freeze: fix schema to support overlapping functions

also, move null extractor to its own namespace

.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
 
@@ -32,9 +32,9 @@ 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)
+  - [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
 
@@ -159,8 +159,8 @@ The process described here has several goals:
 
 Please follow these steps to have your contribution considered by the maintainers:
 
-1. Follow all instructions in [the template](PULL_REQUEST_TEMPLATE.md)
-2. Follow the [styleguides](#styleguides)
+1. Follow the [styleguides](#styleguides)
+2. Update the CHANGELOG and add tests and documentation. In case they are not needed, indicate it in [the PR template](pull_request_template.md).
 3. After you submit your pull request, verify that all [status checks](https://help.github.com/articles/about-status-checks/) are passing <details><summary>What if the status checks are failing? </summary>If a status check is failing, and you believe that the failure is unrelated to your change, please leave a comment on the pull request explaining why you believe the failure is unrelated. A maintainer will re-run the status check for you. If we conclude that the failure was a false positive, then we will open an issue to track that problem with our status check suite.</details>
 
 While the prerequisites above must be satisfied prior to having your pull request reviewed, the reviewer(s) may ask you to complete additional design work, tests, or other changes before your pull request can be ultimately accepted.
@@ -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

@@ -0,0 +1,79 @@
+[mypy]
+
+[mypy-halo.*]
+ignore_missing_imports = True
+
+[mypy-tqdm.*]
+ignore_missing_imports = True
+
+[mypy-ruamel.*]
+ignore_missing_imports = True
+
+[mypy-networkx.*]
+ignore_missing_imports = True
+
+[mypy-pefile.*]
+ignore_missing_imports = True
+
+[mypy-viv_utils.*]
+ignore_missing_imports = True
+
+[mypy-flirt.*]
+ignore_missing_imports = True
+
+[mypy-smda.*]
+ignore_missing_imports = True
+
+[mypy-lief.*]
+ignore_missing_imports = True
+
+[mypy-idc.*]
+ignore_missing_imports = True
+
+[mypy-vivisect.*]
+ignore_missing_imports = True
+
+[mypy-envi.*]
+ignore_missing_imports = True
+
+[mypy-PE.*]
+ignore_missing_imports = True
+
+[mypy-idaapi.*]
+ignore_missing_imports = True
+
+[mypy-idautils.*]
+ignore_missing_imports = True
+
+[mypy-ida_bytes.*]
+ignore_missing_imports = True
+
+[mypy-ida_kernwin.*]
+ignore_missing_imports = True
+
+[mypy-ida_settings.*]
+ignore_missing_imports = True
+
+[mypy-ida_funcs.*]
+ignore_missing_imports = True
+
+[mypy-ida_loader.*]
+ignore_missing_imports = True
+
+[mypy-PyQt5.*]
+ignore_missing_imports = True
+
+[mypy-binaryninja.*]
+ignore_missing_imports = True
+
+[mypy-pytest.*]
+ignore_missing_imports = True
+
+[mypy-devtools.*]
+ignore_missing_imports = True
+
+[mypy-elftools.*]
+ignore_missing_imports = True
+
+[mypy-dncil.*]
+ignore_missing_imports = True

.github/pull_request_template.md

@@ -1,32 +1,22 @@
-<!--
-Thank you for contributing to capa! :heart:
 
-IMPORTANT NOTE
-It's most important that you submit your improvements. So even if you don't use this complete template we look forward to collaborating!
+<!--
+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
 
-PR template based on https://embeddedartistry.com/blog/2017/08/04/a-github-pull-request-template-for-your-projects/
+Please describe the changes in this pull request (PR). Include your motivation and context to help us review.
+
+Please mention the issue your PR addresses (if any):
+closes #issue_number
 -->
 
-### Description
 
-<!-- Please describe the changes in this PR. Including your motivation and context helps us to review. -->
+### Checklist
 
-closes # (issue)
-
-### Type of change
-
-Please update the [CHANGELOG.md](/CHANGELOG.md)
-
-- [ ] Bug fix (non-breaking change which fixes an issue)
-- [ ] New feature (non-breaking change which adds functionality)
-- [ ] Breaking change (fix or feature that would cause existing functionality to not work as expected)
-- [ ] This change requires a documentation update
-  - [ ] I have made the corresponding changes to the documentation
-
-### Tests
-
-- [ ] I have added tests that prove my fix is effective or that my feature works
+<!-- CHANGELOG.md has a `master (unreleased)` section. Please add bug fixes, new features, breaking changes and anything else you think is worthwhile mentioning in the release notes to this file. -->
+- [ ] No CHANGELOG update needed
+<!-- Tests prove that your fix/work as expected and ensure it doesn't break on the feature. -->
 - [ ] No new tests needed
+<!-- Please help us keeping capa documentation up-to-date -->
+- [ ] No documentation update needed

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

@@ -1,4 +1,4 @@
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 import PyInstaller.utils.hooks
 
 # ref: https://groups.google.com/g/pyinstaller/c/amWi0-66uZI/m/miPoKfWjBAAJ

.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
 
@@ -45,8 +45,8 @@ hiddenimports = [
     "vivisect.analysis.crypto",
     "vivisect.analysis.crypto.constants",
     "vivisect.analysis.elf",
-    "vivisect.analysis.elf",
     "vivisect.analysis.elf.elfplt",
+    "vivisect.analysis.elf.elfplt_late",
     "vivisect.analysis.elf.libc_start_main",
     "vivisect.analysis.generic",
     "vivisect.analysis.generic",

.github/pyinstaller/pyinstaller.spec

@@ -1,5 +1,5 @@
 # -*- mode: python -*-
-# Copyright (C) 2020 FireEye, Inc. All Rights Reserved.
+# Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
 import os.path
 import subprocess
 
@@ -33,6 +33,7 @@ a = Analysis(
         # this gets invoked from the directory of the spec file,
         # i.e. ./.github/pyinstaller
         ('../../rules', 'rules'),
+        ('../../sigs', 'sigs'),
 
         # capa.render.default uses tabulate that depends on wcwidth.
         # it seems wcwidth uses a json file `version.json`
@@ -94,7 +95,7 @@ exe = EXE(pyz,
           a.datas,
           exclude_binaries=False,
           name='capa',
-          icon='logo.ico',
+          # TODO not working anymore for unknown reason icon='logo.ico',
           debug=False,
           strip=None,
           upx=True,
@@ -108,5 +109,4 @@ exe = EXE(pyz,
 #               a.datas,
 #               strip=None,
 #               upx=True,
-#               name='capa-dat')
-
+#               name='capa-dat')

.github/workflows/build.yml

@@ -1,6 +1,8 @@
 name: build
 
 on:
+  pull_request:
+    branches: [ master ]
   release:
     types: [edited, published]
 
@@ -9,13 +11,15 @@ jobs:
     name: PyInstaller for ${{ matrix.os }}
     runs-on: ${{ matrix.os }}
     strategy:
+      # set to false for debugging
+      fail-fast: true
       matrix:
         include:
-          - os: ubuntu-16.04
+          - os: ubuntu-18.04
             # use old linux so that the shared library versioning is more portable
             artifact_name: capa
             asset_name: linux
-          - os: windows-2019
+          - os: windows-2022
             artifact_name: capa.exe
             asset_name: windows
           - os: macos-10.15
@@ -26,29 +30,66 @@ jobs:
         uses: actions/checkout@v2
         with:
           submodules: true
-      - name: Set up Python 3.9
+      # 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
         with:
-          python-version: 3.9
-      - if: matrix.os == 'ubuntu-16.04'
+          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: pip install --upgrade pip setuptools
+      - name: Install capa with build requirements
+        run: pip install -e .[build]
       - name: Build standalone executable
-        run: pyinstaller .github/pyinstaller/pyinstaller.spec
-      - name: Does it run?
+        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
         with:
           name: ${{ matrix.asset_name }}
           path: dist/${{ matrix.artifact_name }}
 
-  zip:
-    name: zip ${{ matrix.asset_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]
+    strategy:
+      matrix:
+        include:
+          # OSs not already tested above
+          - os: ubuntu-18.04
+            artifact_name: capa
+            asset_name: linux
+          - os: ubuntu-20.04
+            artifact_name: capa
+            asset_name: linux
+          - os: windows-2022
+            artifact_name: capa.exe
+            asset_name: windows
+    steps:
+      - name: Download ${{ matrix.asset_name }}
+        uses: actions/download-artifact@v2
+        with:
+          name: ${{ matrix.asset_name }}
+      - name: Set executable flag
+        if: matrix.os != 'windows-2022'      
+        run: chmod +x ${{ matrix.artifact_name }}
+      - name: Run capa
+        run: ./${{ matrix.artifact_name }} -h
+
+  zip_and_upload:
+    # upload zipped binaries to Release page
+    if: github.event_name == 'release'
+    name: zip and upload ${{ matrix.asset_name }}
     runs-on: ubuntu-20.04
-    needs: build
+    needs: [build]
     strategy:
       matrix:
         include:

.github/workflows/changelog.yml

@@ -0,0 +1,41 @@
+name: changelog
+
+on:
+  # We need pull_request_target instead of pull_request because a write
+  # repository token is needed to add a review to a PR. DO NOT BUILD
+  # OR RUN UNTRUSTED CODE FROM PRs IN THIS ACTION
+  pull_request_target:
+    types: [opened, edited, synchronize]
+
+jobs:
+  check_changelog:
+    # no need to check for dependency updates via dependabot
+    if: github.actor != 'dependabot[bot]' && github.actor != 'dependabot-preview[bot]'
+    runs-on: ubuntu-20.04
+    env:
+      NO_CHANGELOG: '[x] No CHANGELOG update needed'
+    steps:
+    - name: Get changed files
+      id: files
+      uses: Ana06/get-changed-files@v1.2
+    - name: check changelog updated
+      id: changelog_updated
+      env:
+        PR_BODY: ${{ github.event.pull_request.body }}
+        FILES: ${{ steps.files.outputs.modified }}
+      run: |
+        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
+      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
+      with:
+        repo-token: ${{ secrets.GITHUB_TOKEN }}
+        event: DISMISS
+        body: "CHANGELOG updated or no update needed, thanks! :smile:"

.github/workflows/publish.yml

@@ -15,7 +15,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v2
         with:
-          python-version: '2.7'
+          python-version: '3.7'
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip
@@ -27,3 +27,4 @@ jobs:
         run: |
           python setup.py sdist bdist_wheel
           twine upload --skip-existing dist/*
+

.github/workflows/tag.yml

@@ -12,13 +12,18 @@ jobs:
     - name: Checkout capa-rules
       uses: actions/checkout@v2
       with:
-        repository: fireeye/capa-rules
+        repository: mandiant/capa-rules
         token: ${{ secrets.CAPA_TOKEN }}
     - name: Tag capa-rules
-      run: git tag ${{ github.event.release.tag_name }}
+      run: |
+        # user information is needed to create annotated tags (with a message)
+        git config user.email 'capa-dev@mandiant.com'
+        git config user.name 'Capa Bot'
+        name=${{ github.event.release.tag_name }}
+        git tag $name -m "https://github.com/mandiant/capa/releases/$name"
     - name: Push tag to capa-rules
       uses: ad-m/github-push-action@master
       with:
-        repository: fireeye/capa-rules
+        repository: mandiant/capa-rules
         github_token: ${{ secrets.CAPA_TOKEN }}
         tags: true

.github/workflows/tests.yml

@@ -2,11 +2,27 @@ name: CI
 
 on:
   push:
-    branches: [ master ]
+    branches: [ master, dotnet-main ]
   pull_request:
-    branches: [ master ]
+    # TODO cleanup these
+    branches: [ master, dotnet-main, feature-981 ]
+
+# save workspaces to speed up testing
+env:
+  CAPA_SAVE_WORKSPACE: "True"
 
 jobs:
+  changelog_format:
+    runs-on: ubuntu-20.04
+    steps:
+    - name: Checkout capa
+      uses: actions/checkout@v2
+    # The sync GH action in capa-rules relies on a single '- *$' in the CHANGELOG file
+    - name: Ensure CHANGELOG has '- *$'
+      run: |
+        number=$(grep '\- *$' CHANGELOG.md | wc -l)
+        if [ $number != 1 ]; then exit 1; fi
+
   code_style:
     runs-on: ubuntu-20.04
     steps:
@@ -15,26 +31,29 @@ jobs:
     - name: Set up Python 3.8
       uses: actions/setup-python@v2
       with:
-        python-version: 3.8
+        python-version: "3.8"
     - name: Install dependencies
-      run: pip install 'isort==5.*' black
+      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/
 
   rule_linter:
     runs-on: ubuntu-20.04
     steps:
-    - name: Checkout capa with rules submodule
+    - name: Checkout capa with submodules
       uses: actions/checkout@v2
       with:
-        submodules: true
+        submodules: recursive
     - name: Set up Python 3.8
       uses: actions/setup-python@v2
       with:
-        python-version: 3.8
-    # We don't need vivisect, so we can install capa using Python3
+        python-version: "3.8"
     - name: Install capa
       run: pip install -e .
     - name: Run rule linter
@@ -49,20 +68,18 @@ jobs:
       matrix:
         os: [ubuntu-20.04, windows-2019, macos-10.15]
         # across all operating systems
-        python-version: [3.6, 3.9]
+        python-version: ["3.7", "3.10"]
         include:
           # on Ubuntu run these as well
           - os: ubuntu-20.04
-            python-version: 2.7
+            python-version: "3.8"
           - os: ubuntu-20.04
-            python-version: 3.7
-          - os: ubuntu-20.04
-            python-version: 3.8
+            python-version: "3.9"
     steps:
     - name: Checkout capa with submodules
       uses: actions/checkout@v2
       with:
-        submodules: true
+        submodules: recursive
     - name: Set up Python ${{ matrix.python-version }}
       uses: actions/setup-python@v2
       with:
@@ -70,10 +87,7 @@ jobs:
     - name: Install pyyaml
       if: matrix.os == 'ubuntu-20.04'
       run: sudo apt-get install -y libyaml-dev
-    - name: Install Microsoft Visual C++ 9.0
-      if: matrix.os == 'windows-2019' && matrix.python-version == '2.7'
-      run: choco install vcpython27
     - name: Install capa
       run: pip install -e .[dev]
     - name: Run tests
-      run: pytest tests/
+      run: pytest -v tests/

.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

@@ -4,17 +4,526 @@
 
 ### New Features
 
-### New Rules
+ - 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
+ - render: support Addresses that aren't simple integers, like .NET token+offset #981 @williballenthin
+
+### 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
+
+### New Rules (6)
+
+- 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
+-
 
 ### Bug Fixes
+- improve handling _ prefix compile/link artifact #924 @mike-hunhoff
 
-### Changes
+### capa explorer IDA Pro plugin
+- improve file format extraction #918 @mike-hunhoff
+- remove decorators added by IDA to ELF imports #919 @mike-hunhoff
 
 ### Development
 
 ### Raw diffs
-- [capa v1.6.1...master](https://github.com/fireeye/capa/compare/v1.6.1...master)
-- [capa-rules v1.6.1...master](https://github.com/fireeye/capa-rules/compare/v1.6.1...master)
+- [capa v3.2.0...master](https://github.com/mandiant/capa/compare/v3.2.0...master)
+- [capa-rules v3.2.0...master](https://github.com/mandiant/capa-rules/compare/v3.2.0...master)
+
+## 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.
+  
+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.
+
+### Bug Fixes
+
+- fix bug in PyInstaller config preventing ELF analysis #795 @mr-tz
+
+### Raw diffs
+- [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)
+
+This version updates the version of vivisect used by capa. Users will experience fewer bugs and find improved analysis results.
+
+Thanks to the community for highlighting issues and analysis misses. Your feedback is crucial to further improve capa.
+
+### Bug Fixes
+
+- fix many underlying bugs in vivisect analysis and update to version v1.0.5 #786 @williballenthin
+
+### Raw diffs
+- [capa v3.0.0...v3.0.1](https://github.com/fireeye/capa/compare/v3.0.0...v3.0.1)
+- [capa-rules v3.0.0...v3.0.1](https://github.com/fireeye/capa-rules/compare/v3.0.0...v3.0.1)
+
+## v3.0.0 (2021-09-15)
+
+We are excited to announce version 3.0! :tada:
+
+capa 3.0:
+- adds support for ELF files targeting Linux thanks to [Intezer](https://www.intezer.com/)
+- adds new features to specify OS, CPU architecture, and file format
+- fixes a few bugs that may have led to false negatives (missed capabilities) in older versions
+- adds 80 new rules, including 36 describing techniques for Linux
+
+A huge thanks to everyone who submitted issues, provided feedback, and contributed code and rules.
+Special acknowledgement to @Adir-Shemesh and @TcM1911 of [Intezer](https://www.intezer.com/) for contributing the code to enable ELF support.
+Also, welcome first time contributors:
+  - @jaredscottwilson
+  - @cdong1012
+  - @jlepore-fe 
+
+### New Features
+
+- all: add support for ELF files #700 @Adir-Shemesh @TcM1911
+- rule format: add feature `format: ` for file format, like `format: pe` #723 @williballenthin
+- rule format: add feature `arch: ` for architecture, like `arch: amd64` #723 @williballenthin
+- rule format: add feature `os: ` for operating system, like `os: windows` #723 @williballenthin
+- rule format: add feature `substring: ` for verbatim strings with leading/trailing wildcards #737 @williballenthin
+- scripts: add `profile-memory.py` for profiling memory usage #736 @williballenthin
+- main: add light weight ELF file feature extractor to detect file limitations #770 @mr-tz
+
+### Breaking Changes
+
+- rules using `format`, `arch`, `os`, or `substring` features cannot be used by capa versions prior to v3
+- legacy term `arch` (i.e., "x32") is now called `bitness` @williballenthin
+- freeze format gains new section for "global" features #759 @williballenthin
+
+### New Rules (80)
+
+- collection/webcam/capture-webcam-image @johnk3r
+- 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@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
+- c2/shell/create-unix-reverse-shell joakim@intezer.com
+- c2/shell/execute-shell-command-received-from-socket joakim@intezer.com
+- collection/get-current-user joakim@intezer.com
+- host-interaction/file-system/change-file-permission joakim@intezer.com
+- host-interaction/hardware/memory/get-memory-information joakim@intezer.com
+- host-interaction/mutex/lock-file joakim@intezer.com
+- host-interaction/os/version/get-kernel-version joakim@intezer.com
+- host-interaction/os/version/get-linux-distribution joakim@intezer.com
+- host-interaction/process/terminate/terminate-process-via-kill joakim@intezer.com
+- lib/duplicate-stdin-and-stdout joakim@intezer.com
+- nursery/capture-network-configuration-via-ifconfig joakim@intezeer.com
+- nursery/collect-ssh-keys joakim@intezer.com
+- nursery/enumerate-processes-via-procfs joakim@intezer.com
+- nursery/interact-with-iptables joakim@intezer.com
+- persistence/persist-via-desktop-autostart joakim@intezer.com
+- 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@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@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
+- nursery/write-file-on-linux joakim@intezer.com
+- communication/socket/tcp/send/obtain-transmitpackets-callback-function-via-wsaioctl jonathan.lepore@mandiant.com
+- nursery/linked-against-cpp-http-library @mr-tz
+- nursery/linked-against-cpp-json-library @mr-tz
+
+### Bug Fixes
+
+- main: fix `KeyError: 0` when reporting results @williballehtin #703
+- main: fix potential false negatives due to namespaces across scopes @williballenthin #721
+- linter: suppress some warnings about imports from ntdll/ntoskrnl @williballenthin #743
+- linter: suppress some warnings about missing examples in the nursery @williballenthin #747
+
+### capa explorer IDA Pro plugin
+
+- explorer: add additional filter logic when displaying matches by function #686 @mike-hunhoff
+- explorer: remove duplicate check when saving file #687 @mike-hunhoff
+- explorer: update IDA extractor to use non-canon mnemonics #688 @mike-hunhoff
+- explorer: allow user to add specified number of bytes when adding a Bytes feature in the Rule Generator #689 @mike-hunhoff
+- explorer: enforce max column width Features and Editor panes #691 @mike-hunhoff
+- explorer: add option to limit features to currently selected disassembly address #692 @mike-hunhoff
+- explorer: update support documentation and runtime checks #741 @mike-hunhoff
+- explorer: small performance boost to rule generator search functionality #742 @mike-hunhoff
+- explorer: add support for arch, os, and format features #758 @mike-hunhoff
+- explorer: improve parsing algorithm for rule generator feature editor #768 @mike-hunhoff
+
+### Development
+
+### Raw diffs
+- [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)
+
+We are excited to announce version 2.0! :tada:
+capa 2.0:
+- enables anyone to contribute rules more easily
+- is the first Python 3 ONLY version
+- provides more concise and relevant result via identification of library functions using FLIRT
+  ![capa v2.0 results ignoring library code functions](doc/img/changelog/flirt-ignore.png)
+- includes many features and enhancements for the capa explorer IDA plugin
+- adds 93 new rules, including all new techniques introduced in MITRE ATT&CK v9
+
+A huge thanks to everyone who submitted issues, provided feedback, and contributed code and rules. Many colleagues across dozens of organizations have volunteered their experience to improve this tool! :heart:
+
+
+### New Features
+
+- 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
+- rule: add file-scope feature (`function-name`) for recognized library functions #567 @williballenthin
+- main: auto detect shellcode based on file extension #516 @mr-tz
+- main: more detailed progress bar output when matching functions #562 @mr-tz
+- main: detect file limitations without doing code analysis for better performance #583 @williballenthin
+- show-features: don't show features from library functions #569 @williballenthin
+- linter: summarize results at the end #571 @williballenthin
+- linter: check for `or` with always true child statement, e.g. `optional`, colors #348 @mr-tz
+
+### Breaking Changes
+
+- py3: drop Python 2 support #480 @Ana06
+- meta: added `library_functions` field, `feature_counts.functions` does not include library functions any more #562 @mr-tz
+- json: results document now contains parsed ATT&CK and MBC fields instead of canonical representation #526 @mr-tz
+- json: record all matching strings for regex #159 @williballenthin
+- main: implement file limitations via rules not code #390 @williballenthin
+- json: correctly render negative offsets #619 @williballenthin
+- library: remove logic from `__init__.py` throughout #622 @williballenthin
+
+### New Rules (93)
+
+- anti-analysis/packer/amber/packed-with-amber @gormaniac
+- collection/file-managers/gather-3d-ftp-information @re-fox
+- collection/file-managers/gather-alftp-information @re-fox
+- collection/file-managers/gather-bitkinex-information @re-fox
+- collection/file-managers/gather-blazeftp-information @re-fox
+- collection/file-managers/gather-bulletproof-ftp-information @re-fox
+- collection/file-managers/gather-classicftp-information @re-fox
+- collection/file-managers/gather-coreftp-information @re-fox
+- collection/file-managers/gather-cuteftp-information @re-fox
+- collection/file-managers/gather-cyberduck-information @re-fox
+- collection/file-managers/gather-direct-ftp-information @re-fox
+- collection/file-managers/gather-directory-opus-information @re-fox
+- collection/file-managers/gather-expandrive-information @re-fox
+- collection/file-managers/gather-faststone-browser-information @re-fox
+- collection/file-managers/gather-fasttrack-ftp-information @re-fox
+- collection/file-managers/gather-ffftp-information @re-fox
+- collection/file-managers/gather-filezilla-information @re-fox
+- collection/file-managers/gather-flashfxp-information @re-fox
+- collection/file-managers/gather-fling-ftp-information @re-fox
+- collection/file-managers/gather-freshftp-information @re-fox
+- collection/file-managers/gather-frigate3-information @re-fox
+- collection/file-managers/gather-ftp-commander-information @re-fox
+- collection/file-managers/gather-ftp-explorer-information @re-fox
+- collection/file-managers/gather-ftp-voyager-information @re-fox
+- collection/file-managers/gather-ftpgetter-information @re-fox
+- collection/file-managers/gather-ftpinfo-information @re-fox
+- collection/file-managers/gather-ftpnow-information @re-fox
+- collection/file-managers/gather-ftprush-information @re-fox
+- collection/file-managers/gather-ftpshell-information @re-fox
+- collection/file-managers/gather-global-downloader-information @re-fox
+- collection/file-managers/gather-goftp-information @re-fox
+- collection/file-managers/gather-leapftp-information @re-fox
+- collection/file-managers/gather-netdrive-information @re-fox
+- collection/file-managers/gather-nexusfile-information @re-fox
+- collection/file-managers/gather-nova-ftp-information @re-fox
+- collection/file-managers/gather-robo-ftp-information @re-fox
+- collection/file-managers/gather-securefx-information @re-fox
+- collection/file-managers/gather-smart-ftp-information @re-fox
+- collection/file-managers/gather-softx-ftp-information @re-fox
+- collection/file-managers/gather-southriver-webdrive-information @re-fox
+- collection/file-managers/gather-staff-ftp-information @re-fox
+- collection/file-managers/gather-total-commander-information @re-fox
+- collection/file-managers/gather-turbo-ftp-information @re-fox
+- collection/file-managers/gather-ultrafxp-information @re-fox
+- collection/file-managers/gather-winscp-information @re-fox
+- collection/file-managers/gather-winzip-information @re-fox
+- collection/file-managers/gather-wise-ftp-information @re-fox
+- collection/file-managers/gather-ws-ftp-information @re-fox
+- collection/file-managers/gather-xftp-information @re-fox
+- data-manipulation/compression/decompress-data-using-aplib @r3c0nst @mr-tz
+- host-interaction/bootloader/disable-code-signing @williballenthin
+- host-interaction/bootloader/manipulate-boot-configuration @williballenthin
+- host-interaction/driver/disable-driver-code-integrity @williballenthin
+- host-interaction/file-system/bypass-mark-of-the-web @williballenthin
+- host-interaction/network/domain/get-domain-information @recvfrom
+- host-interaction/session/get-logon-sessions @recvfrom
+- linking/runtime-linking/resolve-function-by-fin8-fasthash @r3c0nst @mr-tz
+- nursery/build-docker-image @williballenthin
+- nursery/create-container @williballenthin
+- nursery/encrypt-data-using-fakem-cipher @mike-hunhoff
+- nursery/list-containers @williballenthin
+- 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@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@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@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@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@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
+
+- build: use Python 3.8 for PyInstaller to support consistently running across multiple operating systems including Windows 7 #505 @mr-tz
+- main: correctly match BB-scope matches at file scope #605 @williballenthin
+- main: do not process non-PE files even when --format explicitly provided #664 @mr-tz
+
+### capa explorer IDA Pro plugin
+- explorer: IDA 7.6 support #497 @williballenthin
+- explorer: explain how to install IDA 7.6 patch to enable the plugin #528 @williballenthin
+- explorer: document IDA 7.6sp1 as alternative to the patch #536 @Ana06
+- explorer: add support for function-name feature #618 @mike-hunhoff
+- explorer: circular import workaround #654 @mike-hunhoff
+- explorer: add argument to control whether to automatically analyze when running capa explorer #548 @Ana06
+- explorer: extract API features via function names recognized by IDA/FLIRT #661 @mr-tz
+
+### Development
+
+- ci: add capa release link to capa-rules tag #517 @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
+
+### 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/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)
+
+This release adds IDA 7.6 support to capa.
+
+### Changes
+
+- IDA 7.6 support @williballenthin @Ana06
+
+### Raw diffs
+
+  - [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)
+
+This release backports a fix to capa 1.6: The Windows binary was built with Python 3.9 which doesn't support Windows 7.
+
+### Bug Fixes
+
+- build: use Python 3.8 for PyInstaller to support consistently running across multiple operating systems including Windows 7 @mr-tz @Ana06
+
+### Raw diffs
+
+  - [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)
@@ -82,8 +591,8 @@ 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)
@@ -92,7 +601,7 @@ 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)
 
@@ -154,8 +663,8 @@ 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)
@@ -168,8 +677,8 @@ 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)
@@ -184,7 +693,7 @@ 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
@@ -267,8 +776,8 @@ 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)
 
@@ -280,8 +789,8 @@ 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)
 
@@ -292,7 +801,7 @@ 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
 
@@ -395,8 +904,8 @@ 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)
 
@@ -410,7 +919,7 @@ 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
 
@@ -420,9 +929,9 @@ 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.
  
