alternatively, see Method 3 below.
-
 ### 1. Install capa module
 Use `pip` to install the capa module to your local Python environment. This fetches the library code to your computer but does not keep editable source files around for you to hack on. If you'd like to edit the source files, see below. `$ pip install flare-capa`
 
+#### *Note on capa rules and library identification signatures*
+This method is appropriate for integrating capa in an existing project.
+This technique doesn't pull the default rule set. You can obtain rule releases from [capa-rules](https://github.com/mandiant/capa-rules/releases) and pass the directory to the entrypoint using `-r`. In the IDA Pro plugin you need to configure the rules directory path once.
+
+```console
+$ wget https://github.com/mandiant/capa-rules/archive/refs/tags/v4.0.0.zip
+$ unzip v4.0.0.zip
+$ capa -r /path/to/capa-rules suspicious.exe
+```
+
+This technique also doesn't set up the default library identification [signatures](https://github.com/mandiant/capa/tree/master/sigs). You can pass the signature directory using the `-s` argument.
+For example, to run capa with both a rule path and a signature path:
+```console
+$ capa -s /path/to/capa-sigs suspicious.exe
+```
+
+Alternatively, see Method 3 below.
+
 ### 2. Use capa
 You can now import the `capa` module from a Python script or use the IDA Pro plugins from the `capa/ida` directory. For more information please see the [usage](usage.md) documentation.
 
@@ -43,18 +50,20 @@ You can now import the `capa` module from a Python script or use the IDA Pro plu
 If you'd like to review and modify the capa source code, you'll need to check it out from GitHub and install it locally. By following these instructions, you'll maintain a local directory of source code that you can modify and run easily. 
 
 ### 1. Check out source code
-Next, clone the capa git repository.
-We use submodules to separate [code](https://github.com/fireeye/capa), [rules](https://github.com/fireeye/capa-rules), and [test data](https://github.com/fireeye/capa-testfiles).
-To clone everything use the `--recurse-submodules` option:
-- CAUTION: The capa testfiles repository contains many malware samples. If you pull down everything using this method, you may want to install to a directory that won't trigger your anti-virus software.
-- `$ git clone --recurse-submodules https://github.com/fireeye/capa.git /local/path/to/src` (HTTPS)
-- `$ git clone --recurse-submodules git@github.com:fireeye/capa.git /local/path/to/src` (SSH)
+Clone the capa git repository.
+We use submodules to separate [code](https://github.com/mandiant/capa), [rules](https://github.com/mandiant/capa-rules), and [test data](https://github.com/mandiant/capa-testfiles).
 
-To only get the source code and our provided rules (common), follow these steps:
+To clone everything use the `--recurse-submodules` option:
+- CAUTION: The capa testfiles repository contains many malware samples. If you pull down everything using this method, you may want to install to a directory that is ignored by your anti-virus software.
+- `$ git clone --recurse-submodules https://github.com/mandiant/capa.git /local/path/to/src` (HTTPS)
+- `$ git clone --recurse-submodules git@github.com:mandiant/capa.git /local/path/to/src` (SSH)
+
+To only get the source code and our provided rules (a more common use-case), follow these steps:
 - clone repository
-  - `$ git clone https://github.com/fireeye/capa.git /local/path/to/src` (HTTPS)
-  - `$ git clone git@github.com:fireeye/capa.git /local/path/to/src` (SSH)
+  - `$ git clone https://github.com/mandiant/capa.git /local/path/to/src` (HTTPS)
+  - `$ git clone git@github.com:mandiant/capa.git /local/path/to/src` (SSH)
 - `$ cd /local/path/to/src`
+- initialize the rules submodule and pull rules
 - `$ git submodule update --init rules`
 
 ### 2. Install the local source code
@@ -70,8 +79,7 @@ You'll find that the `capa.exe` (Windows) or `capa` (Linux/MacOS) executables in
 
 For development, we recommend to use [venv](https://docs.python.org/3/tutorial/venv.html). It allows you to create a virtual environment: a self-contained directory tree that contains a Python installation for a particular version of Python, plus a number of additional packages. This approach avoids conflicts between the requirements of different applications on your computer. It also ensures that you don't overlook to add a new requirement to `setup.up` using a library already installed on your system.
 
-To create an environment (in the parent directory, to avoid commiting it by accident or messing with the linters), run:
-`$ python3 -m venv ../capa-env`
+To create an environment (in the parent directory, to avoid committing it by accident or messing with the linters), run: `$ python3 -m venv ../capa-env`
 
 To activate `capa-env` in Linux or MacOS, run:
 `$ source ../capa-env/bin/activate`
@@ -84,10 +92,10 @@ For more details about creating and using virtual environments, check out the [v
 ##### Install development dependencies
 
 We use the following tools to ensure consistent code style and formatting:
-  - [black](https://github.com/psf/black) code formatter, with `-l 120`
-  - [isort 5](https://pypi.org/project/isort/) code formatter, with `--profile black --length-sort --line-width 120`
+  - [black](https://github.com/psf/black) code formatter
+  - [isort 5](https://pypi.org/project/isort/) code formatter
   - [dos2unix](https://linux.die.net/man/1/dos2unix) for UNIX-style LF newlines
-  - [capafmt](https://github.com/fireeye/capa/blob/master/scripts/capafmt.py) rule formatter
+  - [capafmt](https://github.com/mandiant/capa/blob/master/scripts/capafmt.py) rule formatter
 
 To install these development dependencies, run:
 
@@ -98,7 +106,7 @@ You can run it with the argument `no_tests` to skip the tests and only run the c
 
 ##### Setup hooks [optional]
 
-If you plan to contribute to capa, you may want to setup the hooks.
+If you plan to contribute to capa, you may want to setup the provided hooks.
 Run `scripts/setup-hooks.sh` to set the following hooks up:
 - The `pre-commit` hook runs checks before every `git commit`.
   It runs `scripts/ci.sh no_tests` aborting the commit if there are code style or rule linter offenses you need to fix.
@@ -106,13 +114,17 @@ Run `scripts/setup-hooks.sh` to set the following hooks up:
   It runs `scripts/ci.sh` aborting the push if there are code style or rule linter offenses or if the tests fail.
   This way you can ensure everything is alright before sending a pull request.
 
-You can skip the checks by using the `--no-verify` git option.
+You can skip the checks by using the `-n`/`--no-verify` git option.
 
 ### 3. Compile binary using PyInstaller
-We compile capa standalone binaries using PyInstaller. To reproduce the build process check out the source code as described above and follow these steps.
+We compile capa standalone binaries using PyInstaller. To reproduce the build process check out the source code as described above and follow the following steps.
 
 #### Install PyInstaller:
-`$ pip install pyinstaller` (Python 3)
+`$ pip install pyinstaller`
+
+Or install capa with build dependencies:
+
+`$ pip install -e /local/path/to/src[build]`
 
 #### Run Pyinstaller
 `$ pyinstaller .github/pyinstaller/pyinstaller.spec`

doc/limitations.md

@@ -5,7 +5,7 @@ If capa detects that a program may be packed using its rules it warns the user.
 
 
 # Installers, run-time programs, etc.
-capa cannot handle installers, run-time programs like .NET applications, or other packaged applications like AutoIt well. This means that the results may be misleading or incomplete.
+capa cannot handle installers, run-time programs, or other packaged applications like AutoIt well. This means that the results may be misleading or incomplete.
 
 If capa detects an installer, run-time program, etc. it warns the user.
 
@@ -46,6 +46,6 @@ We need more practical use cases and test samples to justify the additional work
 
 
 # ATT&CK, MAEC, MBC, and other capability tagging
-capa uses namespaces to group capabilities (see https://github.com/fireeye/capa-rules/tree/master#namespace-organization).
+capa uses namespaces to group capabilities (see https://github.com/mandiant/capa-rules/tree/master#namespace-organization).
 
-The `rule.meta` field also supports `att&ck`, `mbc`, and `maec` fields to associate rules with the respective taxonomy (see https://github.com/fireeye/capa-rules/blob/master/doc/format.md#meta-block).
+The `rule.meta` field also supports `att&ck`, `mbc`, and `maec` fields to associate rules with the respective taxonomy (see https://github.com/mandiant/capa-rules/blob/master/doc/format.md#meta-block).

doc/release.md

@@ -1,13 +1,13 @@
 # Release checklist
 
-- [ ] Ensure all [milestoned issues/PRs](https://github.com/fireeye/capa/milestones) are addressed, or reassign to a new milestone.
-- [ ] Add the `dont merge` label to all PRs that are close to be ready to merge (or merge them if they are ready) in [capa](https://github.com/fireeye/capa/pulls) and [capa-rules](https://github.com/fireeye/capa-rules/pulls).
-- [ ] Ensure the [CI workflow succeeds in master](https://github.com/fireeye/capa/actions/workflows/tests.yml?query=branch%3Amaster).
-- [ ] Ensure that `python scripts/lint.py rules/ --thorough` succeeds (only `missing examples`  offenses are allowed in the nursery).
+- [ ] Ensure all [milestoned issues/PRs](https://github.com/mandiant/capa/milestones) are addressed, or reassign to a new milestone.
+- [ ] Add the `dont merge` label to all PRs that are close to be ready to merge (or merge them if they are ready) in [capa](https://github.com/mandiant/capa/pulls) and [capa-rules](https://github.com/mandiant/capa-rules/pulls).
+- [ ] Ensure the [CI workflow succeeds in master](https://github.com/mandiant/capa/actions/workflows/tests.yml?query=branch%3Amaster).
+- [ ] Ensure that `python scripts/lint.py rules/ --thorough` succeeds (only `missing examples` offenses are allowed in the nursery). You can [manually trigger a thorough lint](https://github.com/mandiant/capa-rules/actions/workflows/tests.yml) in CI via the "Run workflow" option. 
 - [ ] Review changes
-  - capa https://github.com/fireeye/capa/compare/\<last-release\>...master
-  - capa-rules https://github.com/fireeye/capa-rules/compare/\<last-release>\...master
-- [ ] Update [CHANGELOG.md](https://github.com/fireeye/capa/blob/master/CHANGELOG.md)
+  - capa https://github.com/mandiant/capa/compare/\<last-release\>...master
+  - capa-rules https://github.com/mandiant/capa-rules/compare/\<last-release>\...master
+- [ ] Update [CHANGELOG.md](https://github.com/mandiant/capa/blob/master/CHANGELOG.md)
   - Do not forget to add a nice introduction thanking contributors
   - Remember that we need a major release if we introduce breaking changes
   - Sections: see template below
@@ -31,13 +31,16 @@
     ### Development
 