@@ -467,8 +976,8 @@ 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)
 
@@ -484,9 +993,9 @@ 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
 
@@ -565,8 +1074,8 @@ 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)
 
@@ -579,7 +1088,7 @@ 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
 
@@ -652,5 +1161,5 @@ Download a standalone binary below and checkout the readme [here on GitHub](http
 
 ### 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

@@ -1,17 +1,20 @@
-![capa](.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-485-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-665-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 file or shellcode and it tells you what it thinks the program can do.
+You run it against a PE, ELF, 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).
+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)
 
 ```
 $ capa.exe suspicious.exe
@@ -63,18 +66,11 @@ $ 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.
 
-<!--
-Alternatively, you can fetch a nightly build of a standalone binary from one of the following links. These are built using the latest development branch.
-- Windows 64bit: TODO
-- Linux: TODO
-- OSX: TODO
--->
+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.
 
-To use capa as a library or integrate with another tool, see [doc/installation.md](doc/installation.md) for further setup instructions.
-
-For more information about how to use capa, see [doc/usage.md](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
 
@@ -91,11 +87,11 @@ This is useful for at least two reasons:
   - it shows where within the binary an experienced analyst might study with IDA Pro
 
 ```
-λ capa.exe suspicious.exe -vv
+$ 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
@@ -131,7 +127,7 @@ rule:
   meta:
     name: hash data with CRC32
     namespace: data-manipulation/checksum/crc32
-    author: moritz.raabe@fireeye.com
+    author: moritz.raabe@mandiant.com
     scope: function
     examples:
       - 2D3EDC218A90F03089CC01715A9F047F:0x403CBD
@@ -146,21 +142,24 @@ rule:
       - 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](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](doc/img/explorer_expanded.png)
+![capa + IDA Pro integration](https://github.com/mandiant/capa/blob/master/doc/img/explorer_expanded.png)
 
 # further information
 ## capa
-- [doc/installation](doc/installation.md)
-- [doc/usage](doc/usage.md)
-- [doc/limitations](doc/limitations.md)
-- [Contributing Guide](.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/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 +8,29 @@
 
 import copy
 import collections
+from typing import TYPE_CHECKING, Set, Dict, List, Tuple, Mapping, Iterable
 
-import capa.features
+import capa.perf
+import capa.features.common
+from capa.features.common import Result, Feature
+from capa.features.address import Address
+
+if TYPE_CHECKING:
+    # circular import, otherwise
+    import capa.rules
 
 
-class Statement(object):
+# a collection of features and the locations at which they are found.
+#
+# used throughout matching as the context in which features are searched:
+# to check if a feature exists, do: `Number(0x10) in features`.
+# 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[Address]]
+
+
+class Statement:
     """
     superclass for structural nodes, such as and/or/not.
     this exists to provide a default impl for `__str__` and `__repr__`,
@@ -33,15 +51,12 @@ class Statement(object):
     def __repr__(self):
         return str(self)
 
-    def evaluate(self, ctx):
+    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()
 
@@ -50,7 +65,7 @@ class Statement(object):
             yield self.child
 
         if hasattr(self, "children"):
-            for child in self.children:
+            for child in getattr(self, "children"):
                 yield child
 
     def replace_child(self, existing, new):
@@ -59,75 +74,76 @@ class Statement(object):
                 self.child = new
 
         if hasattr(self, "children"):
-            for i, child in enumerate(self.children):
+            children = getattr(self, "children")
+            for i, child in enumerate(children):
                 if child is existing:
-                    self.children[i] = new
-
-
-class Result(object):
-    """
-    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, statement, children, 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
+                    children[i] = new
 
 
 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)
         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)
         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):
@@ -137,28 +153,55 @@ class Not(Statement):
         super(Not, self).__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)
         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):
@@ -170,7 +213,10 @@ class Range(Statement):
         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, [])
@@ -190,59 +236,66 @@ 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(Subscope, self).__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!")
 
 
-def topologically_order_rules(rules):
+# mapping from rule name to list of: (location of match, result object)
+#
+# used throughout matching and rendering to collection the results
+#  of statement evaluation and their locations.
+#
+# to check if a rule matched, do: `"TCP client" in matches`.
+# to find where a rule matched, do: `map(first, matches["TCP client"])`
+# to see how a rule matched, do:
+#
+#     for address, match_details in matches["TCP client"]:
+#         inspect(match_details)
+#
+# aliased here so that the type can be documented and xref'd.
+MatchResults = Mapping[str, List[Tuple[Address, Result]]]
+
+
+def index_rule_matches(features: FeatureSet, rule: "capa.rules.Rule", locations: Iterable[Address]):
     """
-    order the given rules such that dependencies show up before dependents.
-    this means that as we match rules, we can add features for the matches, and these
-     will be matched by subsequent rules if they follow this order.
+    record into the given featureset that the given rule matched at the given locations.
 
-    assumes that the rule dependency graph is a DAG.
+    naively, this is just adding a MatchedRule feature;
+    however, we also want to record matches for the rule's namespaces.
+
+    updates `features` in-place. doesn't modify the remaining arguments.
     """
-    # we evaluate `rules` multiple times, so if its a generator, realize it into a list.
-    rules = list(rules)
-    namespaces = capa.rules.index_rules_by_namespace(rules)
-    rules = {rule.name: rule for rule in rules}
-    seen = set([])
-    ret = []
-
-    def rec(rule):
-        if rule.name in seen:
-            return
-
-        for dep in rule.get_dependencies(namespaces):
-            rec(rules[dep])
-
-        ret.append(rule)
-        seen.add(rule.name)
-
-    for rule in rules.values():
-        rec(rule)
-
-    return ret
+    features[capa.features.common.MatchedRule(rule.name)].update(locations)
+    namespace = rule.meta.get("namespace")
+    if namespace:
+        while namespace:
+            features[capa.features.common.MatchedRule(namespace)].update(locations)
+            namespace, _, _ = namespace.rpartition("/")
 
 
-def match(rules, features, va):
+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[List[capa.features.Feature], Dict[str, Tuple[int, capa.engine.Result]]]: two-tuple with entries:
-        - list of features used for matching (which may be greater than 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)
+    results = collections.defaultdict(list)  # type: MatchResults
 
     # copy features so that we can modify it
     # without affecting the caller (keep this function pure)
@@ -251,15 +304,22 @@ def match(rules, features, va):
     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))
-            features[capa.features.MatchedRule(rule.name)].add(va)
+            # 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)
 
-            namespace = rule.meta.get("namespace")
-            if namespace:
-                while namespace:
-                    features[capa.features.MatchedRule(namespace)].add(va)
-                    namespace, _, _ = namespace.rpartition("/")
+            # 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, [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/__init__.py

@@ -1,236 +0,0 @@
-# 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 re
-import sys
-import codecs
-import logging
-
-import capa.engine
-
-logger = logging.getLogger(__name__)
-MAX_BYTES_FEATURE_SIZE = 0x100
-
-# thunks may be chained so we specify a delta to control the depth to which these chains are explored
-THUNK_CHAIN_DEPTH_DELTA = 5
-
-# identifiers for supported architectures names that tweak a feature
-# for example, offset/x32
-ARCH_X32 = "x32"
-ARCH_X64 = "x64"
-VALID_ARCH = (ARCH_X32, ARCH_X64)
-
-
-def bytes_to_str(b):
-    if sys.version_info[0] >= 3:
-        return str(codecs.encode(b, "hex").decode("utf-8"))
-    else:
-        return codecs.encode(b, "hex")
-
-
-def hex_string(h):
-    """ render hex string e.g. "0a40b1" as "0A 40 B1" """
-    return " ".join(h[i : i + 2] for i in range(0, len(h), 2)).upper()
-
-
-def escape_string(s):
-    """escape special characters"""
-    s = repr(s)
-    if not s.startswith(('"', "'")):
-        # u'hello\r\nworld' -> hello\\r\\nworld
-        s = s[2:-1]
-    else:
-        # 'hello\r\nworld' -> hello\\r\\nworld
-        s = s[1:-1]
-    s = s.replace("\\'", "'")  # repr() may escape "'" in some edge cases, remove
-    s = s.replace('"', '\\"')  # repr() does not escape '"', add
-    return s
-
-
-class Feature(object):
-    def __init__(self, value, arch=None, description=None):
-        """
-        Args:
-          value (any): the value of the feature, such as the number or string.
-          arch (str): one of the VALID_ARCH values, or None.
-            When None, then the feature applies to any architecture.
-            Modifies the feature name from `feature` to `feature/arch`, like `offset/x32`.
-          description (str): a human-readable description that explains the feature value.
-        """
-        super(Feature, self).__init__()
-
-        if arch is not None:
-            if arch not in VALID_ARCH:
-                raise ValueError("arch '%s' must be one of %s" % (arch, VALID_ARCH))
-            self.name = self.__class__.__name__.lower() + "/" + arch
-        else:
-            self.name = self.__class__.__name__.lower()
-
-        self.value = value
-        self.arch = arch
-        self.description = description
-
-    def __hash__(self):
-        return hash((self.name, self.value, self.arch))
-
-    def __eq__(self, other):
-        return self.name == other.name and self.value == other.value and self.arch == other.arch
-
-    def get_value_str(self):
-        """
-        render the value of this feature, for use by `__str__` and friends.
-        subclasses should override to customize the rendering.
-
-        Returns: any
-        """
-        return 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)
-            else:
-                return "%s(%s)" % (self.name, self.get_value_str())
-        else:
-            return "%s" % self.name
-
-    def __repr__(self):
-        return str(self)
-
-    def evaluate(self, ctx):
-        return capa.engine.Result(self in ctx, self, [], locations=ctx.get(self, []))
-
-    def freeze_serialize(self):
-        if self.arch is not None:
-            return (self.__class__.__name__, [self.value, {"arch": self.arch}])
-        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)
-
-
-class MatchedRule(Feature):
-    def __init__(self, value, description=None):
-        super(MatchedRule, self).__init__(value, description=description)
-        self.name = "match"
-
-
-class Characteristic(Feature):
-    def __init__(self, value, description=None):
-        super(Characteristic, self).__init__(value, description=description)
-
-
-class String(Feature):
-    def __init__(self, value, description=None):
-        super(String, self).__init__(value, description=description)
-
-
-class Regex(String):
-    def __init__(self, value, description=None):
-        super(Regex, self).__init__(value, description=description)
-        pat = self.value[len("/") : -len("/")]
-        flags = re.DOTALL
-        if value.endswith("/i"):
-            pat = self.value[len("/") : -len("/i")]
-            flags |= re.IGNORECASE
-        try:
-            self.re = re.compile(pat, flags)
-        except re.error:
-            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
-            )
-
-    def evaluate(self, ctx):
-        for feature, locations in ctx.items():
-            if not isinstance(feature, (capa.features.String,)):
-                continue
-
-            # `re.search` finds a match anywhere in the given string
-            # which implies leading and/or trailing whitespace.
-            # 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):
-                # 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 value.
-                # see #262.
-                return capa.engine.Result(True, _MatchedRegex(self, feature.value), [], locations=locations)
-
-        return capa.engine.Result(False, _MatchedRegex(self, None), [])
-
-    def __str__(self):
-        return "regex(string =~ %s)" % self.value
-
-
-class _MatchedRegex(Regex):
-    """
-    this represents a specific instance of a regular expression feature match.
-    treat it the same as a `Regex` except it has the `match` field that contains the complete string that matched.
-
-    note: this type should only ever be constructed by `Regex.evaluate()`. it is not part of the public API.
-    """
-
-    def __init__(self, regex, match):
-        """
-        args:
-          regex (Regex): the regex feature that matches
-          match (string|None): the matching string or None if it doesn't match
-        """
-        super(_MatchedRegex, self).__init__(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"
-        # this may be None if the regex doesn't match
-        self.match = match
-
-    def __str__(self):
-        return 'regex(string =~ %s, matched = "%s")' % (self.value, self.match)
-
-
-class StringFactory(object):
-    def __new__(self, value, description=None):
-        if value.startswith("/") and (value.endswith("/") or value.endswith("/i")):
-            return Regex(value, description=description)
-        return String(value, description=description)
-
-
-class Bytes(Feature):
-    def __init__(self, value, description=None):
-        super(Bytes, self).__init__(value, description=description)
-
-    def evaluate(self, ctx):
-        for feature, locations in ctx.items():
-            if not isinstance(feature, (capa.features.Bytes,)):
-                continue
-
-            if feature.value.startswith(self.value):
-                return capa.engine.Result(True, self, [], locations=locations)
-
-        return capa.engine.Result(False, self, [])
-
-    def get_value_str(self):
-        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])

capa/features/address.py

@@ -0,0 +1,96 @@
+import abc
+
+from dncil.clr.token import Token
+
+
+class Address(abc.ABC):
+    @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 __str__(self):
+        # implement str so the address can be rendered in capa output
+        ...
+
+
+class AbsoluteVirtualAddress(int, Address):
+    """an absolute memory address"""
+
+    def __new__(cls, v):
+        assert v >= 0
+        return int.__new__(cls, v)
+
+    def __str__(self):
+        return str(self)
+
+
+class RelativeVirtualAddress(int, Address):
+    """a memory address relative to a base address"""
+
+    pass
+
+
+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 __str__(self):
+        return str(self)
+
+
+class DNTokenAddress(Address):
+    """a .NET token"""
+
+    def __init__(self, token: Token):
+        self.token = token
+
+    def __lt__(self, other):
+        return self.token.value < other.token.value
+
+    def __hash__(self):
+        return hash(self.token.value)
+
+    def __str__(self):
+        return str(self.token)
+
+
+class DNTokenOffsetAddress(Address):
+    """an offset into an object specified by a .NET token"""
+
+    def __init__(self, token: Token, offset: int):
+        assert offset >= 0
+        self.token = token
+        self.offset = offset
+
+    def __lt__(self, other):
+        return (self.token.value, self.offset) < (other.token.value, other.offset)
+
+    def __hash__(self):
+        return hash((self.token.value, self.offset))
+
+    def __str__(self):
+        return f"{self.token:s}+{hex(self.offset)}"
+
+
+class _NoAddress(Address):
+    def __lt__(self, other):
+        return False
+
+    def __hash__(self):
+        return hash(0)
+
+    def __str__(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
@@ -6,7 +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.
 
-from capa.features import Feature
+from capa.features.common import Feature
 
 
 class BasicBlock(Feature):
@@ -20,7 +20,7 @@ class BasicBlock(Feature):
         return ""
 
     def freeze_serialize(self):
-        return (self.__class__.__name__, [])
+        return (self.__class__.__name__, ())
 
     @classmethod
     def freeze_deserialize(cls, args):

capa/features/common.py

@@ -0,0 +1,439 @@
+# 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 re
+import abc
+import codecs
+import logging
+import collections
+from typing import TYPE_CHECKING, Set, Dict, List, Union, Optional, Sequence
+
+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
+
+# thunks may be chained so we specify a delta to control the depth to which these chains are explored
+THUNK_CHAIN_DEPTH_DELTA = 5
+
+
+def bytes_to_str(b: bytes) -> str:
+    return str(codecs.encode(b, "hex").decode("utf-8"))
+
+
+def hex_string(h: str) -> str:
+    """render hex string e.g. "0a40b1" as "0A 40 B1" """
+    return " ".join(h[i : i + 2] for i in range(0, len(h), 2)).upper()
+
+
+def escape_string(s: str) -> str:
+    """escape special characters"""
+    s = repr(s)
+    if not s.startswith(('"', "'")):
+        # u'hello\r\nworld' -> hello\\r\\nworld
+        s = s[2:-1]
+    else:
+        # 'hello\r\nworld' -> hello\\r\\nworld
+        s = s[1:-1]
+    s = s.replace("\\'", "'")  # repr() may escape "'" in some edge cases, remove
+    s = s.replace('"', '\\"')  # repr() does not escape '"', add
+    return s
+
+
+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(Result, self).__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=None):
+        """
+        Args:
+          value (any): the value of the feature, such as the number or string.
+          description (str): a human-readable description that explains the feature value.
+        """
+        super(Feature, self).__init__()
+        self.name = self.__class__.__name__.lower()
+
+        self.value = value
+        self.description = description
+
+    def __hash__(self):
+        return hash((self.name, self.value))
+
+    def __lt__(self, other):
+        return self.freeze_serialize() < other.freeze_serialize()
+
+    def __eq__(self, other):
+        return self.name == other.name and self.value == other.value
+
+    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)
+            else:
+                return "%s(%s)" % (self.name, self.get_value_str())
+        else:
+            return "%s" % self.name
+
+    def __repr__(self):
+        return str(self)
+
+    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()))
+
+    def freeze_serialize(self):
+        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)
+        else:
+            return cls(*args)
+
+
+class MatchedRule(Feature):
+    def __init__(self, value: str, description=None):
+        super(MatchedRule, self).__init__(value, description=description)
+        self.name = "match"
+
+
+class Characteristic(Feature):
+    def __init__(self, value: str, description=None):
+        super(Characteristic, self).__init__(value, description=description)
+
+
+class String(Feature):
+    def __init__(self, value: str, description=None):
+        super(String, self).__init__(value, description=description)
+
+
+class Substring(String):
+    def __init__(self, value: str, description=None):
+        super(Substring, self).__init__(value, description=description)
+        self.value = value
+
+    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)
+
+        for feature, locations in ctx.items():
+            if not isinstance(feature, (String,)):
+                continue
+
+            if not isinstance(feature.value, str):
+                # this is a programming error: String should only contain str
+                raise ValueError("unexpected feature value type")
+
+            if self.value in feature.value:
+                matches[feature.value].extend(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])
+
+            # 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 Result(True, _MatchedSubstring(self, matches), [], locations=locations)
+        else:
+            return Result(False, _MatchedSubstring(self, None), [])
+
+    def __str__(self):
+        return "substring(%s)" % self.value
+
+
+class _MatchedSubstring(Substring):
+    """
+    this represents specific match instances of a substring feature.
+    treat it the same as a `Substring` except it has the `matches` field that contains the complete strings that matched.
+
+    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):
+        """
+        args:
+          substring (Substring): the substring feature that matches.
+          match (Dict[string, List[int]]|None): mapping from matching string to its locations.
+        """
+        super(_MatchedSubstring, self).__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"
+        # this may be None if the substring doesn't match
+        self.matches = matches
+
+    def __str__(self):
+        return 'substring("%s", matches = %s)' % (
+            self.value,
+            ", ".join(map(lambda s: '"' + s + '"', (self.matches or {}).keys())),
+        )
+
+
+class Regex(String):
+    def __init__(self, value: str, description=None):
+        super(Regex, self).__init__(value, description=description)
+        self.value = value
+
+        pat = self.value[len("/") : -len("/")]
+        flags = re.DOTALL
+        if value.endswith("/i"):
+            pat = self.value[len("/") : -len("/i")]
+            flags |= re.IGNORECASE
+        try:
+            self.re = re.compile(pat, flags)
+        except re.error:
+            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
+            )
+
+    def evaluate(self, ctx, short_circuit=True):
+        capa.perf.counters["evaluate.feature"] += 1
+        capa.perf.counters["evaluate.feature.regex"] += 1
+
+        # mapping from string value to list of locations.
+        # will unique the locations later on.
+        matches = collections.defaultdict(list)
+
+        for feature, locations in ctx.items():
+            if not isinstance(feature, (String,)):
+                continue
+
+            if not isinstance(feature.value, str):
+                # this is a programming error: String should only contain str
+                raise ValueError("unexpected feature value type")
+
+            # `re.search` finds a match anywhere in the given string
+            # which implies leading and/or trailing whitespace.
+            # 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)
+                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])
+
+            # 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 Result(True, _MatchedRegex(self, matches), [], locations=locations)
+        else:
+            return Result(False, _MatchedRegex(self, None), [])
+
+    def __str__(self):
+        return "regex(string =~ %s)" % self.value
+
+
+class _MatchedRegex(Regex):
+    """
+    this represents specific match instances of a regular expression feature.
+    treat it the same as a `Regex` except it has the `matches` field that contains the complete strings that matched.
+
+    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):
+        """
+        args:
+          regex (Regex): the regex feature that matches.
+          match (Dict[string, List[int]]|None): mapping from matching string to its locations.
+        """
+        super(_MatchedRegex, self).__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"
+        # this may be None if the regex doesn't match
+        self.matches = matches
+
+    def __str__(self):
+        return "regex(string =~ %s, matches = %s)" % (
+            self.value,
+            ", ".join(map(lambda s: '"' + s + '"', (self.matches or {}).keys())),
+        )
+
+
+class StringFactory:
+    def __new__(cls, value: str, description=None):
+        if value.startswith("/") and (value.endswith("/") or value.endswith("/i")):
+            return Regex(value, description=description)
+        return String(value, description=description)
+
+
+class Bytes(Feature):
+    def __init__(self, value: bytes, description=None):
+        super(Bytes, self).__init__(value, description=description)
+        self.value = value
+
+    def evaluate(self, ctx, **kwargs):
+        capa.perf.counters["evaluate.feature"] += 1
+        capa.perf.counters["evaluate.feature.bytes"] += 1
+
+        for feature, locations in ctx.items():
+            if not isinstance(feature, (Bytes,)):
+                continue
+
+            if feature.value.startswith(self.value):
+                return Result(True, self, [], locations=locations)
+
+        return Result(False, self, [])
+
+    def get_value_str(self):
+        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])
+
+
+# other candidates here: https://docs.microsoft.com/en-us/windows/win32/debug/pe-format#machine-types
+ARCH_I386 = "i386"
+ARCH_AMD64 = "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)
+        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, OS_ANY})
+
+
+class OS(Feature):
+    def __init__(self, value: str, description=None):
+        super(OS, self).__init__(value, description=description)
+        self.name = "os"
+
+
+FORMAT_PE = "pe"
+FORMAT_ELF = "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)
+        self.name = "format"
+
+
+def is_global_feature(feature):
+    """
+    is this a feature that is extracted at every scope?
+    today, these are OS and arch features.
+    """
+    return isinstance(feature, (OS, Arch))

capa/features/extractors/__init__.py

@@ -1,286 +0,0 @@
-# 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 abc
-
-
-class FeatureExtractor(object):
-    """
-    FeatureExtractor defines the interface for fetching features from a sample.
-
-    There may be multiple backends that support fetching features for capa.
-    For example, we use vivisect by default, but also want to support saving
-     and restoring features from a JSON file.
-    When we restore the features, we'd like to use exactly the same matching logic
-     to find matching rules.
-    Therefore, we can define a FeatureExtractor that provides features from the
-     serialized JSON file and do matching without a binary analysis pass.
-    Also, this provides a way to hook in an IDA backend.
-
-    This class is not instantiated directly; it is the base class for other implementations.
-    """
-
-    __metaclass__ = abc.ABCMeta
-
-    def __init__(self):
-        #
-        # note: a subclass should define ctor parameters for its own use.
-        #  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__()
-
-    @abc.abstractmethod
-    def get_base_address(self):
-        """
-        fetch the preferred load address at which the sample was analyzed.
-
-        returns: int
-        """
-        raise NotImplemented
-
-    @abc.abstractmethod
-    def extract_file_features(self):
-        """
-        extract file-scope features.
-
-        example::
-
-            extractor = VivisectFeatureExtractor(vw, path)
-            for feature, va in extractor.get_file_features():
-                print('0x%x: %s', va, feature)
-
-        yields:
-          Tuple[capa.features.Feature, int]: feature and its location
-        """
-        raise NotImplemented
-
-    @abc.abstractmethod
-    def get_functions(self):
-        """
-        enumerate the functions and provide opaque values that will
-         subsequently be provided to `.extract_function_features()`, etc.
-
-        by "opaque value", we mean that this can be any object, as long as it
-         provides enough context to `.extract_function_features()`.
-
-        the opaque value should support casting to int (`__int__`) for the function start address.
-
-        yields:
-          any: the opaque function value.
-        """
-        raise NotImplemented
-
-    @abc.abstractmethod
-    def extract_function_features(self, f):
-        """
-        extract function-scope features.
-        the arguments are opaque values previously provided by `.get_functions()`, etc.
-
-        example::
-
-            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)
-
-        args:
-          f [any]: an opaque value previously fetched from `.get_functions()`.
-
-        yields:
-          Tuple[capa.features.Feature, int]: feature and its location
-        """
-        raise NotImplemented
-
-    @abc.abstractmethod
-    def get_basic_blocks(self, f):
-        """
-        enumerate the basic blocks in the given function and provide opaque values that will
-         subsequently be provided to `.extract_basic_block_features()`, etc.
-
-        by "opaque value", we mean that this can be any object, as long as it
-         provides enough context to `.extract_basic_block_features()`.
-
-        the opaque value should support casting to int (`__int__`) for the basic block start address.
-
-        yields:
-          any: the opaque basic block value.
-        """
-        raise NotImplemented
-
-    @abc.abstractmethod
-    def extract_basic_block_features(self, f, bb):
-        """
-        extract basic block-scope features.
-        the arguments are opaque values previously provided by `.get_functions()`, etc.
-
-        example::
-
-            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)
-
-        args:
-          f [any]: an opaque value previously fetched from `.get_functions()`.
-          bb [any]: an opaque value previously fetched from `.get_basic_blocks()`.
-
-        yields:
-          Tuple[capa.features.Feature, int]: feature and its location
-        """
-        raise NotImplemented
-
-    @abc.abstractmethod
-    def get_instructions(self, f, bb):
-        """
-        enumerate the instructions in the given basic block and provide opaque values that will
-         subsequently be provided to `.extract_insn_features()`, etc.
-
-        by "opaque value", we mean that this can be any object, as long as it
-         provides enough context to `.extract_insn_features()`.
-
-        the opaque value should support casting to int (`__int__`) for the instruction address.
-
-        yields:
-          any: the opaque function value.
-        """
-        raise NotImplemented
-
-    @abc.abstractmethod
-    def extract_insn_features(self, f, bb, insn):
-        """
-        extract instruction-scope features.
-        the arguments are opaque values previously provided by `.get_functions()`, etc.
-
-        example::
-
-            extractor = VivisectFeatureExtractor(vw, path)
-            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)
-
-        args:
-          f [any]: an opaque value previously fetched from `.get_functions()`.
-          bb [any]: an opaque value previously fetched from `.get_basic_blocks()`.
-          insn [any]: an opaque value previously fetched from `.get_instructions()`.
-
-        yields:
-          Tuple[capa.features.Feature, int]: feature and its location
-        """
-        raise NotImplemented
-
-
-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,
-            '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_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/base_extractor.py

@@ -0,0 +1,259 @@
+# 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
+import dataclasses
+from typing import Any, Dict, Tuple, Iterator
+from dataclasses import dataclass
+
+from capa.features.common import Feature
+from capa.features.address import Address, AbsoluteVirtualAddress
+
+# feature extractors may reference functions, BBs, insns by opaque handle values.
+# 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.
+
+
+@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:
+    """
+    FeatureExtractor defines the interface for fetching features from a sample.
+
+    There may be multiple backends that support fetching features for capa.
+    For example, we use vivisect by default, but also want to support saving
+     and restoring features from a JSON file.
+    When we restore the features, we'd like to use exactly the same matching logic
+     to find matching rules.
+    Therefore, we can define a FeatureExtractor that provides features from the
+     serialized JSON file and do matching without a binary analysis pass.
+    Also, this provides a way to hook in an IDA backend.
+
+    This class is not instantiated directly; it is the base class for other implementations.
+    """
+
+    __metaclass__ = abc.ABCMeta
+
+    def __init__(self):
+        #
+        # note: a subclass should define ctor parameters for its own use.
+        #  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__()
+
+    @abc.abstractmethod
+    def get_base_address(self) -> AbsoluteVirtualAddress:
+        """
+        fetch the preferred load address at which the sample was analyzed.
+        """
+        raise NotImplementedError()
+
+    @abc.abstractmethod
+    def extract_global_features(self) -> Iterator[Tuple[Feature, Address]]:
+        """
+        extract features found at every scope ("global").
+
+        example::
+
+            extractor = VivisectFeatureExtractor(vw, path)
+            for feature, va in extractor.get_global_features():
+                print('0x%x: %s', va, feature)
+
+        yields:
+          Tuple[Feature, Address]: feature and its location
+        """
+        raise NotImplementedError()
+
+    @abc.abstractmethod
+    def extract_file_features(self) -> Iterator[Tuple[Feature, Address]]:
+        """
+        extract file-scope features.
+
+        example::
+
+            extractor = VivisectFeatureExtractor(vw, path)
+            for feature, va in extractor.get_file_features():
+                print('0x%x: %s', va, feature)
+
+        yields:
+          Tuple[Feature, Address]: feature and its location
+        """
+        raise NotImplementedError()
+
+    @abc.abstractmethod
+    def get_functions(self) -> Iterator[FunctionHandle]:
+        """
+        enumerate the functions and provide opaque values that will
+         subsequently be provided to `.extract_function_features()`, etc.
+        """
+        raise NotImplementedError()
+
+    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 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:
+          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, 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:
+          addr (Address): the address of a function.
+
+        returns:
+          str: the function name
+
+        raises:
+          KeyError: when the given function does not have a name.
+        """
+        raise KeyError(addr)
+
+    @abc.abstractmethod
+    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.
+
+        example::
+
+            extractor = VivisectFeatureExtractor(vw, path)
+            for function in extractor.get_functions():
+                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, Address]: feature and its location
+        """
+        raise NotImplementedError()
+
+    @abc.abstractmethod
+    def get_basic_blocks(self, f: FunctionHandle) -> Iterator[BBHandle]:
+        """
+        enumerate the basic blocks in the given function and provide opaque values that will
+         subsequently be provided to `.extract_basic_block_features()`, etc.
+        """
+        raise NotImplementedError()
+
+    @abc.abstractmethod
+    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.
+
+        example::
+
+            extractor = VivisectFeatureExtractor(vw, path)
+            for function in extractor.get_functions():
+                for bb in extractor.get_basic_blocks(function):
+                    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, Address]: feature and its location
+        """
+        raise NotImplementedError()
+
+    @abc.abstractmethod
+    def get_instructions(self, f: FunctionHandle, bb: BBHandle) -> Iterator[InsnHandle]:
+        """
+        enumerate the instructions in the given basic block and provide opaque values that will
+         subsequently be provided to `.extract_insn_features()`, etc.
+        """
+        raise NotImplementedError()
+
+    @abc.abstractmethod
+    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.
+
+        example::
+
+            extractor = VivisectFeatureExtractor(vw, path)
+            for function in extractor.get_functions():
+                for bb in extractor.get_basic_blocks(function):
+                    for insn in extractor.get_instructions(function, bb):
+                        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()`.
+          bb [BBHandle]: an opaque value previously fetched from `.get_basic_blocks()`.
+          insn [InsnHandle]: an opaque value previously fetched from `.get_instructions()`.
+
+        yields:
+          Tuple[Feature, Address]: feature and its location
+        """
+        raise NotImplementedError()

capa/features/extractors/common.py

@@ -0,0 +1,100 @@
+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, 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) -> 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), FileOffsetAddress(s.offset)
+
+    for s in capa.features.extractors.strings.extract_unicode_strings(buf):
+        yield String(s.s), FileOffsetAddress(s.offset)
+
+
+def extract_format(buf) -> Iterator[Tuple[Feature, Address]]:
+    if buf.startswith(b"MZ"):
+        yield Format(FORMAT_PE), NO_ADDRESS
+    elif buf.startswith(b"\x7fELF"):
+        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)
+        #
+        # for (1), this logic will need to be updated as the format is implemented.
+        logger.debug("unsupported file format: %s", binascii.hexlify(buf[:4]).decode("ascii"))
+        return
+
+
+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))
+
+    elif buf.startswith(b"\x7fELF"):
+        with contextlib.closing(io.BytesIO(buf)) as f:
+            arch = capa.features.extractors.elf.detect_elf_arch(f)
+
+        if arch not in capa.features.common.VALID_ARCH:
+            logger.debug("unsupported arch: %s", arch)
+            return
+
+        yield Arch(arch), NO_ADDRESS
+
+    else:
+        # we likely end up here:
+        #  1. handling shellcode, or
+        #  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,
+        # but i think this would be rarely used.
+        # rules that rely on arch conditions will fail to match on shellcode.
+        #
+        # for (2), this logic will need to be updated as the format is implemented.
+        logger.debug("unsupported file format: %s, will not guess Arch", binascii.hexlify(buf[:4]).decode("ascii"))
+        return
+
+
+def extract_os(buf) -> Iterator[Tuple[Feature, Address]]:
+    if buf.startswith(b"MZ"):
+        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)
+
+        if os not in capa.features.common.VALID_OS:
+            logger.debug("unsupported os: %s", os)
+            return
+
+        yield OS(os), NO_ADDRESS
+
+    else:
+        # we likely end up here:
+        #  1. handling shellcode, or
+        #  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,
+        # but i think this would be rarely used.
+        # rules that rely on OS conditions will fail to match on shellcode.
+        #
+        # for (2), this logic will need to be updated as the format is implemented.
+        logger.debug("unsupported file format: %s, will not guess OS", binascii.hexlify(buf[:4]).decode("ascii"))
+        return

capa/features/extractors/dnfile/extractor.py

@@ -0,0 +1,71 @@
+# 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 TYPE_CHECKING, List, Tuple, Iterator
+
+from capa.features.address import Address, DNTokenAddress, DNTokenOffsetAddress, AbsoluteVirtualAddress
+
+if TYPE_CHECKING:
+    from capa.features.common import Feature
+
+import dnfile
+from dncil.clr.token import Token
+
+import capa.features.extractors
+import capa.features.extractors.dnfile.file
+import capa.features.extractors.dnfile.insn
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle, FeatureExtractor
+from capa.features.extractors.dnfile.helpers import get_dotnet_managed_method_bodies
+
+
+class DnfileFeatureExtractor(FeatureExtractor):
+    def __init__(self, path: str):
+        super(DnfileFeatureExtractor, self).__init__()
+        self.pe: dnfile.dnPE = dnfile.dnPE(path)
+
+        # 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_os(pe=self.pe))
+        self.global_features.extend(capa.features.extractors.dotnetfile.extract_file_arch(pe=self.pe))
+
+    def get_base_address(self):
+        return AbsoluteVirtualAddress(0x0)
+
+    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]:
+        for token, f in get_dotnet_managed_method_bodies(self.pe):
+            yield FunctionHandle(address=DNTokenAddress(Token(token)), inner=f, ctx={"pe": self.pe})
+
+    def extract_function_features(self, f):
+        # TODO
+        yield from []
+
+    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.token, insn.offset - fh.inner.offset), 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 +1,54 @@
+# 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 TYPE_CHECKING, Tuple, Iterator
+
+if TYPE_CHECKING:
+    import dnfile
+    from capa.features.common import Feature, Format, String, Characteristic
+    from capa.features.file import Import, FunctionName
+    from capa.features.address import Address
+
+import capa.features.extractors
+
+
+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_mixed_mode_characteristic_features(pe: dnfile.dnPE) -> Iterator[Tuple[Characteristic, Address]]:
+    yield from capa.features.extractors.dotnetfile.extract_mixed_mode_characteristic_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_mixed_mode_characteristic_features,
+)

capa/features/extractors/dnfile/helpers.py

@@ -0,0 +1,213 @@
+# 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 Any, Tuple, 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
+
+logger = logging.getLogger(__name__)
+
+# key indexes to dotnet metadata tables
+DOTNET_META_TABLES_BY_INDEX = {table.value: table.name for table in dnfile.enums.MetadataTables}
+
+
+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 calculate_dotnet_token_value(table: int, rid: int) -> int:
+    return ((table & 0xFF) << Token.TABLE_SHIFT) | (rid & Token.RID_MASK)
+
+
+def resolve_dotnet_token(pe: dnfile.dnPE, token: Token) -> Any:
+    """map generic token to string or table row"""
+    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_name: str = DOTNET_META_TABLES_BY_INDEX.get(token.table, "")
+    if not table_name:
+        # table_index is not valid
+        return InvalidToken(token.value)
+
+    table: Any = getattr(pe.net.mdtables, table_name, None)
+    if table is None:
+        # table index is valid but table is not present
+        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.warn("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"""
+    try:
+        user_string: Optional[dnfile.stream.UserString] = pe.net.user_strings.get_us(token.rid)
+    except UnicodeDecodeError as e:
+        logger.warn("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[Tuple[int, str]]:
+    """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)
+    """
+    if not is_dotnet_table_valid(pe, "MemberRef"):
+        return
+
+    for (rid, row) in enumerate(pe.net.mdtables.MemberRef):
+        if not isinstance(row.Class.row, (dnfile.mdtable.TypeRefRow,)):
+            continue
+
+        # like File::OpenRead
+        name = f"{row.Class.row.TypeName}::{row.Name}"
+
+        # ECMA II.22.38: TypeNamespace can be null or non-null
+        if row.Class.row.TypeNamespace:
+            # like System.IO.File::OpenRead
+            name = f"{row.Class.row.TypeNamespace}.{name}"
+
+        token: int = calculate_dotnet_token_value(pe.net.mdtables.MemberRef.number, rid + 1)
+
+        yield token, name
+
+
+def get_dotnet_unmanaged_imports(pe: dnfile.dnPE) -> Iterator[Tuple[int, str]]:
+    """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)
+    """
+    if not is_dotnet_table_valid(pe, "ImplMap"):
+        return
+
+    for row in pe.net.mdtables.ImplMap:
+        dll: str = row.ImportScope.row.Name
+        symbol: str = row.ImportName
+
+        # 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(row.MemberForwarded.table.number, row.MemberForwarded.row_index)
+
+        # like Kernel32.dll
+        if dll and "." in dll:
+            dll = dll.split(".")[0]
+
+        # like kernel32.CreateFileA
+        name: str = f"{dll}.{symbol}"
+
+        yield token, name
+
+
+def get_dotnet_managed_method_bodies(pe: dnfile.dnPE) -> Iterator[Tuple[int, CilMethodBody]]:
+    """get managed methods from MethodDef table"""
+    if not is_dotnet_table_valid(pe, "MethodDef"):
+        return
+
+    for (rid, row) in enumerate(pe.net.mdtables.MethodDef):
+        if not row.ImplFlags.miIL or any((row.Flags.mdAbstract, row.Flags.mdPinvokeImpl)):
+            # skip methods that do not have a method body
+            continue
+
+        body: Optional[CilMethodBody] = read_dotnet_method_body(pe, row)
+        if body is None:
+            continue
+
+        token: int = calculate_dotnet_token_value(dnfile.enums.MetadataTables.MethodDef.value, rid + 1)
+        yield token, body
+
+
+def is_dotnet_table_valid(pe: dnfile.dnPE, table_name: str) -> bool:
+    return bool(getattr(pe.net.mdtables, table_name, None))
+
+
+def get_dotnet_managed_method_names(pe: dnfile.dnPE) -> Iterator[Tuple[int, str]]:
+    """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 continguous run of Methods owned by this Type)
+    """
+    if not is_dotnet_table_valid(pe, "TypeDef"):
+        return
+
+    for row in pe.net.mdtables.TypeDef:
+        for index in row.MethodList:
+            # like File::OpenRead
+            name = f"{row.TypeName}::{index.row.Name}"
+
+            # ECMA II.22.37: TypeNamespace can be null or non-null
+            if row.TypeNamespace:
+                # like System.IO.File::OpenRead
+                name = f"{row.TypeNamespace}.{name}"
+
+            token = calculate_dotnet_token_value(index.table.number, index.row_index)
+
+            yield token, name
+
+
+def is_dotnet_mixed_mode(pe: dnfile.dnPE) -> bool:
+    return not bool(pe.net.Flags.CLR_ILONLY)

capa/features/extractors/dnfile/insn.py

@@ -0,0 +1,149 @@
+# 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 TYPE_CHECKING, Any, Dict, Tuple, Iterator, Optional
+
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle
+
+if TYPE_CHECKING:
+    from dncil.cil.instruction import Instruction
+    from dncil.cil.body import CilMethodBody
+    from capa.features.common import Feature
+    from capa.features.address import Address
+
+import dnfile
+from dncil.clr.token import StringToken, InvalidToken
+from dncil.cil.opcode import OpCodes
+
+import capa.features.extractors.helpers
+from capa.features.insn import API, Number
+from capa.features.common import String, Characteristic
+from capa.features.extractors.dnfile.helpers import (
+    resolve_dotnet_token,
+    read_dotnet_user_string,
+    get_dotnet_managed_imports,
+    get_dotnet_unmanaged_imports,
+    get_dotnet_managed_method_names,
+)
+
+
+def get_managed_imports(ctx: Dict) -> Dict:
+    if "managed_imports_cache" not in ctx:
+        ctx["managed_imports_cache"] = {}
+        for (token, name) in get_dotnet_managed_imports(ctx["pe"]):
+            ctx["managed_imports_cache"][token] = name
+    return ctx["managed_imports_cache"]
+
+
+def get_unmanaged_imports(ctx: Dict) -> Dict:
+    if "unmanaged_imports_cache" not in ctx:
+        ctx["unmanaged_imports_cache"] = {}
+        for (token, name) in get_dotnet_unmanaged_imports(ctx["pe"]):
+            ctx["unmanaged_imports_cache"][token] = name
+    return ctx["unmanaged_imports_cache"]
+
+
+def get_methods(ctx: Dict) -> Dict:
+    if "methods_cache" not in ctx:
+        ctx["methods_cache"] = {}
+        for (token, name) in get_dotnet_managed_method_names(ctx["pe"]):
+            ctx["methods_cache"][token] = name
+    return ctx["methods_cache"]
+
+
+def get_callee_name(ctx: Dict, token: int) -> str:
+    """map dotnet token to method name"""
+    name: str = get_managed_imports(ctx).get(token, "")
+    if not name:
+        # 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
+        name = get_unmanaged_imports(ctx).get(token, "")
+        if not name:
+            name = get_methods(ctx).get(token, "")
+    return name
+
+
+def extract_insn_api_features(fh: FunctionHandle, bh, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """parse instruction API features"""
+    insn: Instruction = ih.inner
+
+    if insn.opcode not in (OpCodes.Call, OpCodes.Callvirt, OpCodes.Jmp, OpCodes.Calli):
+        return
+
+    name: str = get_callee_name(fh.ctx, insn.operand.value)
+    if not name:
+        return
+
+    if "::" in name:
+        # like System.IO.File::OpenRead
+        yield API(name), ih.address
+    else:
+        # like kernel32.CreateFileA
+        dll, _, symbol = name.rpartition(".")
+        for name_variant in capa.features.extractors.helpers.generate_symbols(dll, symbol):
+            yield API(name_variant), ih.address
+
+
+def extract_insn_number_features(fh, bh, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """parse instruction number features"""
+    insn: Instruction = ih.inner
+
+    if insn.is_ldc():
+        yield Number(insn.get_ldc()), ih.address
+
+
+def extract_insn_string_features(fh: FunctionHandle, bh, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """parse instruction string features"""
+    f: CilMethodBody = fh.inner
+    insn: Instruction = ih.inner
+
+    if not insn.is_ldstr():
+        return
+
+    if not isinstance(insn.operand, StringToken):
+        return
+
+    user_string: Optional[str] = read_dotnet_user_string(fh.ctx["pe"], insn.operand)
+    if user_string is None:
+        return
+
+    yield String(user_string), ih.address
+
+
+def extract_unmanaged_call_characteristic_features(
+    fh: FunctionHandle, bb: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Characteristic, Address]]:
+    insn: Instruction = ih.inner
+    if insn.opcode not in (OpCodes.Call, OpCodes.Callvirt, OpCodes.Jmp, OpCodes.Calli):
+        return
+
+    token: Any = resolve_dotnet_token(fh.ctx["pe"], insn.operand)
+    if isinstance(token, InvalidToken):
+        return
+    if not isinstance(token, dnfile.mdtable.MethodDefRow):
+        return
+
+    if any((token.Flags.mdPinvokeImpl, token.ImplFlags.miUnmanaged, token.ImplFlags.miNative)):
+        yield Characteristic("unmanaged call"), insn.offset
+
+
+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):
+            yield feature, addr
+
+
+INSTRUCTION_HANDLERS = (
+    extract_insn_api_features,
+    extract_insn_number_features,
+    extract_insn_string_features,
+    extract_unmanaged_call_characteristic_features,
+)

capa/features/extractors/dnfile_.py

@@ -0,0 +1,116 @@
+import logging
+from typing import Tuple, Iterator
+
+import dnfile
+import pefile
+
+from capa.features.common import OS, OS_ANY, ARCH_ANY, ARCH_I386, 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_DOTNET), NO_ADDRESS
+
+
+def extract_file_os(**kwargs) -> Iterator[Tuple[Feature, Address]]:
+    yield OS(OS_ANY), NO_ADDRESS
+
+
+def extract_file_arch(pe, **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
+    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(DnfileFeatureExtractor, self).__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
+        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 not bool(self.pe.net.Flags.CLR_ILONLY)
+
+    def get_runtime_version(self) -> Tuple[int, int]:
+        return self.pe.net.struct.MajorRuntimeVersion, self.pe.net.struct.MinorRuntimeVersion
+
+    def get_meta_version_string(self) -> str:
+        return self.pe.net.metadata.struct.Version.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 +1,163 @@
+import logging
+from typing import Tuple, Iterator
+
+import dnfile
+import pefile
+from dncil.clr.token import Token
+
+import capa.features.extractors.helpers
+from capa.features.file import Import, FunctionName
+from capa.features.common import (
+    OS,
+    OS_ANY,
+    ARCH_ANY,
+    ARCH_I386,
+    ARCH_AMD64,
+    FORMAT_DOTNET,
+    Arch,
+    Format,
+    String,
+    Feature,
+    Characteristic,
+)
+from capa.features.address import NO_ADDRESS, Address, DNTokenAddress, DNTokenOffsetAddress, AbsoluteVirtualAddress
+from capa.features.extractors.base_extractor import FeatureExtractor
+from capa.features.extractors.dnfile.helpers import (
+    is_dotnet_mixed_mode,
+    get_dotnet_managed_imports,
+    calculate_dotnet_token_value,
+    get_dotnet_unmanaged_imports,
+    get_dotnet_managed_method_names,
+)
+
+logger = logging.getLogger(__name__)
+
+
+def extract_file_format(**kwargs) -> Iterator[Tuple[Format, Address]]:
+    yield Format(FORMAT_DOTNET), NO_ADDRESS
+
+
+def extract_file_import_names(pe: dnfile.dnPE, **kwargs) -> Iterator[Tuple[Import, Address]]:
+    for (token, name) in get_dotnet_managed_imports(pe):
+        # like System.IO.File::OpenRead
+        yield Import(name), DNTokenAddress(Token(token))
+
+    for (token, name) in get_dotnet_unmanaged_imports(pe):
+        # like kernel32.CreateFileA
+        dll, _, symbol = name.rpartition(".")
+        for name_variant in capa.features.extractors.helpers.generate_symbols(dll, symbol):
+            yield Import(name_variant), DNTokenAddress(Token(token))
+
+
+def extract_file_function_names(pe: dnfile.dnPE, **kwargs) -> Iterator[Tuple[FunctionName, Address]]:
+    for (token, name) in get_dotnet_managed_method_names(pe):
+        yield FunctionName(name), DNTokenAddress(Token(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
+    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_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_mixed_mode_characteristic_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(DotnetFileFeatureExtractor, self).__init__()
+        self.path: str = path
+        self.pe: dnfile.dnPE = dnfile.dnPE(path)
+
+    def get_base_address(self):
+        return AbsoluteVirtualAddress(0x0)
+
+    def get_entry_point(self) -> int:
+        # self.pe.net.Flags.CLT_NATIVE_ENTRYPOINT
+        #  True: native EP: Token
+        #  False: managed EP: RVA
+        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]:
+        return self.pe.net.struct.MajorRuntimeVersion, self.pe.net.struct.MinorRuntimeVersion
+
+    def get_meta_version_string(self) -> str:
+        return self.pe.net.metadata.struct.Version.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