     ### Raw diffs
-    - [capa <release>...master](https://github.com/fireeye/capa/compare/<release>...master)
-    - [capa-rules <release>...master](https://github.com/fireeye/capa-rules/compare/<release>...master)
+    - [capa <release>...master](https://github.com/mandiant/capa/compare/<release>...master)
+    - [capa-rules <release>...master](https://github.com/mandiant/capa-rules/compare/<release>...master)
     ```
-- [ ] Update [capa/version.py](https://github.com/fireeye/capa/blob/master/capa/version.py)
-- [ ] Create a PR with the updated [CHANGELOG.md](https://github.com/fireeye/capa/blob/master/CHANGELOG.md) and [capa/version.py](https://github.com/fireeye/capa/blob/master/capa/version.py). Copy this checklist in the PR description.
-- [ ] After PR review, merge the PR and [create the release in GH](https://github.com/fireeye/capa/releases/new) using text from the [CHANGELOG.md](https://github.com/fireeye/capa/blob/master/CHANGELOG.md).
-- [ ] Verify GH actions [upload artifacts](https://github.com/fireeye/capa/releases), [publish to PyPI](https://pypi.org/project/flare-capa) and [create a tag in capa rules](https://github.com/fireeye/capa-rules/tags) upon completion.
+- [ ] Update [capa/version.py](https://github.com/mandiant/capa/blob/master/capa/version.py)
+- [ ] Create a PR with the updated [CHANGELOG.md](https://github.com/mandiant/capa/blob/master/CHANGELOG.md) and [capa/version.py](https://github.com/mandiant/capa/blob/master/capa/version.py). Copy this checklist in the PR description.
+- [ ] After PR review, merge the PR and [create the release in GH](https://github.com/mandiant/capa/releases/new) using text from the [CHANGELOG.md](https://github.com/mandiant/capa/blob/master/CHANGELOG.md).
+- Verify GH actions
+  - [ ] [upload artifacts](https://github.com/mandiant/capa/releases)
+  - [ ] [publish to PyPI](https://pypi.org/project/flare-capa)
+  - [ ] [create tag in capa rules](https://github.com/mandiant/capa-rules/tags)
+  - [ ] [create release in capa rules](https://github.com/mandiant/capa-rules/releases)
 - [ ] [Spread the word](https://twitter.com)
 - [ ] Update internal service
-

doc/rules.md

@@ -0,0 +1,43 @@
+### rules
+
+capa uses a collection of rules to identify capabilities within a program.
+The [github.com/mandiant/capa-rules](https://github.com/mandiant/capa-rules) repository contains hundreds of standard library rules that are distributed with capa.
+
+When you download a standalone version of capa, this standard library is embedded within the executable and capa will use these rules by default:
+
+```console
+$ capa suspicious.exe
+```
+
+However, you may want to modify the rules for a variety of reasons:
+
+  - develop new rules to find behaviors,
+  - tweak existing rules to reduce false positives,
+  - collect a private selection of rules not shared publicly.
+
+Or, you may want to use capa as a Python library within another application.
+
+In these scenarios, you must provide the rule set to capa as a directory on your file system. Do this using the `-r`/`--rules` parameter:
+
+```console
+$ capa --rules /local/path/to/rules suspicious.exe
+```
+
+You can download the standard set of rules as ZIP or TGZ archives from the [capa-rules release page](https://github.com/mandiant/capa-rules/releases).
+
+Note that you must use match the rules major version with the capa major version, i.e., use `v1` rules with `v1` of capa.
+This is so that new versions of capa can update rule syntax, such as by adding new fields and logic.
+
+Otherwise, using rules with a mismatched version of capa may lead to errors like:
+
+```
+$ capa --rules /path/to/mismatched/rules suspicious.exe
+ERROR:lint:invalid rule: injection.yml: invalid rule: unexpected statement: instruction
+```                
+
+You can check the version of capa you're currently using like this:
+                         
+```console
+$ capa --version
+capa 3.0.3
+```