@@ -0,0 +1,326 @@
+# 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 struct
+import logging
+from enum import Enum
+from typing import BinaryIO
+
+logger = logging.getLogger(__name__)
+
+
+def align(v, alignment):
+    remainder = v % alignment
+    if remainder == 0:
+        return v
+    else:
+        return v + (alignment - remainder)
+
+
+class CorruptElfFile(ValueError):
+    pass
+
+
+class OS(str, Enum):
+    HPUX = "hpux"
+    NETBSD = "netbsd"
+    LINUX = "linux"
+    HURD = "hurd"
+    _86OPEN = "86open"
+    SOLARIS = "solaris"
+    AIX = "aix"
+    IRIX = "irix"
+    FREEBSD = "freebsd"
+    TRU64 = "tru64"
+    MODESTO = "modesto"
+    OPENBSD = "openbsd"
+    OPENVMS = "openvms"
+    NSK = "nsk"
+    AROS = "aros"
+    FENIXOS = "fenixos"
+    CLOUD = "cloud"
+    SYLLABLE = "syllable"
+    NACL = "nacl"
+
+
+def detect_elf_os(f) -> str:
+    """
+    f: type Union[BinaryIO, IDAIO]
+    """
+    f.seek(0x0)
+    file_header = f.read(0x40)
+
+    # 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")
+
+    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)
+
+    if bitness == 32:
+        (e_phoff, e_shoff) = struct.unpack_from(endian + "II", file_header, 0x1C)
+        e_phentsize, e_phnum = struct.unpack_from(endian + "HH", file_header, 0x2A)
+        e_shentsize, e_shnum = struct.unpack_from(endian + "HH", file_header, 0x2E)
+    elif bitness == 64:
+        (e_phoff, e_shoff) = struct.unpack_from(endian + "QQ", file_header, 0x20)
+        e_phentsize, e_phnum = struct.unpack_from(endian + "HH", file_header, 0x36)
+        e_shentsize, e_shnum = struct.unpack_from(endian + "HH", file_header, 0x3A)
+    else:
+        raise NotImplementedError()
+
+    logger.debug("e_phoff: 0x%02x e_phentsize: 0x%02x e_phnum: %d", e_phoff, e_phentsize, e_phnum)
+
+    (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,
+        # at least when guessing OSes
+        # 0: "SYSV",  # too often used when OS is not SYSV
+        1: OS.HPUX,
+        2: OS.NETBSD,
+        3: OS.LINUX,
+        4: OS.HURD,
+        5: OS._86OPEN,
+        6: OS.SOLARIS,
+        7: OS.AIX,
+        8: OS.IRIX,
+        9: OS.FREEBSD,
+        10: OS.TRU64,
+        11: OS.MODESTO,
+        12: OS.OPENBSD,
+        13: OS.OPENVMS,
+        14: OS.NSK,
+        15: OS.AROS,
+        16: OS.FENIXOS,
+        17: OS.CLOUD,
+        # 53: "SORTFIX",      # i can't find any reference to this OS, i dont think it exists
+        # 64: "ARM_AEABI",    # not an OS
+        # 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]
+
+    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
+
+    # 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]
+
+        PT_NOTE = 0x4
+
+        (p_type,) = struct.unpack_from(endian + "I", phent, 0x0)
+        logger.debug("p_type: 0x%04x", p_type)
+        if p_type != PT_NOTE:
+            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)
+        else:
+            raise NotImplementedError()
+
+        logger.debug("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
+
+        namesz, descsz, type_ = struct.unpack_from(endian + "III", note, 0x0)
+        name_offset = 0xC
+        desc_offset = name_offset + align(namesz, 0x4)
+
+        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:
+                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)
+
+            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)
+        else:
+            raise NotImplementedError()
+
+        f.seek(p_offset)
+        interp = f.read(p_filesz)
+        if len(interp) != p_filesz:
+            logger.warning("failed to read interp content")
+            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
+
+    f.seek(e_shoff)
+    section_header_size = e_shnum * e_shentsize
+    section_headers = f.read(section_header_size)
+    if len(section_headers) != section_header_size:
+        logger.warning("failed to read section headers")
+        e_shnum = 0
+
+    # search for notes stored in sections that aren't visible in program headers.
+    # e.g. .note.Linux in Linux kernel modules.
+    for i in range(e_shnum):
+        offset = i * e_shentsize
+        shent = section_headers[offset : offset + e_shentsize]
+
+        if bitness == 32:
+            sh_name, sh_type, _, sh_addr, sh_offset, sh_size = struct.unpack_from(endian + "IIIIII", shent, 0x0)
+        elif bitness == 64:
+            sh_name, sh_type, _, sh_addr, sh_offset, sh_size = struct.unpack_from(endian + "IIQQQQ", shent, 0x0)
+        else:
+            raise NotImplementedError()
+
+        SHT_NOTE = 0x7
+        if sh_type != SHT_NOTE:
+            continue
+
+        logger.debug("sh_offset: 0x%02x sh_size: 0x%04x", sh_offset, sh_size)
+
+        f.seek(sh_offset)
+        note = f.read(sh_size)
+        if len(note) != sh_size:
+            logger.warning("failed to read note content")
+            continue
+
+        namesz, descsz, type_ = struct.unpack_from(endian + "III", note, 0x0)
+        name_offset = 0xC
+        desc_offset = name_offset + align(namesz, 0x4)
+
+        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 name == "Linux":
+            logger.debug("note owner: %s", "LINUX")
+            ret = OS.LINUX if not ret else ret
+
+    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"

capa/features/extractors/elffile.py

@@ -0,0 +1,160 @@
+# 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 io
+import logging
+import contextlib
+from typing import Tuple, Iterator
+
+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.address import NO_ADDRESS, FileOffsetAddress, AbsoluteVirtualAddress
+from capa.features.extractors.elf import Arch as ElfArch
+from capa.features.extractors.base_extractor import FeatureExtractor
+
+logger = logging.getLogger(__name__)
+
+
+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:
+        if not isinstance(section, SymbolTableSection):
+            continue
+
+        if section["sh_entsize"] == 0:
+            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()))
+
+        for nsym, 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), FileOffsetAddress(0x0)
+
+
+def extract_file_section_names(elf, **kwargs):
+    for section in elf.iter_sections():
+        if section.name:
+            yield Section(section.name), AbsoluteVirtualAddress(section.header.sh_addr)
+        elif section.is_null():
+            yield Section("NULL"), AbsoluteVirtualAddress(section.header.sh_addr)
+
+
+def extract_file_strings(buf, **kwargs):
+    yield from capa.features.extractors.common.extract_file_strings(buf)
+
+
+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_tuple = next(capa.features.extractors.common.extract_os(buf))
+        yield os_tuple
+    except StopIteration:
+        yield OS("unknown"), NO_ADDRESS
+
+
+def extract_file_format(**kwargs):
+    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), NO_ADDRESS
+    elif arch == "x64":
+        yield Arch(ElfArch.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, addr in file_handler(elf=elf, buf=buf):  # type: ignore
+            yield feature, addr
+
+
+FILE_HANDLERS = (
+    # TODO extract_file_export_names,
+    extract_file_import_names,
+    extract_file_section_names,
+    extract_file_strings,
+    # no library matching
+    extract_file_format,
+)
+
+
+def extract_global_features(elf: ELFFile, buf: bytes) -> Iterator[Tuple[Feature, int]]:
+    for global_handler in GLOBAL_HANDLERS:
+        for feature, addr in global_handler(elf=elf, buf=buf):  # type: ignore
+            yield feature, addr
+
+
+GLOBAL_HANDLERS = (
+    extract_file_os,
+    extract_file_arch,
+)
+
+
+class ElfFeatureExtractor(FeatureExtractor):
+    def __init__(self, path: str):
+        super(ElfFeatureExtractor, self).__init__()
+        self.path = path
+        with open(self.path, "rb") as f:
+            self.elf = ELFFile(io.BytesIO(f.read()))
+
+    def get_base_address(self):
+        # virtual address of the first segment with type LOAD
+        for segment in self.elf.iter_segments():
+            if segment.header.p_type == "PT_LOAD":
+                return AbsoluteVirtualAddress(segment.header.p_vaddr)
+
+    def extract_global_features(self):
+        with open(self.path, "rb") as f:
+            buf = f.read()
+
+        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, 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")
+
+    def extract_function_features(self, f):
+        raise NotImplementedError("ElfFeatureExtractor can only be used to extract file features")
+
+    def get_basic_blocks(self, f):
+        raise NotImplementedError("ElfFeatureExtractor can only be used to extract file features")
+
+    def extract_basic_block_features(self, f, bb):
+        raise NotImplementedError("ElfFeatureExtractor can only be used to extract file features")
+
+    def get_instructions(self, f, bb):
+        raise NotImplementedError("ElfFeatureExtractor can only be used to extract file features")
+
+    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):
+        raise NotImplementedError("ElfFeatureExtractor can only be used to extract file features")
+
+    def get_function_name(self, va):
+        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
@@ -6,23 +6,18 @@
 #  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 sys
+import struct
 import builtins
-
-from capa.features.file import Import
-from capa.features.insn import API
+from typing import Tuple, Iterator
 
 MIN_STACKSTRING_LEN = 8
 
 
-def xor_static(data, i):
-    if sys.version_info >= (3, 0):
-        return bytes(c ^ i for c in data)
-    else:
-        return "".join(chr(ord(c) ^ i) for c in data)
+def xor_static(data: bytes, i: int) -> bytes:
+    return bytes(c ^ i for c in data)
 
 
-def is_aw_function(symbol):
+def is_aw_function(symbol: str) -> bool:
     """
     is the given function name an A/W function?
     these are variants of functions that, on Windows, accept either a narrow or wide string.
@@ -34,11 +29,10 @@ def is_aw_function(symbol):
     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):
+def is_ordinal(symbol: str) -> bool:
     """
     is the given symbol an ordinal that is prefixed by "#"?
     """
@@ -47,7 +41,7 @@ def is_ordinal(symbol):
     return False
 
 
-def generate_symbols(dll, symbol):
+def generate_symbols(dll: str, symbol: str) -> Iterator[str]:
     """
     for a given dll and symbol name, generate variants.
     we over-generate features to make matching easier.
@@ -57,6 +51,9 @@ def generate_symbols(dll, symbol):
       - CreateFileA
       - CreateFile
     """
+    # normalize dll name
+    dll = dll.lower()
+
     # kernel32.CreateFileA
     yield "%s.%s" % (dll, symbol)
 
@@ -73,11 +70,11 @@ def generate_symbols(dll, symbol):
             yield symbol[:-1]
 
 
-def all_zeros(bytez):
+def all_zeros(bytez: bytes) -> bool:
     return all(b == 0 for b in builtins.bytes(bytez))
 
 
-def twos_complement(val, bits):
+def twos_complement(val: int, bits: int) -> int:
     """
     compute the 2's complement of int value val
 
@@ -90,3 +87,49 @@ def twos_complement(val, bits):
     else:
         # return positive value as is
         return val
+
+
+def carve_pe(pbytes: bytes, offset: int = 0) -> Iterator[Tuple[int, int]]:
+    """
+    Generate (offset, key) tuples of embedded PEs
+
+    Based on the version from vivisect:
+      https://github.com/vivisect/vivisect/blob/7be4037b1cecc4551b397f840405a1fc606f9b53/PE/carve.py#L19
+    And its IDA adaptation:
+      capa/features/extractors/ida/file.py
+    """
+    mz_xor = [
+        (
+            xor_static(b"MZ", key),
+            xor_static(b"PE", key),
+            key,
+        )
+        for key in range(256)
+    ]
+
+    pblen = len(pbytes)
+    todo = [(pbytes.find(mzx, offset), mzx, pex, key) for mzx, pex, key in mz_xor]
+    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
+        # e_lfanew is at 0x3c
+        e_lfanew = off + 0x3C
+        if pblen < (e_lfanew + 4):
+            continue
+
+        newoff = struct.unpack("<I", xor_static(pbytes[e_lfanew : e_lfanew + 4], key))[0]
+
+        nextres = pbytes.find(mzx, off + 1)
+        if nextres != -1:
+            todo.append((nextres, mzx, pex, key))
+
+        peoff = off + newoff
+        if pblen < (peoff + 2):
+            continue
+
+        if pbytes[peoff : peoff + 2] == pex:
+            yield (off, key)

capa/features/extractors/ida/__init__.py

@@ -1,93 +0,0 @@
-# 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 sys
-import types
-
-import idaapi
-
-import capa.features.extractors.ida.file
-import capa.features.extractors.ida.insn
-import capa.features.extractors.ida.function
-import capa.features.extractors.ida.basicblock
-from capa.features.extractors import FeatureExtractor
-
-
-def get_ea(self):
-    """ """
-    if isinstance(self, (idaapi.BasicBlock, idaapi.func_t)):
-        return self.start_ea
-    if isinstance(self, idaapi.insn_t):
-        return self.ea
-    raise TypeError
-
-
-def add_ea_int_cast(o):
-    """
-    dynamically add a cast-to-int (`__int__`) method to the given object
-    that returns the value of the `.ea` property.
-    this bit of skullduggery lets use cast viv-utils objects as ints.
-    the correct way of doing this is to update viv-utils (or subclass the objects here).
-    """
-    if sys.version_info[0] >= 3:
-        setattr(o, "__int__", types.MethodType(get_ea, o))
-    else:
-        setattr(o, "__int__", types.MethodType(get_ea, o, type(o)))
-    return o
-
-
-class IdaFeatureExtractor(FeatureExtractor):
-    def __init__(self):
-        super(IdaFeatureExtractor, self).__init__()
-
-    def get_base_address(self):
-        return idaapi.get_imagebase()
-
-    def extract_file_features(self):
-        for (feature, ea) in capa.features.extractors.ida.file.extract_features():
-            yield feature, ea
-
-    def get_functions(self):
-        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 add_ea_int_cast(f)
-
-    @staticmethod
-    def get_function(ea):
-        f = idaapi.get_func(ea)
-        setattr(f, "ctx", {})
-        return add_ea_int_cast(f)
-
-    def extract_function_features(self, f):
-        for (feature, ea) in capa.features.extractors.ida.function.extract_features(f):
-            yield feature, ea
-
-    def get_basic_blocks(self, f):
-        for bb in capa.features.extractors.ida.helpers.get_function_blocks(f):
-            yield add_ea_int_cast(bb)
-
-    def extract_basic_block_features(self, f, bb):
-        for (feature, ea) in capa.features.extractors.ida.basicblock.extract_features(f, bb):
-            yield feature, ea
-
-    def get_instructions(self, f, bb):
-        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 add_ea_int_cast(insn)
-
-    def extract_insn_features(self, f, bb, insn):
-        for (feature, ea) in capa.features.extractors.ida.insn.extract_features(f, bb, insn):
-            yield feature, ea

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
@@ -6,25 +6,23 @@
 #  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 sys
 import string
 import struct
+from typing import Tuple, Iterator
 
 import idaapi
 
 import capa.features.extractors.ida.helpers
-from capa.features 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:
@@ -38,19 +36,12 @@ def get_printable_len(op):
     else:
         raise ValueError("Unhandled operand data type 0x%x." % op.dtype)
 
-    def is_printable_ascii(chars):
-        if sys.version_info[0] >= 3:
-            return all(c < 127 and chr(c) in string.printable for c in chars)
-        else:
-            return all(ord(c) < 127 and 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 sys.version_info[0] >= 3:
-            if all(c == 0x00 for c in chars[1::2]):
-                return is_printable_ascii(chars[::2])
-        else:
-            if all(c == "\x00" 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)
@@ -61,12 +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
 
@@ -79,14 +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):
@@ -97,39 +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 = (
@@ -140,9 +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

@@ -0,0 +1,70 @@
+# 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
+import capa.features.extractors.elf
+import capa.features.extractors.ida.file
+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.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: List[Tuple[Feature, Address]] = []
+        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 AbsoluteVirtualAddress(idaapi.get_imagebase())
+
+    def extract_global_features(self):
+        yield from self.global_features
+
+    def extract_file_features(self):
+        yield from capa.features.extractors.ida.file.extract_features()
+
+    def get_functions(self) -> Iterator[FunctionHandle]:
+        import capa.features.extractors.ida.helpers as ida_helpers
+
+        # ignore library functions and thunk functions as identified by IDA
+        yield from ida_helpers.get_functions(skip_thunks=True, skip_libs=True)
+
+    @staticmethod
+    def get_function(ea: int) -> FunctionHandle:
+        f = idaapi.get_func(ea)
+        return FunctionHandle(address=AbsoluteVirtualAddress(f.start_ea), inner=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, fh: FunctionHandle) -> Iterator[BBHandle]:
+        import capa.features.extractors.ida.helpers as ida_helpers
+
+        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, fh: FunctionHandle, bbh: BBHandle) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.ida.basicblock.extract_features(fh, bbh)
+
+    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(bbh.inner.start_ea, bbh.inner.end_ea):
+            yield InsnHandle(address=AbsoluteVirtualAddress(insn.ea), inner=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
@@ -7,26 +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 capa.features.extractors.common
 import capa.features.extractors.helpers
 import capa.features.extractors.strings
 import capa.features.extractors.ida.helpers
-from capa.features import String, Characteristic
-from capa.features.file import Export, Import, Section
+from capa.features.file import Export, Import, Section, FunctionName
+from capa.features.common import FORMAT_PE, FORMAT_ELF, Format, 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 = [
@@ -59,13 +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
@@ -74,16 +74,16 @@ def extract_file_embedded_pe():
     """
     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
+            yield Characteristic("embedded pe"), FileOffsetAddress(ea)
 
 
-def extract_file_export_names():
-    """ extract function exports """
+def extract_file_export_names() -> Iterator[Tuple[Feature, Address]]:
+    """extract function exports"""
     for (_, _, ea, name) in idautils.Entries():
-        yield Export(name), ea
+        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:
@@ -95,11 +95,12 @@ def extract_file_import_names():
      - importname
     """
     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]:
@@ -112,10 +113,10 @@ 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
 
 
-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
@@ -123,10 +124,10 @@ 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
@@ -135,19 +136,45 @@ 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)
-
-        for s in capa.features.extractors.strings.extract_ascii_strings(seg_buff):
-            yield String(s.s), (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 from capa.features.extractors.common.extract_file_strings(seg_buff)
 
 
-def extract_features():
-    """ extract file features """
+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), 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() -> Iterator[Tuple[Feature, Address]]:
+    file_info = idaapi.get_inf_structure()
+
+    if file_info.filetype == idaapi.f_PE:
+        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: %d" % file_info.filetype)
+
+
+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 = (
@@ -156,6 +183,8 @@ FILE_HANDLERS = (
     extract_file_strings,
     extract_file_section_names,
     extract_file_embedded_pe,
+    extract_file_function_names,
+    extract_file_format,
 )
 
 

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 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
@@ -41,25 +37,16 @@ def extract_function_loop(f):
         yield Characteristic("loop"), f.start_ea
 
 
-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

@@ -0,0 +1,58 @@
+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, Feature
+from capa.features.address import NO_ADDRESS, Address
+
+logger = logging.getLogger(__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), 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), NO_ADDRESS
+
+    else:
+        # we likely end up here:
+        #  1. handling shellcode, or
+        #  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,
+        # but i think this would be rarely used.
+        # rules that rely on OS conditions will fail to match on shellcode.
+        #
+        # for (2), this logic will need to be updated as the format is implemented.
+        logger.debug("unsupported file format: %s, will not guess OS", format_name)
+        return
+
+
+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:
+        # 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", info.procname)
+        return

capa/features/extractors/ida/helpers.py

@@ -1,21 +1,22 @@
-# 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 sys
-import string
+from typing import Any, Dict, Tuple, Iterator
 
 import idc
 import idaapi
 import idautils
 import ida_bytes
 
+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:
@@ -23,36 +24,32 @@ def find_byte_sequence(start, end, seq):
         end: max virtual address
         seq: bytes to search e.g. b"\x01\x03"
     """
-    if sys.version_info[0] >= 3:
-        seq = " ".join(["%02x" % b for b in seq])
-    else:
-        seq = " ".join(["%02x" % ord(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: int = None, end: 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:
@@ -64,7 +61,7 @@ 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
@@ -82,8 +79,8 @@ def get_segment_buffer(seg):
     return buff if buff else b""
 
 
-def get_file_imports():
-    """ get file imports """
+def get_file_imports() -> Dict[int, Tuple[str, str, int]]:
+    """get file imports"""
     imports = {}
 
     for idx in range(idaapi.get_import_module_qty()):
@@ -92,10 +89,18 @@ 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
 
@@ -104,14 +109,12 @@ def get_file_imports():
     return imports
 
 
-def get_instructions_in_range(start, end):
+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)
@@ -119,8 +122,8 @@ def get_instructions_in_range(start, end):
             yield insn
 
 
-def is_operand_equal(op1, op2):
-    """ compare two IDA op_t """
+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
 
@@ -145,8 +148,8 @@ def is_operand_equal(op1, op2):
     return True
 
 
-def is_basic_block_equal(bb1, bb2):
-    """ compare two IDA BasicBlock """
+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
 
@@ -159,12 +162,12 @@ def is_basic_block_equal(bb1, bb2):
     return True
 
 
-def basic_block_size(bb):
-    """ calculate size of basic block """
+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):
@@ -177,10 +180,10 @@ def read_bytes_at(ea, count):
         return idc.get_bytes(ea, count)
 
 
-def find_string_at(ea, min=4):
-    """ check if ASCII string exists at a given virtual address """
+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
@@ -194,7 +197,7 @@ 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:
@@ -231,24 +234,24 @@ def get_op_phrase_info(op):
     return {"base": base, "index": index, "scale": scale, "offset": offset}
 
 
-def is_op_write(insn, op):
-    """ Check if an operand is written to (destination operand) """
+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):
-    """ Check if an operand is read from (source operand) """
+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):
-    """ Check is an operand has been marked as an offset (by auto-analysis or manually) """
+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):
-    """ determine if instruction modifies SP, ESP, RSP """
+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):
             # register is stack and written
@@ -256,8 +259,8 @@ def is_sp_modified(insn):
     return False
 
 
-def is_bp_modified(insn):
-    """ check if instruction modifies BP, EBP, RBP """
+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):
             # register is base and written
@@ -265,13 +268,13 @@ def is_bp_modified(insn):
     return False
 
 
-def is_frame_register(reg):
-    """ check if register is sp or bp """
+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=()):
-    """ yield op_t for instruction, filter on type if specified """
+def get_insn_ops(insn: idaapi.insn_t, target_ops: 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:
             # avoid looping all 6 ops if only subset exists
@@ -281,12 +284,12 @@ def get_insn_ops(insn, target_ops=()):
         yield op
 
 
-def is_op_stack_var(ea, index):
-    """ check if operand is a stack variable """
+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
@@ -306,26 +309,18 @@ 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:
@@ -335,8 +330,8 @@ def is_basic_block_tight_loop(bb):
     return False
 
 
-def find_data_reference_from_insn(insn, max_depth=10):
-    """ search for data reference from instruction, return address of instruction if no reference exists """
+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
 
@@ -365,19 +360,18 @@ 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):
-    """ check if basic block is return block """
+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
@@ -12,51 +13,24 @@ import idautils
 
 import capa.features.extractors.helpers
 import capa.features.extractors.ida.helpers
-from capa.features import (
-    ARCH_X32,
-    ARCH_X64,
-    MAX_BYTES_FEATURE_SIZE,
-    THUNK_CHAIN_DEPTH_DELTA,
-    Bytes,
-    String,
-    Characteristic,
-)
-from capa.features.insn import API, Number, Offset, Mnemonic
+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_arch(ctx):
-    """
-    fetch the ARCH_* constant for the currently open workspace.
-
-    via Tamir Bahar/@tmr232
-    https://reverseengineering.stackexchange.com/a/11398/17194
-    """
-    if "arch" not in ctx:
-        info = idaapi.get_inf_structure()
-        if info.is_64bit():
-            ctx["arch"] = ARCH_X64
-        elif info.is_32bit():
-            ctx["arch"] = ARCH_X32
-        else:
-            raise ValueError("unexpected architecture")
-    return ctx["arch"]
-
-
-def get_imports(ctx):
+def get_imports(ctx: Dict[str, Any]) -> Dict[str, 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):
-    """ check instruction for API call """
-    if not insn.get_canon_mnem() in ("call", "jmp"):
-        return
-
+def check_for_api_call(ctx: Dict[str, Any], insn: idaapi.insn_t) -> Iterator[str]:
+    """check instruction for API call"""
     info = ()
     ref = insn.ea
 
@@ -84,34 +58,55 @@ def check_for_api_call(ctx, insn):
         yield "%s.%s" % (info[0], info[1])
 
 
-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]
     """
-    for api in check_for_api_call(f.ctx, insn):
+    insn: idaapi.insn_t = ih.inner
+
+    if not insn.get_canon_mnem() in ("call", "jmp"):
+        return
+
+    for api in check_for_api_call(fh.ctx, insn):
         dll, _, symbol = api.rpartition(".")
         for name in capa.features.extractors.helpers.generate_symbols(dll, symbol):
-            yield API(name), insn.ea
+            yield API(name), ih.address
+
+    # extract IDA/FLIRT recognized API functions
+    targets = tuple(idautils.CodeRefsFrom(insn.ea, False))
+    if not targets:
+        return
+
+    target = targets[0]
+    target_func = idaapi.get_func(target)
+    if not target_func or target_func.start_ea != target:
+        # not a function (start)
+        return
+
+    if target_func.flags & idaapi.FUNC_LIB:
+        name = idaapi.get_name(target_func.start_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
@@ -122,7 +117,11 @@ 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):
@@ -133,21 +132,27 @@ def extract_insn_number_features(f, bb, insn):
         else:
             const = op.addr
 
-        yield Number(const), insn.ea
-        yield Number(const, arch=get_arch(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
 
@@ -155,41 +160,46 @@ 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
+            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)
         if idaapi.is_mapped(op_off):
@@ -202,12 +212,32 @@ 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, arch=get_arch(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
@@ -221,7 +251,7 @@ 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
@@ -255,8 +285,8 @@ def bb_stack_cookie_registers(bb):
                     yield op.reg
 
 
-def is_nzxor_stack_cookie_delta(f, bb, insn):
-    """ check if nzxor exists within stack cookie delta """
+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):
         return False
@@ -278,8 +308,8 @@ def is_nzxor_stack_cookie_delta(f, bb, insn):
     return False
 
 
-def is_nzxor_stack_cookie(f, bb, insn):
-    """ check if nzxor is related to stack cookie """
+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:
         #   xor     ecx, ebp        ; StackCookie
@@ -295,37 +325,49 @@ 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(insn.get_canon_mnem()), 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
@@ -333,6 +375,8 @@ 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
 
@@ -344,15 +388,19 @@ 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
@@ -361,23 +409,21 @@ 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):
@@ -385,50 +431,40 @@ 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
 
 
@@ -440,6 +476,7 @@ 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
@@ -6,7 +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.
 
-from networkx import nx
+import networkx
 from networkx.algorithms.components import strongly_connected_components
 
 
@@ -20,6 +20,6 @@ def has_loop(edges, threshold=2):
     returns:
         bool
     """
-    g = nx.DiGraph()
+    g = networkx.DiGraph()
     g.add_edges_from(edges)
     return any(len(comp) >= threshold for comp in strongly_connected_components(g))

capa/features/extractors/null.py

@@ -0,0 +1,77 @@
+from typing import Dict, List, Tuple
+from dataclasses import dataclass
+
+from capa.features.common import Feature
+from capa.features.address import 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 address, feature in self.global_features:
+            yield feature, 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.get(f.address, {}).features:
+            yield feature, address
+
+    def get_basic_blocks(self, f):
+        for address in sorted(self.functions.get(f.address, {}).basic_blocks.keys()):
+            yield BBHandle(address, None)
+
+    def extract_basic_block_features(self, f, bb):
+        for address, feature in self.functions.get(f.address, {}).basic_blocks.get(bb.address, {}).features:
+            yield feature, address
+
+    def get_instructions(self, f, bb):
+        for address in sorted(self.functions.get(f.address, {}).basic_blocks.get(bb.address, {}).instructions.keys()):
+            yield InsnHandle(address, None)
+
+    def extract_insn_features(self, f, bb, insn):
+        for address, feature in (
+            self.functions.get(f.address, {})
+            .basic_blocks.get(bb.address, {})
+            .instructions.get(insn.address, {})
+            .features
+        ):
+            yield feature, address

capa/features/extractors/pefile.py

@@ -0,0 +1,216 @@
+# 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 pefile
+
+import capa.features.common
+import capa.features.extractors
+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 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__)
+
+
+def extract_file_embedded_pe(buf, **kwargs):
+    for offset, _ in capa.features.extractors.helpers.carve_pe(buf, 1):
+        yield Characteristic("embedded pe"), FileOffsetAddress(offset)
+
+
+def extract_file_export_names(pe, **kwargs):
+    base_address = pe.OPTIONAL_HEADER.ImageBase
+
+    if hasattr(pe, "DIRECTORY_ENTRY_EXPORT"):
+        for export in pe.DIRECTORY_ENTRY_EXPORT.symbols:
+            if not export.name:
+                continue
+            try:
+                name = export.name.partition(b"\x00")[0].decode("ascii")
+            except UnicodeDecodeError:
+                continue
+            va = base_address + export.address
+            yield Export(name), AbsoluteVirtualAddress(va)
+
+
+def extract_file_import_names(pe, **kwargs):
+    """
+    extract imported function names
+    1. imports by ordinal:
+     - modulename.#ordinal
+    2. imports by name, results in two features to support importname-only matching:
+     - modulename.importname
+     - importname
+    """
+    if hasattr(pe, "DIRECTORY_ENTRY_IMPORT"):
+        for dll in pe.DIRECTORY_ENTRY_IMPORT:
+            try:
+                modname = dll.dll.partition(b"\x00")[0].decode("ascii")
+            except UnicodeDecodeError:
+                continue
+
+            # strip extension
+            modname = modname.rpartition(".")[0].lower()
+
+            for imp in dll.imports:
+                if imp.import_by_ordinal:
+                    impname = "#%s" % imp.ordinal
+                else:
+                    try:
+                        impname = imp.name.partition(b"\x00")[0].decode("ascii")
+                    except UnicodeDecodeError:
+                        continue
+
+                for name in capa.features.extractors.helpers.generate_symbols(modname, impname):
+                    yield Import(name), AbsoluteVirtualAddress(imp.address)
+
+
+def extract_file_section_names(pe, **kwargs):
+    base_address = pe.OPTIONAL_HEADER.ImageBase
+
+    for section in pe.sections:
+        try:
+            name = section.Name.partition(b"\x00")[0].decode("ascii")
+        except UnicodeDecodeError:
+            continue
+
+        yield Section(name), AbsoluteVirtualAddress(base_address + section.VirtualAddress)
+
+
+def extract_file_strings(buf, **kwargs):
+    yield from capa.features.extractors.common.extract_file_strings(buf)
+
+
+def extract_file_function_names(**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("pefile doesn't have library matching")
+    return
+
+
+def extract_file_os(**kwargs):
+    # assuming PE -> Windows
+    # though i suppose they're also used by UEFI
+    yield OS(OS_WINDOWS), NO_ADDRESS
+
+
+def extract_file_format(**kwargs):
+    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), NO_ADDRESS
+    elif pe.FILE_HEADER.Machine == pefile.MACHINE_TYPE["IMAGE_FILE_MACHINE_AMD64"]:
+        yield Arch(ARCH_AMD64), NO_ADDRESS
+    else:
+        logger.warning("unsupported architecture: %s", pefile.MACHINE_TYPE[pe.FILE_HEADER.Machine])
+
+
+def extract_file_features(pe, buf):
+    """
+    extract file features from given workspace
+
+    args:
+      pe (pefile.PE): the parsed PE
+      buf: the raw sample bytes
+
+    yields:
+      Tuple[Feature, VA]: a feature and its location.
+    """
+
+    for file_handler in FILE_HANDLERS:
+        for feature, va in file_handler(pe=pe, 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,
+)
+
+
+def extract_global_features(pe, buf):
+    """
+    extract global features from given workspace
+
+    args:
+      pe (pefile.PE): the parsed PE
+      buf: the raw sample bytes
+
+    yields:
+      Tuple[Feature, VA]: a feature and its location.
+    """
+    for handler in GLOBAL_HANDLERS:
+        for feature, va in handler(pe=pe, buf=buf):
+            yield feature, va
+
+
+GLOBAL_HANDLERS = (
+    extract_file_os,
+    extract_file_arch,
+)
+
+
+class PefileFeatureExtractor(FeatureExtractor):
+    def __init__(self, path: str):
+        super(PefileFeatureExtractor, self).__init__()
+        self.path = path
+        self.pe = pefile.PE(path)
+
+    def get_base_address(self):
+        return AbsoluteVirtualAddress(self.pe.OPTIONAL_HEADER.ImageBase)
+
+    def extract_global_features(self):
+        with open(self.path, "rb") as f:
+            buf = f.read()
+
+        yield from extract_global_features(self.pe, buf)
+
+    def extract_file_features(self):
+        with open(self.path, "rb") as f:
+            buf = f.read()
+
+        yield from extract_file_features(self.pe, buf)
+
+    def get_functions(self):
+        raise NotImplementedError("PefileFeatureExtract can only be used to extract file features")
+
+    def extract_function_features(self, f):
+        raise NotImplementedError("PefileFeatureExtract can only be used to extract file features")
+
+    def get_basic_blocks(self, f):
+        raise NotImplementedError("PefileFeatureExtract can only be used to extract file features")
+
+    def extract_basic_block_features(self, f, bb):
+        raise NotImplementedError("PefileFeatureExtract can only be used to extract file features")
+
+    def get_instructions(self, f, bb):
+        raise NotImplementedError("PefileFeatureExtract can only be used to extract file features")
+
+    def extract_insn_features(self, f, bb, insn):
+        raise NotImplementedError("PefileFeatureExtract can only be used to extract file features")
+
+    def is_library_function(self, va):
+        raise NotImplementedError("PefileFeatureExtract can only be used to extract file features")
+
+    def get_function_name(self, va):
+        raise NotImplementedError("PefileFeatureExtract can only be used to extract file features")

capa/features/extractors/smda/__init__.py

@@ -1,52 +0,0 @@
-import sys
-import types
-
-from smda.common.SmdaReport import SmdaReport
-from smda.common.SmdaInstruction import SmdaInstruction
-
-import capa.features.extractors.smda.file
-import capa.features.extractors.smda.insn
-import capa.features.extractors.smda.function
-import capa.features.extractors.smda.basicblock
-from capa.main import UnsupportedRuntimeError
-from capa.features.extractors import FeatureExtractor
-
-
-class SmdaFeatureExtractor(FeatureExtractor):
-    def __init__(self, smda_report: SmdaReport, path):
-        super(SmdaFeatureExtractor, self).__init__()
-        if sys.version_info < (3, 0):
-            raise UnsupportedRuntimeError("SMDA should only be used with Python 3.")
-        self.smda_report = smda_report
-        self.path = path
-
-    def get_base_address(self):
-        return self.smda_report.base_addr
-
-    def extract_file_features(self):
-        for feature, va in capa.features.extractors.smda.file.extract_features(self.smda_report, self.path):
-            yield feature, va
-
-    def get_functions(self):
-        for function in self.smda_report.getFunctions():
-            yield function
-
-    def extract_function_features(self, f):
-        for feature, va in capa.features.extractors.smda.function.extract_features(f):
-            yield feature, va
-
-    def get_basic_blocks(self, f):
-        for bb in f.getBlocks():
-            yield bb
-
-    def extract_basic_block_features(self, f, bb):
-        for feature, va in capa.features.extractors.smda.basicblock.extract_features(f, bb):
-            yield feature, va
-
-    def get_instructions(self, f, bb):
-        for smda_ins in bb.getInstructions():
-            yield smda_ins
-
-    def extract_insn_features(self, f, bb, insn):
-        for feature, va in capa.features.extractors.smda.insn.extract_features(f, bb, insn):
-            yield feature, va

capa/features/extractors/smda/basicblock.py

@@ -1,10 +1,12 @@
-import sys
 import string
 import struct
+from typing import Tuple, Iterator
 
-from capa.features 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.helpers import MIN_STACKSTRING_LEN
+from capa.features.extractors.base_extractor import BBHandle, FunctionHandle
 
 
 def _bb_has_tight_loop(f, bb):
@@ -14,10 +16,10 @@ def _bb_has_tight_loop(f, bb):
     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 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.inner, bb.inner):
+        yield Characteristic("tight loop"), bb.address
 
 
 def _bb_has_stackstring(f, bb):
@@ -38,10 +40,10 @@ 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 extract_stackstring(f: FunctionHandle, bb: BBHandle) -> Iterator[Tuple[Feature, Address]]:
+    """check basic block for stackstring indicators"""
+    if _bb_has_stackstring(f.inner, bb.inner):
+        yield Characteristic("stack string"), bb.address
 
 
 def is_mov_imm_to_stack(smda_ins):
@@ -108,21 +110,21 @@ def get_printable_len(instr):
     return 0
 
 
-def extract_features(f, bb):
+def extract_features(f: FunctionHandle, bb: BBHandle) -> Iterator[Tuple[Feature, Address]]:
     """
     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.
+      f: the function from which to extract features
+      bb: the basic block to process.
 
     yields:
-      Feature, set[VA]: the features and their location found in this basic block.
+      Tuple[Feature, Address]: the features and their location found in this basic block.
     """
-    yield BasicBlock(), bb.offset
+    yield BasicBlock(), bb.address
     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/smda/extractor.py

@@ -0,0 +1,57 @@
+from typing import List, Tuple
+
+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.common import Feature
+from capa.features.address import Address, AbsoluteVirtualAddress
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle, 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: List[Tuple[Feature, Address]] = []
+        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 AbsoluteVirtualAddress(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 FunctionHandle(address=AbsoluteVirtualAddress(function.offset), inner=function)
+
+    def extract_function_features(self, fh):
+        yield from capa.features.extractors.smda.function.extract_features(fh)
+
+    def get_basic_blocks(self, fh):
+        for bb in fh.inner.getBlocks():
+            yield BBHandle(address=AbsoluteVirtualAddress(bb.offset), inner=bb)
+
+    def extract_basic_block_features(self, fh, bbh):
+        yield from capa.features.extractors.smda.basicblock.extract_features(fh, bbh)
+
+    def get_instructions(self, fh, bbh):
+        for smda_ins in bbh.inner.getInstructions():
+            yield InsnHandle(address=AbsoluteVirtualAddress(smda_ins.offset), inner=smda_ins)
+
+    def extract_insn_features(self, fh, bbh, ih):
+        yield from capa.features.extractors.smda.insn.extract_features(fh, bbh, ih)

capa/features/extractors/smda/file.py

@@ -1,133 +1,95 @@
-import struct
-
 # 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 import String, Characteristic
 from capa.features.file import Export, Import, Section
+from capa.features.common import String, Characteristic
+from capa.features.address import FileOffsetAddress, AbsoluteVirtualAddress
 
 
-def carve(pbytes, offset=0):
-    """
-    Return a list of (offset, size, xor) tuples of embedded PEs
-
-    Based on the version from vivisect:
-    https://github.com/vivisect/vivisect/blob/7be4037b1cecc4551b397f840405a1fc606f9b53/PE/carve.py#L19
-    And its IDA adaptation:
-    capa/features/extractors/ida/file.py
-    """
-    mz_xor = [
-        (
-            capa.features.extractors.helpers.xor_static(b"MZ", i),
-            capa.features.extractors.helpers.xor_static(b"PE", i),
-            i,
-        )
-        for i in range(256)
-    ]
-
-    pblen = len(pbytes)
-    todo = [(pbytes.find(mzx, offset), mzx, pex, i) for mzx, pex, i in mz_xor]
-    todo = [(off, mzx, pex, i) for (off, mzx, pex, i) in todo if off != -1]
-
-    while len(todo):
-
-        off, mzx, pex, i = todo.pop()
-
-        # The MZ header has one field we will check
-        # e_lfanew is at 0x3c
-        e_lfanew = off + 0x3C
-        if pblen < (e_lfanew + 4):
-            continue
-
-        newoff = struct.unpack("<I", capa.features.extractors.helpers.xor_static(pbytes[e_lfanew : e_lfanew + 4], i))[0]
-
-        nextres = pbytes.find(mzx, off + 1)
-        if nextres != -1:
-            todo.append((nextres, mzx, pex, i))
-
-        peoff = off + newoff
-        if pblen < (peoff + 2):
-            continue
-
-        if pbytes[peoff : peoff + 2] == pex:
-            yield (off, i)
+def extract_file_embedded_pe(buf, **kwargs):
+    for offset, _ in capa.features.extractors.helpers.carve_pe(buf, 1):
+        yield Characteristic("embedded pe"), FileOffsetAddress(offset)
 
 
-def extract_file_embedded_pe(smda_report, file_path):
-    with open(file_path, "rb") as f:
-        fbytes = f.read()
+def extract_file_export_names(buf, **kwargs):
+    lief_binary = lief.parse(buf)
 
-    for offset, i in carve(fbytes, 1):
-        yield Characteristic("embedded pe"), offset
-
-
-def extract_file_export_names(smda_report, file_path):
-    lief_binary = lief.parse(file_path)
     if lief_binary is not None:
         for function in lief_binary.exported_functions:
-            yield Export(function.name), function.address
+            yield Export(function.name), AbsoluteVirtualAddress(function.address)
 
 
-def extract_file_import_names(smda_report, file_path):
+def extract_file_import_names(smda_report, buf):
     # extract import table info via LIEF
-    lief_binary = lief.parse(file_path)
+    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:
-                va = func.iat_address + smda_report.base_addr
                 for name in capa.features.extractors.helpers.generate_symbols(library_name, func.name):
-                    yield Import(name), va
+                    yield Import(name), AbsoluteVirtualAddress(va)
             elif func.is_ordinal:
                 for name in capa.features.extractors.helpers.generate_symbols(library_name, "#%s" % func.ordinal):
-                    yield Import(name), va
+                    yield Import(name), AbsoluteVirtualAddress(va)
 
 
-def extract_file_section_names(smda_report, file_path):
-    lief_binary = lief.parse(file_path)
+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
+            yield Section(section.name), AbsoluteVirtualAddress(base_address + section.virtual_address)
 
 
-def extract_file_strings(smda_report, file_path):
+def extract_file_strings(buf, **kwargs):
     """
     extract ASCII and UTF-16 LE strings from file
     """
-    with open(file_path, "rb") as f:
-        b = f.read()
+    for s in capa.features.extractors.strings.extract_ascii_strings(buf):
+        yield String(s.s), FileOffsetAddress(s.offset)
 
-    for s in capa.features.extractors.strings.extract_ascii_strings(b):
-        yield String(s.s), s.offset
-
-    for s in capa.features.extractors.strings.extract_unicode_strings(b):
-        yield String(s.s), s.offset
+    for s in capa.features.extractors.strings.extract_unicode_strings(buf):
+        yield String(s.s), FileOffsetAddress(s.offset)
 
 
-def extract_features(smda_report, file_path):
+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
-      file_path: path to the input file
+      buf: the raw bytes of the sample
 
     yields:
       Tuple[Feature, VA]: a feature and its location.
     """
 
     for file_handler in FILE_HANDLERS:
-        result = file_handler(smda_report, file_path)
-        for feature, va in file_handler(smda_report, file_path):
-            yield feature, va
+        for feature, addr in file_handler(smda_report=smda_report, buf=buf):
+            yield feature, addr
 
 
 FILE_HANDLERS = (
@@ -136,4 +98,6 @@ FILE_HANDLERS = (
     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 +1,42 @@
-from capa.features import Characteristic
+from typing import Tuple, Iterator
+
+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):
-    for inref in f.inrefs:
-        yield Characteristic("calls to"), inref
+def extract_function_calls_to(f: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
+    for inref in f.inner.inrefs:
+        yield Characteristic("calls to"), AbsoluteVirtualAddress(inref)
 
 
-def extract_function_loop(f):
+def extract_function_loop(f: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
     """
     parse if a function has a loop
     """
     edges = []
-    for bb_from, bb_tos in f.blockrefs.items():
+    for bb_from, bb_tos in f.inner.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
+        yield Characteristic("loop"), f.address
 
 
-def extract_features(f):
+def extract_features(f: FunctionHandle):
     """
     extract features from the given function.
 
     args:
-      f (smda.common.SmdaFunction): the function from which to extract features
+      f: the function from which to extract features
 
     yields:
-      Feature, set[VA]: the features and their location found in this function.
+      Tuple[Feature, Address]: 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(f):
+            yield feature, addr
 
 
 FUNCTION_HANDLERS = (extract_function_calls_to, extract_function_loop)

capa/features/extractors/smda/global_.py

@@ -0,0 +1,21 @@
+import logging
+
+from capa.features.common import ARCH_I386, ARCH_AMD64, Arch
+from capa.features.address import NO_ADDRESS
+
+logger = logging.getLogger(__name__)
+
+
+def extract_arch(smda_report):
+    if smda_report.architecture == "intel":
+        if smda_report.bitness == 32:
+            yield Arch(ARCH_I386), NO_ADDRESS
+        elif smda_report.bitness == 64:
+            yield Arch(ARCH_AMD64), NO_ADDRESS
+    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

@@ -1,20 +1,15 @@
 import re
 import string
 import struct
+from typing import Tuple, Iterator
 
-from smda.common.SmdaReport import SmdaReport
+import smda
 
 import capa.features.extractors.helpers
-from capa.features import (
-    ARCH_X32,
-    ARCH_X64,
-    MAX_BYTES_FEATURE_SIZE,
-    THUNK_CHAIN_DEPTH_DELTA,
-    Bytes,
-    String,
-    Characteristic,
-)
-from capa.features.insn import API, Number, Offset, Mnemonic
+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
@@ -23,27 +18,20 @@ PATTERN_HEXNUM = re.compile(r"[+\-] (?P<num>0x[a-fA-F0-9]+)")
 PATTERN_SINGLENUM = re.compile(r"[+\-] (?P<num>[0-9])")
 
 
-def get_arch(smda_report):
-    if smda_report.architecture == "intel":
-        if smda_report.bitness == 32:
-            return ARCH_X32
-        elif smda_report.bitness == 64:
-            return ARCH_X64
-    else:
-        raise NotImplementedError
-
-
-def extract_insn_api_features(f, bb, insn):
+def extract_insn_api_features(fh: FunctionHandle, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
     """parse API features from the given instruction."""
-    if insn.offset in f.apirefs:
-        api_entry = f.apirefs[insn.offset]
+    f: smda.Function = fh.inner
+    insn: smda.Insn = ih.inner
+
+    if ih.address in f.apirefs:
+        api_entry = f.apirefs[ih.address]
         # 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:
+            yield API(name), ih.address
+    elif ih.address in f.outrefs:
         current_function = f
         current_instruction = insn
         for index in range(THUNK_CHAIN_DEPTH_DELTA):
@@ -62,7 +50,7 @@ def extract_insn_api_features(f, bb, insn):
                             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
+                                yield API(name), ih.address
                     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]
@@ -70,11 +58,14 @@ def extract_insn_api_features(f, bb, insn):
                     return
 
 
-def extract_insn_number_features(f, bb, insn):
+def extract_insn_number_features(fh: FunctionHandle, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
     """parse number features from the given instruction."""
     # example:
     #
     #     push    3136B0h         ; dwControlCode
+    f: smda.Function = fh.inner
+    insn: smda.Insn = ih.inner
+
     operands = [o.strip() for o in insn.operands.split(",")]
     if insn.mnemonic == "add" and operands[0] in ["esp", "rsp"]:
         # skip things like:
@@ -82,12 +73,25 @@ def extract_insn_number_features(f, bb, insn):
         #    .text:00401140                 call    sub_407E2B
         #    .text:00401145                 add     esp, 0Ch
         return
-    for operand in operands:
+    for i, operand in enumerate(operands):
         try:
-            yield Number(int(operand, 16)), insn.offset
-            yield Number(int(operand, 16), arch=get_arch(f.smda_report)), insn.offset
-        except:
+            # The result of bitwise operations is calculated as though carried out
+            # in two’s complement with an infinite number of sign bits
+            value = int(operand, 16) & ((1 << f.smda_report.bitness) - 1)
+        except ValueError:
             continue
+        else:
+            yield Number(value), ih.address
+            yield OperandNumber(i, value), ih.address
+
+            if insn.mnemonic == "add" and 0 < value < MAX_STRUCTURE_SIZE:
+                # for pattern like:
+                #
+                #     add eax, 0x10
+                #
+                # assume 0x10 is also an offset (imagine eax is a pointer).
+                yield Offset(value), ih.address
+                yield OperandOffset(i, value), ih.address
 
 
 def read_bytes(smda_report, va, num_bytes=None):
@@ -97,7 +101,7 @@ def read_bytes(smda_report, va, num_bytes=None):
 
     rva = va - smda_report.base_addr
     if smda_report.buffer is None:
-        return
+        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:
@@ -136,12 +140,15 @@ def derefs(smda_report, p):
         p = val
 
 
-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
     """
+    f: smda.Function = fh.inner
+    insn: smda.Insn = ih.inner
+
     for data_ref in insn.getDataRefs():
         for v in derefs(f.smda_report, data_ref):
             bytes_read = read_bytes(f.smda_report, v)
@@ -150,7 +157,7 @@ def extract_insn_bytes_features(f, bb, insn):
             if capa.features.extractors.helpers.all_zeros(bytes_read):
                 continue
 
-            yield Bytes(bytes_read), insn.offset
+            yield Bytes(bytes_read), ih.address
 
 
 def detect_ascii_len(smda_report, offset):
@@ -194,30 +201,34 @@ def read_string(smda_report, offset):
         return read_bytes(smda_report, offset, ulen).decode("utf-16")
 
 
-def extract_insn_string_features(f, bb, insn):
+def extract_insn_string_features(fh: FunctionHandle, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
     """parse string features from the given instruction."""
     # example:
     #
     #     push    offset aAcr     ; "ACR  > "
+    f: smda.Function = fh.inner
+    insn: smda.Insn = ih.inner
+
     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
+                yield String(string_read.rstrip("\x00")), ih.address
 
 
-def extract_insn_offset_features(f, bb, insn):
+def extract_insn_offset_features(f, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
     """parse structure offset features from the given instruction."""
     # examples:
     #
     #     mov eax, [esi + 4]
     #     mov eax, [esi + ecx + 16384]
+    insn: smda.Insn = ih.inner
+
     operands = [o.strip() for o in insn.operands.split(",")]
-    for operand in operands:
-        if not "ptr" in operand:
-            continue
+    for i, operand in enumerate(operands):
         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)
@@ -227,8 +238,26 @@ def extract_insn_offset_features(f, bb, insn):
         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, arch=get_arch(f.smda_report)), insn.offset
+
+        if "ptr" not in operand:
+            if (
+                insn.mnemonic == "lea"
+                and i == 1
+                and (operand.count("+") + operand.count("-")) == 1
+                and operand.count("*") == 0
+            ):
+                # for pattern like:
+                #
+                #     lea eax, [ebx + 1]
+                #
+                # assume 1 is also an offset (imagine ebx is a zero register).
+                yield Number(number), ih.address
+                yield OperandNumber(i, number), ih.address
+
+            continue
+
+        yield Offset(number), ih.address
+        yield OperandOffset(i, number), ih.address
 
 
 def is_security_cookie(f, bb, insn):
@@ -252,11 +281,16 @@ def is_security_cookie(f, bb, insn):
     return False
 
 
-def extract_insn_nzxor_characteristic_features(f, bb, insn):
+def extract_insn_nzxor_characteristic_features(
+    fh: FunctionHandle, bh: 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.
     """
+    f: smda.Function = fh.inner
+    bb: smda.BasicBlock = bh.inner
+    insn: smda.Insn = ih.inner
 
     if insn.mnemonic not in ("xor", "xorpd", "xorps", "pxor"):
         return
@@ -268,18 +302,35 @@ def extract_insn_nzxor_characteristic_features(f, bb, insn):
     if is_security_cookie(f, bb, insn):
         return
 
-    yield Characteristic("nzxor"), insn.offset
+    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.mnemonic), insn.offset
+    yield Mnemonic(ih.inner.mnemonic), 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: smda.Insn = ih.inner
+
+    if insn.mnemonic != "call":
+        return
+
+    if not insn.operands.startswith("0x"):
+        return
+
+    if int(insn.operands, 16) == insn.offset + 5:
+        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
     """
+    insn: smda.Insn = ih.inner
 
     if insn.mnemonic not in ["push", "mov"]:
         return
@@ -287,65 +338,75 @@ def extract_insn_peb_access_characteristic_features(f, bb, insn):
     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
+            yield Characteristic("peb access"), ih.address
         elif "gs:" in operand and "0x60" in operand:
-            yield Characteristic("peb access"), insn.offset
+            yield Characteristic("peb access"), ih.address
 
 
-def extract_insn_segment_access_features(f, bb, insn):
-    """ parse the instruction for access to fs or gs """
+def extract_insn_segment_access_features(f, bb, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    """parse the instruction for access to fs or gs"""
+    insn: smda.Insn = ih.inner
+
     operands = [o.strip() for o in insn.operands.split(",")]
     for operand in operands:
         if "fs:" in operand:
-            yield Characteristic("fs access"), insn.offset
+            yield Characteristic("fs access"), ih.address
         elif "gs:" in operand:
-            yield Characteristic("gs access"), insn.offset
+            yield Characteristic("gs access"), ih.address
 
 
-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.
     """
+    f: smda.Function = fh.inner
+    insn: smda.Insn = ih.inner
+
     if insn.mnemonic in ["call", "jmp"]:
-        if insn.offset in f.apirefs:
+        if ih.address 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
+        if ih.address in f.outrefs:
+            for target in f.outrefs[ih.address]:
+                if smda_report.getSection(ih.address) != smda_report.getSection(target):
+                    yield Characteristic("cross section flow"), ih.address
         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
+            if smda_report.getSection(ih.address) != smda_report.getSection(target):
+                yield Characteristic("cross section flow"), ih.address
 
 
 # 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]]:
+    f: smda.Function = fh.inner
+    insn: smda.Insn = ih.inner
+
     if insn.mnemonic != "call":
         return
 
-    if insn.offset in f.outrefs:
-        for outref in f.outrefs[insn.offset]:
-            yield Characteristic("calls from"), outref
+    if ih.address in f.outrefs:
+        for outref in f.outrefs[ih.address]:
+            yield Characteristic("calls from"), AbsoluteVirtualAddress(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
+                yield Characteristic("recursive call"), AbsoluteVirtualAddress(outref)
+    if ih.address in f.apirefs:
+        yield Characteristic("calls from"), ih.address
 
 
 # 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: smda.Insn = ih.inner
+
     if insn.mnemonic != "call":
         return
     if insn.operands.startswith("0x"):
@@ -357,7 +418,7 @@ def extract_function_indirect_call_characteristic_features(f, bb, insn):
     # call edx
     # call dword ptr [eax+50h]
     # call qword ptr [rsp+78h]
-    yield Characteristic("indirect call"), insn.offset
+    yield Characteristic("indirect call"), ih.address
 
 
 def extract_features(f, bb, insn):
@@ -365,16 +426,16 @@ 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.
+      f: the function to process.
+      bb: the basic block to process.
+      insn: the instruction to process.
 
     yields:
-      Feature, set[VA]: 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 = (
@@ -385,6 +446,7 @@ 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/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/__init__.py

@@ -1,81 +0,0 @@
-# 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 types
-
-import viv_utils
-
-import capa.features.extractors
-import capa.features.extractors.viv.file
-import capa.features.extractors.viv.insn
-import capa.features.extractors.viv.function
-import capa.features.extractors.viv.basicblock
-from capa.features.extractors import FeatureExtractor
-
-__all__ = ["file", "function", "basicblock", "insn"]
-
-
-def get_va(self):
-    try:
-        # vivisect type
-        return self.va
-    except AttributeError:
-        pass
-
-    raise TypeError()
-
-
-def add_va_int_cast(o):
-    """
-    dynamically add a cast-to-int (`__int__`) method to the given object
-    that returns the value of the `.va` property.
-
-    this bit of skullduggery lets use cast viv-utils objects as ints.
-    the correct way of doing this is to update viv-utils (or subclass the objects here).
-    """
-    setattr(o, "__int__", types.MethodType(get_va, o))
-    return o
-
-
-class VivisectFeatureExtractor(FeatureExtractor):
-    def __init__(self, vw, path):
-        super(VivisectFeatureExtractor, self).__init__()
-        self.vw = vw
-        self.path = path
-
-    def get_base_address(self):
-        # assume there is only one file loaded into the vw
-        return list(self.vw.filemeta.values())[0]["imagebase"]
-
-    def extract_file_features(self):
-        for feature, va in capa.features.extractors.viv.file.extract_features(self.vw, self.path):
-            yield feature, va
-
-    def get_functions(self):
-        for va in sorted(self.vw.getFunctions()):
-            yield add_va_int_cast(viv_utils.Function(self.vw, va))
-
-    def extract_function_features(self, f):
-        for feature, va in capa.features.extractors.viv.function.extract_features(f):
-            yield feature, va
-
-    def get_basic_blocks(self, f):
-        for bb in f.basic_blocks:
-            yield add_va_int_cast(bb)
-
-    def extract_basic_block_features(self, f, bb):
-        for feature, va in capa.features.extractors.viv.basicblock.extract_features(f, bb):
-            yield feature, va
-
-    def get_instructions(self, f, bb):
-        for insn in bb.instructions:
-            yield add_va_int_cast(insn)
-
-    def extract_insn_features(self, f, bb, insn):
-        for feature, va in capa.features.extractors.viv.insn.extract_features(f, bb, insn):
-            yield feature, va

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 vivisect.const
+import envi.archs.i386.disasm
 
-from capa.features 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")
+    ...
 
 
 def _bb_has_tight_loop(f, bb):
@@ -37,17 +40,17 @@ def _bb_has_tight_loop(f, bb):
     """
     if len(bb.instructions) > 0:
         for bva, bflags in bb.instructions[-1].getBranches():
-            if bflags & vivisect.envi.BR_COND:
+            if bflags & envi.BR_COND:
                 if bva == bb.va:
                     return True
 
     return 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.va
+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.inner):
+        yield Characteristic("tight loop"), bb.address
 
 
 def _bb_has_stackstring(f, bb):
@@ -67,13 +70,13 @@ def _bb_has_stackstring(f, bb):
     return False
 
 
-def extract_stackstring(f, bb):
-    """ check basic block for stackstring indicators """
-    if _bb_has_stackstring(f, bb):
-        yield Characteristic("stack string"), bb.va
+def extract_stackstring(f: FunctionHandle, bb: BBHandle) -> Iterator[Tuple[Feature, Address]]:
+    """check basic block for stackstring indicators"""
+    if _bb_has_stackstring(f, bb.inner):
+        yield Characteristic("stack string"), bb.address
 
 
-def is_mov_imm_to_stack(instr):
+def is_mov_imm_to_stack(instr: envi.archs.i386.disasm.i386Opcode) -> bool:
     """
     Return if instruction moves immediate onto stack
     """
@@ -105,7 +108,7 @@ def is_mov_imm_to_stack(instr):
     return True
 
 
-def get_printable_len(oper):
+def get_printable_len(oper: envi.archs.i386.disasm.i386ImmOper) -> int:
     """
     Return string length if all operand bytes are ascii or utf16-le printable
     """
@@ -117,14 +120,18 @@ def get_printable_len(oper):
         chars = struct.pack("<I", oper.imm)
     elif oper.tsize == 8:
         chars = struct.pack("<Q", oper.imm)
+    else:
+        raise ValueError("unexpected oper.tsize: %d" % (oper.tsize))
+
     if is_printable_ascii(chars):
         return oper.tsize
-    if is_printable_utf16le(chars):
+    elif is_printable_utf16le(chars):
         return oper.tsize / 2
-    return 0
+    else:
+        return 0
 
 
-def is_printable_ascii(chars):
+def is_printable_ascii(chars: bytes) -> bool:
     try:
         chars_str = chars.decode("ascii")
     except UnicodeDecodeError:
@@ -133,12 +140,13 @@ def is_printable_ascii(chars):
         return all(c in string.printable for c in chars_str)
 
 
-def is_printable_utf16le(chars):
+def is_printable_utf16le(chars: bytes) -> bool:
     if all(c == b"\x00" for c in chars[1::2]):
         return is_printable_ascii(chars[::2])
+    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.
 
@@ -147,12 +155,12 @@ def extract_features(f, bb):
       bb (viv_utils.BasicBlock): the basic block to process.
 
     yields:
-      Feature, set[VA]: the features and their location found in this basic block.
+      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

@@ -0,0 +1,92 @@
+# 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
+from typing import List, Tuple, Iterator
+
+import viv_utils
+import viv_utils.flirt
+
+import capa.features.extractors.common
+import capa.features.extractors.viv.file
+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.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 VivInstructionHandle:
+    """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__()
+        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: List[Tuple[Feature, Address]] = []
+        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 AbsoluteVirtualAddress(list(self.vw.filemeta.values())[0]["imagebase"])
+
+    def extract_global_features(self):
+        yield from self.global_features
+
+    def extract_file_features(self):
+        yield from capa.features.extractors.viv.file.extract_features(self.vw, self.buf)
+
+    def get_functions(self) -> Iterator[FunctionHandle]:
+        for va in sorted(self.vw.getFunctions()):
+            yield FunctionHandle(address=AbsoluteVirtualAddress(va), inner=viv_utils.Function(self.vw, va))
+
+    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, 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, fh: FunctionHandle, bbh) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.viv.basicblock.extract_features(fh, bbh)
+
+    def get_instructions(self, fh: FunctionHandle, bbh: BBHandle) -> Iterator[InsnHandle]:
+        bb: viv_utils.BasicBlock = bbh.inner
+        for insn in bb.instructions:
+            yield InsnHandle(address=AbsoluteVirtualAddress(insn.va), inner=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, addr):
+        return viv_utils.flirt.is_library_function(self.vw, addr)
+
+    def get_function_name(self, addr):
+        return viv_utils.get_function_name(self.vw, addr)

capa/features/extractors/viv/file.py

@@ -1,33 +1,36 @@
-# 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
+import viv_utils.flirt
 
+import capa.features.insn
+import capa.features.extractors.common
 import capa.features.extractors.helpers
 import capa.features.extractors.strings
-from capa.features import String, Characteristic
-from capa.features.file import Export, Import, Section
+from capa.features.file import Export, Import, Section, FunctionName
+from capa.features.common import String, Feature, Characteristic
+from capa.features.address import Address, FileOffsetAddress, AbsoluteVirtualAddress
 
 
-def extract_file_embedded_pe(vw, file_path):
-    with open(file_path, "rb") as f:
-        fbytes = f.read()
-
-    for offset, i in pe_carve.carve(fbytes, 1):
-        yield Characteristic("embedded pe"), offset
+def extract_file_embedded_pe(buf, **kwargs) -> Iterator[Tuple[Feature, Address]]:
+    for offset, _ in pe_carve.carve(buf, 1):
+        yield Characteristic("embedded pe"), FileOffsetAddress(offset)
 
 
-def extract_file_export_names(vw, file_path):
-    for va, etype, name, _ in vw.getExports():
-        yield Export(name), va
+def extract_file_export_names(vw, **kwargs) -> Iterator[Tuple[Feature, Address]]:
+    for va, _, name, _ in vw.getExports():
+        yield Export(name), AbsoluteVirtualAddress(va)
 
 
-def extract_file_import_names(vw, file_path):
+def extract_file_import_names(vw, **kwargs) -> Iterator[Tuple[Feature, Address]]:
     """
     extract imported function names
     1. imports by ordinal:
@@ -38,16 +41,17 @@ def extract_file_import_names(vw, file_path):
     """
     for va, _, _, tinfo in vw.getImports():
         # vivisect source: tinfo = "%s.%s" % (libname, impname)
-        modname, impname = tinfo.split(".")
+        modname, impname = tinfo.split(".", 1)
         if is_viv_ord_impname(impname):
             # 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):
+def is_viv_ord_impname(impname: str) -> bool:
     """
     return if import name matches vivisect's ordinal naming scheme `'ord%d' % ord`
     """
@@ -61,40 +65,51 @@ def is_viv_ord_impname(impname):
         return True
 
 
-def extract_file_section_names(vw, file_path):
+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(vw, file_path):
+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) -> Iterator[Tuple[Feature, Address]]:
     """
-    extract ASCII and UTF-16 LE strings from file
+    extract the names of statically-linked library functions.
     """
-    with open(file_path, "rb") as f:
-        b = f.read()
-
-    for s in capa.features.extractors.strings.extract_ascii_strings(b):
-        yield String(s.s), s.offset
-
-    for s in capa.features.extractors.strings.extract_unicode_strings(b):
-        yield String(s.s), s.offset
+    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), 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_features(vw, file_path):
+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) -> Iterator[Tuple[Feature, Address]]:
     """
     extract file features from given workspace
 
     args:
       vw (vivisect.VivWorkspace): the vivisect workspace
-      file_path: path to the input file
+      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, file_path):
-            yield feature, va
+        for feature, addr in file_handler(vw=vw, buf=buf):  # type: ignore
+            yield feature, addr
 
 
 FILE_HANDLERS = (
@@ -103,4 +118,6 @@ FILE_HANDLERS = (
     extract_file_import_names,
     extract_file_section_names,
     extract_file_strings,
+    extract_file_function_names,
+    extract_file_format,
 )

capa/features/extractors/viv/function.py

@@ -1,39 +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 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")
+    ...
 
 
-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:
@@ -41,30 +49,30 @@ def extract_function_loop(f):
             for bva, bflags in bb.instructions[-1].getBranches():
                 # vivisect does not set branch flags for non-conditional jmp so add explicit check
                 if (
-                    bflags & vivisect.envi.BR_COND
-                    or bflags & vivisect.envi.BR_FALL
-                    or bflags & vivisect.envi.BR_TABLE
+                    bflags & envi.BR_COND
+                    or bflags & envi.BR_FALL
+                    or bflags & envi.BR_TABLE
                     or bb.instructions[-1].mnem == "jmp"
                 ):
                     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:
-      Feature, set[VA]: the features and their location found in this function.
+      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

@@ -0,0 +1,26 @@
+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, Feature
+from capa.features.address import NO_ADDRESS, Address
+
+logger = logging.getLogger(__name__)
+
+
+def extract_arch(vw) -> Iterator[Tuple[Feature, Address]]:
+    if isinstance(vw.arch, envi.archs.amd64.Amd64Module):
+        yield Arch(ARCH_AMD64), NO_ADDRESS
+
+    elif isinstance(vw.arch, envi.archs.i386.i386Module):
+        yield Arch(ARCH_I386), NO_ADDRESS
+
+    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", vw.arch.__class__.__name__)
+        return

capa/features/extractors/viv/helpers.py

@@ -1,14 +1,17 @@
-# 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 Optional
+
+from vivisect import VivWorkspace
 from vivisect.const import XR_TO, REF_CODE
 
 
-def get_coderef_from(vw, va):
+def get_coderef_from(vw: VivWorkspace, va: int) -> Optional[int]:
     """
     return first code `tova` whose origin is the specified va
     return None if no code reference is found

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
@@ -7,11 +7,16 @@
 # 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
 
 import envi
 import vivisect.const
 import envi.archs.i386.disasm
 import envi.archs.amd64.disasm
+from vivisect import VivWorkspace
+
+if TYPE_CHECKING:
+    from capa.features.extractors.viv.extractor import VivInstructionHandle
 
 # pull out consts for lookup performance
 i386RegOper = envi.archs.i386.disasm.i386RegOper
@@ -26,7 +31,7 @@ FAR_BRANCH_MASK = envi.BR_PROC | envi.BR_DEREF | envi.BR_ARCH
 DESTRUCTIVE_MNEMONICS = ("mov", "lea", "pop", "xor")
 
 
-def get_previous_instructions(vw, va):
+def get_previous_instructions(vw: VivWorkspace, va: int) -> List[int]:
     """
     collect the instructions that flow to the given address, local to the current function.
 
@@ -43,12 +48,14 @@ def get_previous_instructions(vw, va):
     # ensure that it fallsthrough to this one.
     loc = vw.getPrevLocation(va, adjacent=True)
     if loc is not None:
-        # from vivisect.const:
-        # location: (L_VA, L_SIZE, L_LTYPE, L_TINFO)
-        (pva, _, ptype, pinfo) = vw.getPrevLocation(va, adjacent=True)
+        ploc = vw.getPrevLocation(va, adjacent=True)
+        if ploc is not None:
+            # from vivisect.const:
+            # location: (L_VA, L_SIZE, L_LTYPE, L_TINFO)
+            (pva, _, ptype, pinfo) = ploc
 
-        if ptype == LOC_OP and not (pinfo & IF_NOFALL):
-            ret.append(pva)
+            if ptype == LOC_OP and not (pinfo & IF_NOFALL):
+                ret.append(pva)
 
     # find any code refs, e.g. jmp, to this location.
     # ignore any calls.
@@ -67,7 +74,7 @@ class NotFoundError(Exception):
     pass
 
 
-def find_definition(vw, va, reg):
+def find_definition(vw: VivWorkspace, va: int, reg: int) -> Tuple[int, Union[int, None]]:
     """
     scan backwards from the given address looking for assignments to the given register.
     if a constant, return that value.
@@ -83,8 +90,8 @@ def find_definition(vw, va, reg):
     raises:
       NotFoundError: when the definition cannot be found.
     """
-    q = collections.deque()
-    seen = set([])
+    q = collections.deque()  # type: Deque[int]
+    seen = set([])  # type: Set[int]
 
     q.extend(get_previous_instructions(vw, va))
     while q:
@@ -128,14 +135,16 @@ def find_definition(vw, va, reg):
     raise NotFoundError()
 
 
-def is_indirect_call(vw, va, insn=None):
+def is_indirect_call(vw: VivWorkspace, va: int, insn: Optional["VivInstructionHandle"] = None) -> 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, va, insn=None):
+def resolve_indirect_call(
+    vw: VivWorkspace, va: int, insn: Optional["VivInstructionHandle"] = None
+) -> Tuple[int, Optional[int]]:
     """
     inspect the given indirect call instruction and attempt to resolve the target address.
 

capa/features/extractors/viv/insn.py

@@ -1,26 +1,28 @@
-# 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
 import envi.memory
+import viv_utils.flirt
+import envi.archs.i386.regs
+import envi.archs.amd64.regs
 import envi.archs.i386.disasm
+import envi.archs.amd64.disasm
 
 import capa.features.extractors.helpers
 import capa.features.extractors.viv.helpers
-from capa.features import (
-    ARCH_X32,
-    ARCH_X64,
-    MAX_BYTES_FEATURE_SIZE,
-    THUNK_CHAIN_DEPTH_DELTA,
-    Bytes,
-    String,
-    Characteristic,
-)
-from capa.features.insn import API, Number, Offset, Mnemonic
+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
@@ -28,27 +30,21 @@ from capa.features.extractors.viv.indirect_calls import NotFoundError, resolve_i
 SECURITY_COOKIE_BYTES_DELTA = 0x40
 
 
-def get_arch(vw):
-    arch = vw.getMeta("Architecture")
-    if arch == "i386":
-        return ARCH_X32
-    elif arch == "amd64":
-        return ARCH_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")
+    ...
 
 
 def get_imports(vw):
@@ -68,13 +64,15 @@ def get_imports(vw):
         return imports
 
 
-def extract_insn_api_features(f, bb, insn):
-    """parse API features from the given instruction."""
-
-    # example:
-    #
-    #    call dword [0x00473038]
+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]
+    """
+    insn: envi.Opcode = ih.inner
+    f: viv_utils.Function = fh.inner
     if insn.mnem not in ("call", "jmp"):
         return
 
@@ -91,12 +89,12 @@ 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
     #
-    # this is also how calls to internal functions may be decoded on x64.
+    # this is also how calls to internal functions may be decoded on x32 and x64.
     # see Lab21-01.exe_:0x140001178
     #
     # follow chained thunks, e.g. in 82bf6347acf15e5d883715dc289d8a2b at 0x14005E0FF in
@@ -111,11 +109,26 @@ def extract_insn_api_features(f, bb, insn):
         if not target:
             return
 
+        if viv_utils.flirt.is_library_function(f.vw, target):
+            name = viv_utils.get_function_name(f.vw, target)
+            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"):
+                target += 4
 
             target = capa.features.extractors.viv.helpers.get_coderef_from(f.vw, target)
             if not target:
@@ -131,7 +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:
@@ -148,38 +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.disasm.REG_ESP:
-            # skip things like:
-            #
-            #    .text:00401140                 call    sub_407E2B
-            #    .text:00401145                 add     esp, 0Ch
-            return
-
-        yield Number(v), insn.va
-        yield Number(v, arch=get_arch(f.vw)), insn.va
+                yield API(name), ih.address
 
 
 def derefs(vw, p):
@@ -214,7 +196,7 @@ def derefs(vw, p):
         p = next
 
 
-def read_memory(vw, va, size):
+def read_memory(vw, va: int, size: int) -> bytes:
     # as documented in #176, vivisect will not readMemory() when the section is not marked readable.
     #
     # but here, we don't care about permissions.
@@ -227,10 +209,10 @@ def read_memory(vw, va, size):
             mva, msize, mperms, mfname = mmap
             offset = va - mva
             return mbytes[offset : offset + size]
-    raise envi.SegmentationViolation(va)
+    raise envi.exc.SegmentationViolation(va)
 
 
-def read_bytes(vw, va):
+def read_bytes(vw, va: int) -> bytes:
     """
     read up to MAX_BYTES_FEATURE_SIZE from the given address.
 
@@ -239,7 +221,7 @@ def read_bytes(vw, va):
     """
     segm = vw.getSegment(va)
     if not segm:
-        raise envi.SegmentationViolation(va)
+        raise envi.exc.SegmentationViolation(va)
 
     segm_end = segm[0] + segm[1]
     try:
@@ -248,16 +230,19 @@ def read_bytes(vw, va):
             return read_memory(vw, va, segm_end - va)
         else:
             return read_memory(vw, va, MAX_BYTES_FEATURE_SIZE)
-    except envi.SegmentationViolation:
+    except envi.exc.SegmentationViolation:
         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
 
@@ -280,19 +265,19 @@ def extract_insn_bytes_features(f, bb, insn):
         for v in derefs(f.vw, v):
             try:
                 buf = read_bytes(f.vw, v)
-            except envi.SegmentationViolation:
+            except envi.exc.SegmentationViolation:
                 continue
 
             if capa.features.extractors.helpers.all_zeros(buf):
                 continue
 
-            yield Bytes(buf), insn.va
+            yield Bytes(buf), ih.address
 
 
-def read_string(vw, offset):
+def read_string(vw, offset: int) -> str:
     try:
         alen = vw.detectString(offset)
-    except envi.SegmentationViolation:
+    except envi.exc.SegmentationViolation:
         pass
     else:
         if alen > 0:
@@ -300,7 +285,7 @@ def read_string(vw, offset):
 
     try:
         ulen = vw.detectUnicode(offset)
-    except envi.SegmentationViolation:
+    except envi.exc.SegmentationViolation:
         pass
     except IndexError:
         # potential vivisect bug detecting Unicode at segment end
@@ -320,87 +305,18 @@ def read_string(vw, offset):
     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.disasm.REG_ESP:
-                continue
-
-            if oper.reg == envi.archs.i386.disasm.REG_EBP:
-                continue
-
-            # TODO: do x64 support for real.
-            if oper.reg == envi.archs.amd64.disasm.REG_RBP:
-                continue
-
-            # viv already decodes offsets as signed
-            v = oper.disp
-
-            yield Offset(v), insn.va
-            yield Offset(v, arch=get_arch(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, arch=get_arch(f.vw)), insn.va
-
-
-def is_security_cookie(f, bb, insn):
+def is_security_cookie(f, bb, insn) -> bool:
     """
     check if an instruction is related to security cookie checks
     """
     # security cookie check should use SP or BP
     oper = insn.opers[1]
     if oper.isReg() and oper.reg not in [
-        envi.archs.i386.disasm.REG_ESP,
-        envi.archs.i386.disasm.REG_EBP,
+        envi.archs.i386.regs.REG_ESP,
+        envi.archs.i386.regs.REG_EBP,
         # TODO: do x64 support for real.
-        envi.archs.amd64.disasm.REG_RBP,
-        envi.archs.amd64.disasm.REG_RSP,
+        envi.archs.amd64.regs.REG_RBP,
+        envi.archs.amd64.regs.REG_RSP,
     ]:
         return False
 
@@ -417,11 +333,17 @@ def is_security_cookie(f, bb, insn):
     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
 
@@ -431,19 +353,40 @@ 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
@@ -462,7 +405,7 @@ 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 (
@@ -470,23 +413,25 @@ 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):
-    """ parse the instruction for access to fs or gs """
+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):
+def get_section(vw, va: int):
     for start, length, _, __ in vw.getMemoryMaps():
         if start <= va < start + length:
             return start
@@ -494,10 +439,13 @@ def get_section(vw, va):
     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`.
@@ -524,7 +472,7 @@ 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
@@ -532,7 +480,10 @@ 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
 
@@ -542,7 +493,7 @@ 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
+        yield Characteristic("calls from"), AbsoluteVirtualAddress(target)
 
     # call via thunk on x86,
     # see 9324d1a8ae37a36ae560c37448c9705a at 0x407985
@@ -551,43 +502,190 @@ 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
+        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
+        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)
+        except ValueError:
+            continue
+        else:
+            yield String(s.rstrip("\x00")), 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.
 
@@ -597,24 +695,23 @@ def extract_features(f, bb, insn):
       insn (vivisect...Instruction): the instruction to process.
 
     yields:
-      Feature, set[VA]: 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
@@ -6,22 +6,33 @@
 #  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 capa.features import Feature
+from capa.features.common import Feature
 
 
 class Export(Feature):
-    def __init__(self, value, description=None):
+    def __init__(self, value: str, description=None):
         # value is export name
         super(Export, self).__init__(value, description=description)
 
 
 class Import(Feature):
-    def __init__(self, value, description=None):
+    def __init__(self, value: str, description=None):
         # value is import name
         super(Import, self).__init__(value, description=description)
 
 
 class Section(Feature):
-    def __init__(self, value, description=None):
+    def __init__(self, value: str, description=None):
         # value is section name
         super(Section, self).__init__(value, description=description)
+
+
+class FunctionName(Feature):
+    """recognized name for statically linked function"""
+
+    def __init__(self, name: str, description=None):
+        # value is function name
+        super(FunctionName, self).__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

@@ -1,47 +1,7 @@
 """
 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': {
-        '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.
+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
@@ -49,190 +9,351 @@ 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 json
 import zlib
 import logging
+import collections
+from enum import Enum
+from typing import Any, Set, Dict, List, Type, Tuple
 
-import capa.features
+from pydantic import Field, BaseModel
+
+import capa.helpers
 import capa.features.file
 import capa.features.insn
+import capa.features.common
+import capa.features.address
 import capa.features.basicblock
-import capa.features.extractors
-from capa.helpers import hex
+import capa.features.extractors.base_extractor
+from capa.helpers import assert_never
 
 logger = logging.getLogger(__name__)
 
 
-def serialize_feature(feature):
-    return feature.freeze_serialize()
+class HashableModel(BaseModel):
+    class Config:
+        frozen = True
 
 
-KNOWN_FEATURES = {F.__name__: F for F in capa.features.Feature.__subclasses__()}
+class AddressType(str, Enum):
+    ABSOLUTE = "absolute"
+    RELATIVE = "relative"
+    FILE = "file"
+    DN_TOKEN = "dn token"
+    DN_TOKEN_OFFSET = "dn token offset"
+    NO_ADDRESS = "no address"
 
 
-def deserialize_feature(doc):
-    F = KNOWN_FEATURES[doc[0]]
-    return F.freeze_deserialize(doc[1])
+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):
+            # TODO: probably need serialization here
+            return cls(type=AddressType.DN_TOKEN, value=a.token)
+
+        elif isinstance(a, capa.features.address.DNTokenOffsetAddress):
+            # TODO: probably need serialization here
+            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:
+            return capa.features.address.DNTokenOffsetAddress(*self.value)
+
+        elif self.type is AddressType.NO_ADDRESS:
+            return capa.features.address.NO_ADDRESS
+
+        else:
+            assert_never(self.type)
 
 
-def dumps(extractor):
+KNOWN_FEATURES: Dict[str, Type[capa.features.common.Feature]] = {
+    F.__name__: F for F in capa.features.common.Feature.__subclasses__()
+}
+KNOWN_FEATURES.update({F.__name__: F for F in capa.features.insn._Operand.__subclasses__()})  # type: ignore
+
+
+class Feature(HashableModel):
+    name: str
+    args: Tuple[Any, ...]
+
+    @classmethod
+    def from_capa(cls, f: capa.features.common.Feature) -> "Feature":
+        name, args = f.freeze_serialize()
+        return cls(name=name, args=tuple(args))
+
+    def to_capa(self) -> capa.features.common.Feature:
+        F = KNOWN_FEATURES[self.name]
+        return F.freeze_deserialize(self.args)
+
+
+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
+    address: Address
+    feature: Feature
+
+
+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: List[InstructionFeature]
+
+
+class BasicBlockFeatures(BaseModel):
+    address: Address
+    features: List[BasicBlockFeature]
+    instructions: List[InstructionFeatures]
+
+
+class FunctionFeatures(BaseModel):
+    address: Address
+    features: List[FunctionFeature]
+    basic_blocks: List[BasicBlockFeatures] = Field(alias="basic block")
+
+    class Config:
+        allow_population_by_field_name = True
+
+
+class Features(BaseModel):
+    global_: List[GlobalFeature] = Field(alias="global")
+    file: List[FileFeature]
+    functions: List[FunctionFeatures]
+
+    class Config:
+        allow_population_by_field_name = True
+
+
+class Freeze(BaseModel):
+    version: int = 2
+    base_address: Address = Field(alias="base address")
+    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
-
-    args:
-      extractor: capa.features.extractor.FeatureExtractor:
-
-    returns:
-      str: the serialized features.
     """
-    ret = {
-        "version": 1,
-        "base address": extractor.get_base_address(),
-        "functions": {},
-        "scopes": {
-            "file": [],
-            "function": [],
-            "basic block": [],
-            "instruction": [],
-        },
-    }
 
-    for feature, va in extractor.extract_file_features():
-        ret["scopes"]["file"].append(serialize_feature(feature) + (hex(va), ()))
+    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():
-        ret["functions"][hex(f)] = {}
-
-        for feature, va in extractor.extract_function_features(f):
-            ret["scopes"]["function"].append(serialize_feature(feature) + (hex(va), (hex(f),)))
+        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):
-            ret["functions"][hex(f)][hex(bb)] = []
+            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)
+            ]
 
-            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),
-                        ),
+            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,
                     )
                 )
 
-            for insnva, insn in sorted(
-                [(insn.__int__(), insn) for insn in extractor.get_instructions(f, bb)], key=lambda p: p[0]
-            ):
-                ret["functions"][hex(f)][hex(bb)].append(hex(insnva))
+            basic_blocks.append(
+                BasicBlockFeatures(
+                    address=bbaddr,
+                    features=bbfeatures,
+                    instructions=instructions,
+                )
+            )
 
-                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)
+        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()),
+        features=features,
+    )
+
+    return freeze.json()
 
 
-def loads(s):
+def loads(s: str) -> capa.features.extractors.base_extractor.FeatureExtractor:
     """deserialize a set of features (as a NullFeatureExtractor) from a string."""
-    doc = json.loads(s)
+    import capa.features.extractors.null as null
 
-    if doc.get("version") != 1:
-        raise ValueError("unsupported freeze format version: %d" % (doc.get("version")))
+    freeze = Freeze.parse_raw(s)
+    if freeze.version != 2:
+        raise ValueError("unsupported freeze format version: %d", freeze.version)
 
-    features = {
-        "base address": doc.get("base address"),
-        "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("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.NullFeatureExtractor(features)
+    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):
+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):
+def is_freeze(buf: bytes) -> bool:
     return buf[: len(MAGIC)] == MAGIC
 
 
-def load(buf):
+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")
@@ -248,45 +369,16 @@ def main(argv=None):
     if argv is None:
         argv = sys.argv[1:]
 
-    formats = [
-        ("auto", "(default) detect file type automatically"),
-        ("pe", "Windows PE file"),
-        ("sc32", "32-bit shellcode"),
-        ("sc64", "64-bit shellcode"),
-    ]
-    format_help = ", ".join(["%s: %s" % (f[0], f[1]) for f in formats])
-
     parser = argparse.ArgumentParser(description="save capa features to a file")
-    parser.add_argument("sample", type=str, help="Path to sample to analyze")
+    capa.main.install_common_args(parser, {"sample", "format", "backend", "signatures"})
     parser.add_argument("output", type=str, help="Path to output file")
-    parser.add_argument("-v", "--verbose", action="store_true", help="Enable verbose output")
-    parser.add_argument("-q", "--quiet", action="store_true", help="Disable all output but errors")
-    parser.add_argument(
-        "-f", "--format", choices=[f[0] for f in formats], default="auto", help="Select sample format, %s" % format_help
-    )
-    if sys.version_info >= (3, 0):
-        parser.add_argument(
-            "-b",
-            "--backend",
-            type=str,
-            help="select the backend to use",
-            choices=(capa.main.BACKEND_VIV, capa.main.BACKEND_SMDA),
-            default=capa.main.BACKEND_VIV,
-        )
     args = parser.parse_args(args=argv)
+    capa.main.handle_common_args(args)
 
-    if args.quiet:
-        logging.basicConfig(level=logging.ERROR)
-        logging.getLogger().setLevel(logging.ERROR)
-    elif args.verbose:
-        logging.basicConfig(level=logging.DEBUG)
-        logging.getLogger().setLevel(logging.DEBUG)
-    else:
-        logging.basicConfig(level=logging.INFO)
-        logging.getLogger().setLevel(logging.INFO)
+    sigpaths = capa.main.get_signatures(args.signatures)
+
+    extractor = capa.main.get_extractor(args.sample, args.format, args.backend, sigpaths, False)
 
-    backend = args.backend if sys.version_info > (3, 0) else capa.main.BACKEND_VIV
-    extractor = capa.main.get_extractor(args.sample, args.format, backend)
     with open(args.output, "wb") as f:
         f.write(dump(extractor))
 

capa/features/insn.py

@@ -1,40 +1,99 @@
-# 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
 
-from capa.features import Feature
+import capa.render.utils
+from capa.features.common import Feature
 
 
 class API(Feature):
-    def __init__(self, name, description=None):
-        # Downcase library name if given
-        if "." in name:
-            modname, _, impname = name.rpartition(".")
-            name = modname.lower() + "." + impname
-
+    def __init__(self, name: str, description=None):
         super(API, self).__init__(name, description=description)
 
 
 class Number(Feature):
-    def __init__(self, value, arch=None, description=None):
-        super(Number, self).__init__(value, arch=arch, description=description)
+    def __init__(self, value: Union[int, float], description=None):
+        super(Number, self).__init__(value, description=description)
 
     def get_value_str(self):
-        return "0x%X" % self.value
+        if isinstance(self.value, int):
+            return capa.render.utils.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, arch=None, description=None):
-        super(Offset, self).__init__(value, arch=arch, description=description)
+    def __init__(self, value: int, description=None):
+        super(Offset, self).__init__(value, description=description)
 
     def get_value_str(self):
-        return "0x%X" % self.value
+        return capa.render.utils.hex(self.value)
 
 
 class Mnemonic(Feature):
-    def __init__(self, value, description=None):
+    def __init__(self, value: str, description=None):
         super(Mnemonic, self).__init__(value, description=description)
+
+
+# max number of operands to consider for a given instrucion.
+# 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(_Operand, self).__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
+
+    def freeze_serialize(self):
+        return (self.__class__.__name__, (self.index, self.value))
+
+
+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(OperandNumber, self).__init__(index, value, description=description)
+        self.name = self.NAMES[index]
+
+    def get_value_str(self) -> str:
+        assert isinstance(self.value, int)
+        return capa.render.utils.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(OperandOffset, self).__init__(index, value, description=description)
+        self.name = self.NAMES[index]
+
+    def get_value_str(self) -> str:
+        assert isinstance(self.value, int)
+        return capa.render.utils.hex(self.value)

capa/helpers.py

@@ -1,36 +1,118 @@
-# 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
+
+from capa.exceptions import UnsupportedFormatError
+from capa.features.common import FORMAT_SC32, FORMAT_SC64, FORMAT_UNKNOWN
+
+EXTENSIONS_SHELLCODE_32 = ("sc32", "raw32")
+EXTENSIONS_SHELLCODE_64 = ("sc64", "raw64")
+
+
+logger = logging.getLogger("capa")
 
 _hex = hex
 
 
 def hex(i):
-    # under py2.7, long integers get formatted with a trailing `L`
-    # and this is not pretty. so strip it out.
-    return _hex(oint(i)).rstrip("L")
+    return _hex(int(i))
 
 
-def oint(i):
-    # there seems to be some trouble with using `int(viv_utils.Function)`
-    # with the black magic we do with binding the `__int__()` routine.
-    # i haven't had a chance to debug this yet (and i have no hotel wifi).
-    # so in the meantime, detect this, and call the method directly.
-    try:
-        return int(i)
-    except TypeError:
-        return i.__int__()
-
-
-def get_file_taste(sample_path):
+def get_file_taste(sample_path: str) -> bytes:
     if not os.path.exists(sample_path):
         raise IOError("sample path %s does not exist or cannot be accessed" % sample_path)
     with open(sample_path, "rb") as f:
         taste = f.read(8)
     return taste
+
+
+def is_runtime_ida():
+    try:
+        import idc
+    except ImportError:
+        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
+
+    with open(sample, "rb") as f:
+        buf = f.read()
+
+    for feature, _ in extract_format(buf):
+        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

@@ -0,0 +1,163 @@
+# 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 datetime
+
+import idc
+import idaapi
+import idautils
+import ida_bytes
+import ida_loader
+
+import capa
+import capa.version
+import capa.features.common
+
+logger = logging.getLogger("capa")
+
+# file type as returned by idainfo.file_type
+SUPPORTED_FILE_TYPES = (
+    idaapi.f_PE,
+    idaapi.f_ELF,
+    idaapi.f_BIN,
+    # idaapi.f_MACHO,
+)
+
+# arch type as returned by idainfo.procname
+SUPPORTED_ARCH_TYPES = ("metapc",)
+
+
+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 = float(idaapi.get_kernel_version())
+    if version < 7.4 or version >= 8:
+        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: IDA >= 7.4 and IDA < 8.0." % version)
+        return False
+    return True
+
+
+def is_supported_file_type():
+    file_info = idaapi.get_inf_structure()
+    if file_info.filetype not in SUPPORTED_FILE_TYPES:
+        logger.error("-" * 80)
+        logger.error(" Input file does not appear to be a supported file type.")
+        logger.error(" ")
+        logger.error(
+            " capa currently only supports analyzing PE, ELF, or binary files containing x86 (32- and 64-bit) shellcode."
+        )
+        logger.error(" If you don't know the input file type, you can try using the `file` utility to guess it.")
+        logger.error("-" * 80)
+        return False
+    return True
+
+
+def is_supported_arch_type():
+    file_info = idaapi.get_inf_structure()
+    if file_info.procname not in SUPPORTED_ARCH_TYPES or not any((file_info.is_32bit(), file_info.is_64bit())):
+        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)
+        return False
+    return True
+
+
+def get_disasm_line(va):
+    """ """
+    return idc.generate_disasm_line(va, idc.GENDSM_FORCE_CODE)
+
+
+def is_func_start(ea):
+    """check if function stat exists at virtual address"""
+    f = idaapi.get_func(ea)
+    return f and f.start_ea == ea
+
+
+def get_func_start_ea(ea):
+    """ """
+    f = idaapi.get_func(ea)
+    return f if f is None else f.start_ea
+
+
+def get_file_md5():
+    """ """
+    md5 = idautils.GetInputFileMD5()
+    if not isinstance(md5, str):
+        md5 = capa.features.common.bytes_to_str(md5)
+    return md5
+
+
+def get_file_sha256():
+    """ """
+    sha256 = idaapi.retrieve_input_file_sha256()
+    if not isinstance(sha256, str):
+        sha256 = capa.features.common.bytes_to_str(sha256)
+    return sha256
+
+
+def collect_metadata():
+    """ """
+    md5 = get_file_md5()
+    sha256 = get_file_sha256()
+
+    return {
+        "timestamp": datetime.datetime.now().isoformat(),
+        # "argv" is not relevant here
+        "sample": {
+            "md5": md5,
+            "sha1": "",  # not easily accessible
+            "sha256": sha256,
+            "path": idaapi.get_input_file_path(),
+        },
+        "analysis": {
+            "format": idaapi.get_file_type_name(),
+            "extractor": "ida",
+            "base_address": idaapi.get_imagebase(),
+            "layout": {
+                # this is updated after capabilities have been collected.
+                # will look like:
+                #
+                # "functions": { 0x401000: { "matched_basic_blocks": [ 0x401000, 0x401005, ... ] }, ... }
+            },
+        },
+        "version": capa.version.__version__,
+    }
+
+
+class IDAIO:
+    """
+    An object that acts as a file-like object,
+    using bytes from the current IDB workspace.
+    """
+
+    def __init__(self):
+        super(IDAIO, self).__init__()
+        self.offset = 0
+
+    def seek(self, offset, whence=0):
+        assert whence == 0
+        self.offset = offset
+
+    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("reading 0x%x bytes at 0x%x (ea: 0x%x)", size, self.offset, ea)
+        return ida_bytes.get_bytes(ea, size)
+
+    def close(self):
+        return

capa/ida/helpers/__init__.py

@@ -1,121 +0,0 @@
-# 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 logging
-import datetime
-
-import idc
-import six
-import idaapi
-import idautils
-
-import capa
-
-logger = logging.getLogger("capa")
-
-SUPPORTED_IDA_VERSIONS = [
-    "7.1",
-    "7.2",
-    "7.3",
-    "7.4",
-    "7.5",
-]
-
-# file type names as returned by idaapi.get_file_type_name()
-SUPPORTED_FILE_TYPES = [
-    "Portable executable for 80386 (PE)",
-    "Portable executable for AMD64 (PE)",
-    "Binary file",  # x86/AMD64 shellcode support
-]
-
-
-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:
-        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))
-        )
-        return False
-    return True
-
-
-def is_supported_file_type():
-    file_type = idaapi.get_file_type_name()
-    if file_type not in SUPPORTED_FILE_TYPES:
-        logger.error("-" * 80)
-        logger.error(" Input file does not appear to be a PE file.")
-        logger.error(" ")
-        logger.error(
-            " capa currently only supports analyzing PE files (or binary files containing x86/AMD64 shellcode) with IDA."
-        )
-        logger.error(" If you don't know the input file type, you can try using the `file` utility to guess it.")
-        logger.error("-" * 80)
-        return False
-    return True
-
-
-def get_disasm_line(va):
-    """ """
-    return idc.generate_disasm_line(va, idc.GENDSM_FORCE_CODE)
-
-
-def is_func_start(ea):
-    """ check if function stat exists at virtual address """
-    f = idaapi.get_func(ea)
-    return f and f.start_ea == ea
-
-
-def get_func_start_ea(ea):
-    """ """
-    f = idaapi.get_func(ea)
-    return f if f is None else f.start_ea
-
-
-def get_file_md5():
-    """ """
-    md5 = idautils.GetInputFileMD5()
-    if not isinstance(md5, six.string_types):
-        md5 = capa.features.bytes_to_str(md5)
-    return md5
-
-
-def get_file_sha256():
-    """ """
-    sha256 = idaapi.retrieve_input_file_sha256()
-    if not isinstance(sha256, six.string_types):
-        sha256 = capa.features.bytes_to_str(sha256)
-    return sha256
-
-
-def collect_metadata():
-    """ """
-    md5 = get_file_md5()
-    sha256 = get_file_sha256()
-
-    return {
-        "timestamp": datetime.datetime.now().isoformat(),
-        # "argv" is not relevant here
-        "sample": {
-            "md5": md5,
-            "sha1": "",  # not easily accessible
-            "sha256": sha256,
-            "path": idaapi.get_input_file_path(),
-        },
-        "analysis": {
-            "format": idaapi.get_file_type_name(),
-            "extractor": "ida",
-            "base_address": idaapi.get_imagebase(),
-        },
-        "version": capa.version.__version__,
-    }

capa/ida/plugin/README.md

@@ -34,18 +34,49 @@ For more information on the FLARE team's open-source framework, capa, check out
 
 ### Requirements
 
-capa explorer supports the following IDA setups:
+capa explorer supports Python versions >= 3.7.x and the following IDA Pro versions:
 
-* IDA Pro 7.4+ with Python 2.7 or Python 3.
+* IDA 7.4
+* IDA 7.5
+* IDA 7.6 (caveat below)
+* IDA 7.7
+
+capa explorer is however limited to the Python versions supported by your IDA installation (which may not include all Python versions >= 3.7.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 3.7.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/mandiant/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)
 
-If you encounter issues with your specific setup, please open a new [Issue](https://github.com/fireeye/capa/issues).
 
 ### Supported File Types
 
 capa explorer is limited to the file types supported by capa, which include:
 
-* Windows 32-bit and 64-bit PE files
-* Windows 32-bit and 64-bit shellcode
+* Windows x86 (32- and 64-bit) PE and ELF files
+* Windows x86 (32- and 64-bit) shellcode
 
 ### Installation
 
@@ -55,19 +86,20 @@ You can install capa explorer using the following steps:
     ```
     $ 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
+3. Download the [standard collection of capa rules](https://github.com/mandiant/capa-rules) (capa explorer needs capa rules to analyze a database)
+4. Copy [capa_explorer.py](https://raw.githubusercontent.com/mandiant/capa/master/capa/ida/plugin/capa_explorer.py) to your IDA plugins directory
 
 ### Usage
 
 1. Open IDA and analyze a supported file type (select the `Manual Load` and `Load Resources` options in IDA for best results)
 2. Open capa explorer in IDA by navigating to `Edit > Plugins > FLARE capa explorer` or using the keyboard shortcut `Alt+F5`
+   You can also use `ida_loader.load_and_run_plugin("capa_explorer", arg)`. `arg` is a bitflag for which setting the LSB enables automatic analysis. See `capa.ida.plugin.form.Options` for more details.
 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.
+downloading and using the [standard collection of capa rules](https://github.com/mandiant/capa-rules) when getting started with the plugin.
 
 #### Tips for Program Analysis
 
@@ -93,15 +125,15 @@ downloading and using the [standard collection of capa rules](https://github.com
 ## 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

@@ -1,17 +1,15 @@
-# 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
 import ida_kernwin
 
-from capa.ida.helpers import is_supported_file_type, is_supported_ida_version
 from capa.ida.plugin.form import CapaExplorerForm
 from capa.ida.plugin.icon import ICON
 
@@ -28,8 +26,8 @@ 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
 
@@ -41,10 +39,14 @@ class CapaExplorerPlugin(idaapi.plugin_t):
         """called when IDA is loading the plugin"""
         logging.basicConfig(level=logging.INFO)
 
+        import capa.ida.helpers
+
         # do not load plugin if IDA version/file type not supported
-        if not is_supported_ida_version():
+        if not capa.ida.helpers.is_supported_ida_version():
             return idaapi.PLUGIN_SKIP
-        if not is_supported_file_type():
+        if not capa.ida.helpers.is_supported_file_type():
+            return idaapi.PLUGIN_SKIP
+        if not capa.ida.helpers.is_supported_arch_type():
             return idaapi.PLUGIN_SKIP
         return idaapi.PLUGIN_OK
 
@@ -53,8 +55,14 @@ class CapaExplorerPlugin(idaapi.plugin_t):
         pass
 
     def run(self, arg):
-        """called when IDA is running the plugin as a script"""
-        self.form = CapaExplorerForm(self.PLUGIN_NAME)
+        """
+        called when IDA is running the plugin as a script
+
+        args:
+          arg (int): bitflag. Setting LSB enables automatic analysis upon
+          loading. The other bits are currently undefined. See `form.Options`.
+        """
+        self.form = CapaExplorerForm(self.PLUGIN_NAME, arg)
         return True
 
 

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/form.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
@@ -12,6 +12,7 @@ import json
 import logging
 import itertools
 import collections
+from typing import Set, Dict
 
 import idaapi
 import ida_kernwin
@@ -20,9 +21,14 @@ from PyQt5 import QtGui, QtCore, QtWidgets
 
 import capa.main
 import capa.rules
+import capa.engine
 import capa.ida.helpers
-import capa.render.utils as rutils
-import capa.features.extractors.ida
+import capa.render.json
+import capa.features.common
+import capa.render.result_document
+import capa.features.extractors.ida.extractor
+from capa.engine import FeatureSet
+from capa.features.common import Feature
 from capa.ida.plugin.icon import QICON
 from capa.ida.plugin.view import (
     CapaExplorerQtreeView,
@@ -30,9 +36,11 @@ from capa.ida.plugin.view import (
     CapaExplorerRulgenPreview,
     CapaExplorerRulegenFeatures,
 )
+from capa.features.address import NO_ADDRESS, Address
 from capa.ida.plugin.hooks import CapaExplorerIdaHooks
 from capa.ida.plugin.model import CapaExplorerDataModel
 from capa.ida.plugin.proxy import CapaExplorerRangeProxyModel, CapaExplorerSearchProxyModel
+from capa.features.extractors.base_extractor import FunctionHandle
 
 logger = logging.getLogger(__name__)
 settings = ida_settings.IDASettings("capa")
@@ -41,11 +49,16 @@ CAPA_SETTINGS_RULE_PATH = "rule_path"
 CAPA_SETTINGS_RULEGEN_AUTHOR = "rulegen_author"
 CAPA_SETTINGS_RULEGEN_SCOPE = "rulegen_scope"
 
+from enum import IntFlag
+
+
+class Options(IntFlag):
+    DEFAULT = 0
+    ANALYZE = 1  # Runs the analysis when starting the explorer
+
 
 def write_file(path, data):
     """ """
-    if os.path.exists(path) and 1 != idaapi.ask_yn(1, "The file already exists. Overwrite?"):
-        return
     with open(path, "wb") as save_file:
         save_file.write(data)
 
@@ -58,32 +71,32 @@ def trim_function_name(f, max_length=25):
     return n
 
 
-def find_func_features(f, extractor):
+def find_func_features(fh: FunctionHandle, extractor):
     """ """
-    func_features = collections.defaultdict(set)
-    bb_features = collections.defaultdict(dict)
+    func_features: Dict[Feature, Set] = collections.defaultdict(set)
+    bb_features: Dict[Address, Dict] = collections.defaultdict(dict)
 
-    for (feature, ea) in extractor.extract_function_features(f):
-        func_features[feature].add(ea)
+    for (feature, addr) in extractor.extract_function_features(fh):
+        func_features[feature].add(addr)
 
-    for bb in extractor.get_basic_blocks(f):
+    for bbh in extractor.get_basic_blocks(fh):
         _bb_features = collections.defaultdict(set)
 
-        for (feature, ea) in extractor.extract_basic_block_features(f, bb):
-            _bb_features[feature].add(ea)
-            func_features[feature].add(ea)
+        for (feature, addr) in extractor.extract_basic_block_features(fh, bbh):
+            _bb_features[feature].add(addr)
+            func_features[feature].add(addr)
 
-        for insn in extractor.get_instructions(f, bb):
-            for (feature, ea) in extractor.extract_insn_features(f, bb, insn):
-                _bb_features[feature].add(ea)
-                func_features[feature].add(ea)
+        for insn in extractor.get_instructions(fh, bbh):
+            for (feature, addr) in extractor.extract_insn_features(fh, bbh, insn):
+                _bb_features[feature].add(addr)
+                func_features[feature].add(addr)
 
-        bb_features[capa.helpers.oint(bb)] = _bb_features
+        bb_features[bbh.address] = _bb_features
 
     return func_features, bb_features
 
 
-def find_func_matches(f, ruleset, func_features, bb_features):
+def find_func_matches(f: FunctionHandle, ruleset, func_features, bb_features):
     """ """
     func_matches = collections.defaultdict(list)
     bb_matches = collections.defaultdict(list)
@@ -97,10 +110,10 @@ def find_func_matches(f, ruleset, func_features, bb_features):
         for (name, res) in matches.items():
             bb_matches[name].extend(res)
             for (ea, _) in res:
-                func_features[capa.features.MatchedRule(name)].add(ea)
+                func_features[capa.features.common.MatchedRule(name)].add(ea)
 
     # find rule matches for function, function features include rule matches for basic blocks
-    _, matches = capa.engine.match(ruleset.function_rules, func_features, capa.helpers.oint(f))
+    _, matches = capa.engine.match(ruleset.function_rules, func_features, f.address)
     for (name, res) in matches.items():
         func_matches[name].extend(res)
 
@@ -109,19 +122,19 @@ def find_func_matches(f, ruleset, func_features, bb_features):
 
 def find_file_features(extractor):
     """ """
-    file_features = collections.defaultdict(set)
-    for (feature, ea) in extractor.extract_file_features():
-        if ea:
-            file_features[feature].add(ea)
+    file_features = collections.defaultdict(set)  # type: FeatureSet
+    for (feature, addr) in extractor.extract_file_features():
+        if addr:
+            file_features[feature].add(addr)
         else:
             if feature not in file_features:
                 file_features[feature] = set()
     return file_features
 
 
-def find_file_matches(ruleset, file_features):
+def find_file_matches(ruleset, file_features: FeatureSet):
     """ """
-    _, matches = capa.engine.match(ruleset.file_rules, file_features, 0x0)
+    _, matches = capa.engine.match(ruleset.file_rules, file_features, NO_ADDRESS)
     return matches
 
 
@@ -155,7 +168,7 @@ class CapaExplorerProgressIndicator(QtCore.QObject):
         self.progress.emit("extracting features from %s" % text)
 
 
-class CapaExplorerFeatureExtractor(capa.features.extractors.ida.IdaFeatureExtractor):
+class CapaExplorerFeatureExtractor(capa.features.extractors.ida.extractor.IdaFeatureExtractor):
     """subclass the IdaFeatureExtractor
 
     track progress during feature extraction, also allow user to cancel feature extraction
@@ -165,9 +178,9 @@ class CapaExplorerFeatureExtractor(capa.features.extractors.ida.IdaFeatureExtrac
         super(CapaExplorerFeatureExtractor, self).__init__()
         self.indicator = CapaExplorerProgressIndicator()
 
-    def extract_function_features(self, f):
-        self.indicator.update("function at 0x%X" % f.start_ea)
-        return super(CapaExplorerFeatureExtractor, self).extract_function_features(f)
+    def extract_function_features(self, fh: FunctionHandle):
+        self.indicator.update("function at 0x%X" % fh.inner.start_ea)
+        return super(CapaExplorerFeatureExtractor, self).extract_function_features(fh)
 
 
 class QLineEditClicked(QtWidgets.QLineEdit):
@@ -227,7 +240,7 @@ class CapaSettingsInputDialog(QtWidgets.QDialog):
 class CapaExplorerForm(idaapi.PluginForm):
     """form element for plugin interface"""
 
-    def __init__(self, name):
+    def __init__(self, name, option=Options.DEFAULT):
         """initialize form elements"""
         super(CapaExplorerForm, self).__init__()
 
@@ -267,6 +280,7 @@ class CapaExplorerForm(idaapi.PluginForm):
         self.view_rulegen_editor = None
         self.view_rulegen_header_label = None
         self.view_rulegen_search = None
+        self.view_rulegen_limit_features_by_ea = None
         self.rulegen_current_function = None
         self.rulegen_bb_features_cache = {}
         self.rulegen_func_features_cache = {}
@@ -275,6 +289,9 @@ class CapaExplorerForm(idaapi.PluginForm):
 
         self.Show()
 
+        if (option & Options.ANALYZE) == Options.ANALYZE:
+            self.analyze_program()
+
     def OnCreate(self, form):
         """called when plugin form is created
 
@@ -454,6 +471,10 @@ class CapaExplorerForm(idaapi.PluginForm):
         label2.setText("Editor")
         label2.setFont(font)
 
+        self.view_rulegen_limit_features_by_ea = QtWidgets.QCheckBox("Limit features to current dissasembly address")
+        self.view_rulegen_limit_features_by_ea.setChecked(False)
+        self.view_rulegen_limit_features_by_ea.stateChanged.connect(self.slot_checkbox_limit_features_by_ea)
+
         self.view_rulegen_status_label = QtWidgets.QLabel()
         self.view_rulegen_status_label.setAlignment(QtCore.Qt.AlignLeft)
         self.view_rulegen_status_label.setText("")
@@ -484,6 +505,7 @@ class CapaExplorerForm(idaapi.PluginForm):
         layout3.addWidget(self.view_rulegen_editor, 65)
 
         layout2.addWidget(self.view_rulegen_header_label)
+        layout2.addWidget(self.view_rulegen_limit_features_by_ea)
         layout2.addWidget(self.view_rulegen_search)
         layout2.addWidget(self.view_rulegen_features)
 
@@ -548,6 +570,10 @@ class CapaExplorerForm(idaapi.PluginForm):
         self.limit_results_to_function(idaapi.get_func(ea))
         self.view_tree.reset_ui()
 
+    def update_rulegen_tree_limit_features_to_selection(self, ea):
+        """ """
+        self.view_rulegen_features.filter_items_by_ea(ea)
+
     def ida_hook_screen_ea_changed(self, widget, new_ea, old_ea):
         """function hook for IDA "screen ea changed" action
 
@@ -568,6 +594,9 @@ class CapaExplorerForm(idaapi.PluginForm):
         if not idaapi.get_func(new_ea):
             return
 
+        if self.view_tabs.currentIndex() == 1 and self.view_rulegen_limit_features_by_ea.isChecked():
+            return self.update_rulegen_tree_limit_features_to_selection(new_ea)
+
         if idaapi.get_func(new_ea) == idaapi.get_func(old_ea):
             # user navigated same function - ignore
             return
@@ -603,7 +632,7 @@ class CapaExplorerForm(idaapi.PluginForm):
                 path = self.ask_user_directory()
                 if not path:
                     logger.warning(
-                        "You must select a file directory containing capa rules before analysis can be run. The standard collection of capa rules can be downloaded from https://github.com/fireeye/capa-rules."
+                        "You must select a file directory containing capa rules before analysis can be run. The standard collection of capa rules can be downloaded from https://github.com/mandiant/capa-rules."
                     )
                     return False
                 settings.user[CAPA_SETTINGS_RULE_PATH] = path
@@ -670,7 +699,7 @@ class CapaExplorerForm(idaapi.PluginForm):
             )
             logger.error("Failed to load rules from %s (error: %s).", settings.user[CAPA_SETTINGS_RULE_PATH], e)
             logger.error(
-                "Make sure your file directory contains properly formatted capa rules. You can download the standard collection of capa rules from https://github.com/fireeye/capa-rules."
+                "Make sure your file directory contains properly formatted capa rules. You can download the standard collection of capa rules from https://github.com/mandiant/capa-rules."
             )
             settings.user[CAPA_SETTINGS_RULE_PATH] = ""
             return False
@@ -727,6 +756,7 @@ class CapaExplorerForm(idaapi.PluginForm):
                 meta = capa.ida.helpers.collect_metadata()
                 capabilities, counts = capa.main.find_capabilities(self.ruleset_cache, extractor, disable_progress=True)
                 meta["analysis"].update(counts)
+                meta["analysis"]["layout"] = capa.main.compute_layout(self.ruleset_cache, extractor, capabilities)
             except UserCancelledError:
                 logger.info("User cancelled analysis.")
                 return False
@@ -770,7 +800,9 @@ class CapaExplorerForm(idaapi.PluginForm):
             update_wait_box("rendering results")
 
             try:
-                self.doc = capa.render.convert_capabilities_to_result_document(meta, self.ruleset_cache, capabilities)
+                self.doc = capa.render.result_document.convert_capabilities_to_result_document(
+                    meta, self.ruleset_cache, capabilities
+                )
             except Exception as e:
                 logger.error("Failed to render results (error: %s)", e)
                 return False
@@ -834,7 +866,7 @@ class CapaExplorerForm(idaapi.PluginForm):
             # must use extractor to get function, as capa analysis requires casted object
             extractor = CapaExplorerFeatureExtractor()
         except Exception as e:
-            logger.error("Failed to load IDA feature extractor (error: %s)" % e)
+            logger.error("Failed to load IDA feature extractor (error: %s)", e)
             return False
 
         if ida_kernwin.user_cancelled():
@@ -845,10 +877,10 @@ class CapaExplorerForm(idaapi.PluginForm):
         try:
             f = idaapi.get_func(idaapi.get_screen_ea())
             if f:
-                f = extractor.get_function(f.start_ea)
-                self.rulegen_current_function = f
+                fh: FunctionHandle = extractor.get_function(f.start_ea)
+                self.rulegen_current_function = fh
 
-                func_features, bb_features = find_func_features(f, extractor)
+                func_features, bb_features = find_func_features(fh, extractor)
                 self.rulegen_func_features_cache = collections.defaultdict(set, copy.copy(func_features))
                 self.rulegen_bb_features_cache = collections.defaultdict(dict, copy.copy(bb_features))
 
@@ -859,15 +891,15 @@ class CapaExplorerForm(idaapi.PluginForm):
 
                 try:
                     # add function and bb rule matches to function features, for display purposes
-                    func_matches, bb_matches = find_func_matches(f, self.ruleset_cache, func_features, bb_features)
-                    for (name, res) in itertools.chain(func_matches.items(), bb_matches.items()):
+                    func_matches, bb_matches = find_func_matches(fh, self.ruleset_cache, func_features, bb_features)
+                    for (name, addrs) in itertools.chain(func_matches.items(), bb_matches.items()):
                         rule = self.ruleset_cache[name]
                         if rule.meta.get("capa/subscope-rule"):
                             continue
-                        for (ea, _) in res:
-                            func_features[capa.features.MatchedRule(name)].add(ea)
+                        for (addr, _) in addrs:
+                            func_features[capa.features.common.MatchedRule(name)].add(addr)
                 except Exception as e:
-                    logger.error("Failed to match function/basic block rule scope (error: %s)" % e)
+                    logger.error("Failed to match function/basic block rule scope (error: %s)", e)
                     return False
             else:
                 func_features = {}
@@ -875,7 +907,7 @@ class CapaExplorerForm(idaapi.PluginForm):
             logger.info("User cancelled analysis.")
             return False
         except Exception as e:
-            logger.error("Failed to extract function features (error: %s)" % e)
+            logger.error("Failed to extract function features (error: %s)", e)
             return False
 
         if ida_kernwin.user_cancelled():
@@ -885,7 +917,7 @@ class CapaExplorerForm(idaapi.PluginForm):
 
         try:
             file_features = find_file_features(extractor)
-            self.rulegen_file_features_cache = collections.defaultdict(dict, copy.copy(file_features))
+            self.rulegen_file_features_cache = copy.copy(file_features)
 
             if ida_kernwin.user_cancelled():
                 logger.info("User cancelled analysis.")
@@ -894,17 +926,17 @@ class CapaExplorerForm(idaapi.PluginForm):
 
             try:
                 # add file matches to file features, for display purposes
-                for (name, res) in find_file_matches(self.ruleset_cache, file_features).items():
+                for (name, addrs) in find_file_matches(self.ruleset_cache, file_features).items():
                     rule = self.ruleset_cache[name]
                     if rule.meta.get("capa/subscope-rule"):
                         continue
-                    for (ea, _) in res:
-                        file_features[capa.features.MatchedRule(name)].add(ea)
+                    for (addr, _) in addrs:
+                        file_features[capa.features.common.MatchedRule(name)].add(addr)
             except Exception as e:
-                logger.error("Failed to match file scope rules (error: %s)" % e)
+                logger.error("Failed to match file scope rules (error: %s)", e)
                 return False
         except Exception as e:
-            logger.error("Failed to extract file features (error: %s)" % e)
+            logger.error("Failed to extract file features (error: %s)", e)
             return False
 
         if ida_kernwin.user_cancelled():
@@ -915,7 +947,7 @@ class CapaExplorerForm(idaapi.PluginForm):
         try:
             # load preview and feature tree
             self.view_rulegen_preview.load_preview_meta(
-                f.start_ea if f else None,
+                fh.address if fh else None,
                 settings.user.get(CAPA_SETTINGS_RULEGEN_AUTHOR, "<insert_author>"),
                 settings.user.get(CAPA_SETTINGS_RULEGEN_SCOPE, "function"),
             )
@@ -926,7 +958,7 @@ class CapaExplorerForm(idaapi.PluginForm):
                 "capa rules directory: %s (%d rules)" % (settings.user[CAPA_SETTINGS_RULE_PATH], len(self.rules_cache))
             )
         except Exception as e:
-            logger.error("Failed to render views (error: %s)" % e)
+            logger.error("Failed to render views (error: %s)", e, exc_info=True)
             return False
 
         return True
@@ -967,6 +999,7 @@ class CapaExplorerForm(idaapi.PluginForm):
         self.view_rulegen_editor.reset_view()
         self.view_rulegen_preview.reset_view()
         self.view_rulegen_search.clear()
+        self.view_rulegen_limit_features_by_ea.setChecked(False)
         self.set_rulegen_preview_border_neutral()
         self.rulegen_current_function = None
         self.rulegen_func_features_cache = {}
@@ -1005,7 +1038,7 @@ class CapaExplorerForm(idaapi.PluginForm):
 
     def update_rule_status(self, rule_text):
         """ """
-        if self.view_rulegen_editor.root is None:
+        if not self.view_rulegen_editor.invisibleRootItem().childCount():
             self.set_rulegen_preview_border_neutral()
             self.view_rulegen_status_label.clear()
             return
@@ -1123,9 +1156,9 @@ class CapaExplorerForm(idaapi.PluginForm):
             idaapi.info("No program analysis to save.")
             return
 
-        s = json.dumps(self.doc, sort_keys=True, cls=capa.render.CapaJsonObjectEncoder).encode("utf-8")
+        s = json.dumps(self.doc, sort_keys=True, cls=capa.render.json.CapaJsonObjectEncoder).encode("utf-8")
 
-        path = idaapi.ask_file(True, "*.json", "Choose file to save capa program analysis JSON")
+        path = self.ask_user_capa_json_file()
         if not path:
             return
 
@@ -1158,6 +1191,13 @@ class CapaExplorerForm(idaapi.PluginForm):
 
         self.view_tree.reset_ui()
 
+    def slot_checkbox_limit_features_by_ea(self, state):
+        """ """
+        if state == QtCore.Qt.Checked:
+            self.view_rulegen_features.filter_items_by_ea(idaapi.get_screen_ea())
+        else:
+            self.view_rulegen_features.show_all_items()
+
     def slot_checkbox_show_results_by_function_changed(self, state):
         """slot activated if checkbox clicked
 
@@ -1201,7 +1241,16 @@ class CapaExplorerForm(idaapi.PluginForm):
     def ask_user_capa_rule_file(self):
         """ """
         return QtWidgets.QFileDialog.getSaveFileName(
-            None, "Please select a capa rule to edit", settings.user.get(CAPA_SETTINGS_RULE_PATH, ""), "*.yml"
+            None,
+            "Please select a location to save capa rule file",
+            settings.user.get(CAPA_SETTINGS_RULE_PATH, ""),
+            "*.yml",
+        )[0]
+
+    def ask_user_capa_json_file(self):
+        """ """
+        return QtWidgets.QFileDialog.getSaveFileName(
+            None, "Please select a location to save capa JSON file", "", "*.json"
         )[0]
 
     def set_view_status_label(self, text):

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

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
@@ -6,7 +6,6 @@
 #  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 sys
 import codecs
 
 import idc
@@ -32,7 +31,7 @@ def location_to_hex(location):
     return "%08X" % location
 
 
-class CapaExplorerDataItem(object):
+class CapaExplorerDataItem:
     """store data for CapaExplorerDataModel"""
 
     def __init__(self, parent, data, can_check=True):
@@ -202,7 +201,7 @@ class CapaExplorerRuleMatchItem(CapaExplorerDataItem):
 
     @property
     def source(self):
-        """ return rule contents for display """
+        """return rule contents for display"""
         return self._source
 
 
@@ -217,6 +216,8 @@ class CapaExplorerFunctionItem(CapaExplorerDataItem):
         @param parent: parent node
         @param location: virtual address of function as seen by IDA
         """
+        # location can be an Address now, so need to get the VA
+        location = int(location)
         super(CapaExplorerFunctionItem, self).__init__(
             parent, [self.fmt % idaapi.get_name(location), location_to_hex(location), ""], can_check
         )
@@ -328,14 +329,10 @@ class CapaExplorerByteViewItem(CapaExplorerFeatureItem):
         """
         byte_snap = idaapi.get_bytes(location, 32)
 
+        details = ""
         if byte_snap:
             byte_snap = codecs.encode(byte_snap, "hex").upper()
-            if sys.version_info >= (3, 0):
-                details = " ".join([byte_snap[i : i + 2].decode() for i in range(0, len(byte_snap), 2)])
-            else:
-                details = " ".join([byte_snap[i : i + 2] for i in range(0, len(byte_snap), 2)])
-        else:
-            details = ""
+            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)

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
@@ -6,14 +6,17 @@
 #  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 List
+from collections import deque
 
 import idc
 import idaapi
 from PyQt5 import QtGui, QtCore
 
+import capa.rules
 import capa.ida.helpers
 import capa.render.utils as rutils
+import capa.features.common
 from capa.ida.plugin.item import (
     CapaExplorerDataItem,
     CapaExplorerRuleItem,
@@ -433,12 +436,18 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
             for ea in rule["matches"].keys():
                 ea = capa.ida.helpers.get_func_start_ea(ea)
                 if ea is None:
-                    # file scope, skip for rendering in this mode
+                    # file scope, skip rendering in this mode
                     continue
-                if None is matches_by_function.get(ea, None):
-                    matches_by_function[ea] = CapaExplorerFunctionItem(self.root_node, ea, can_check=False)
+                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
+                    continue
+                match_cache.append(rule["meta"]["name"])
                 CapaExplorerRuleItem(
-                    matches_by_function[ea],
+                    function_root,
                     rule["meta"]["name"],
                     rule["meta"].get("namespace"),
                     len(rule["matches"]),
@@ -492,7 +501,7 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
         value = feature[feature["type"]]
         if value:
             if key == "string":
-                value = '"%s"' % capa.features.escape_string(value)
+                value = '"%s"' % capa.features.common.escape_string(value)
             if feature.get("description", ""):
                 return "%s(%s = %s)" % (key, value, feature["description"])
             else:
@@ -537,6 +546,14 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
         @param location: address of feature
         @param display: text to display in plugin UI
         """
+
+        # convert to offset from locations: List[Address]
+        try:
+            location = int(location)
+        except TypeError:
+            # e.g. capa.features.address._NoAddress, global features
+            return
+
         # special handling for characteristic pending type
         if feature["type"] == "characteristic":
             if feature[feature["type"]] in ("embedded pe",):
@@ -554,10 +571,15 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
                 parent, display, source=doc["rules"].get(feature[feature["type"]], {}).get("source", "")
             )
 
-        if feature["type"] == "regex":
-            return CapaExplorerStringViewItem(
-                parent, display, location, '"%s"' % capa.features.escape_string(feature["match"])
-            )
+        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) + '"'
+                    )
+
+            # 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":
             return CapaExplorerBlockItem(parent, location)
@@ -568,10 +590,6 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
             "mnemonic",
             "number",
             "offset",
-            "number/x32",
-            "number/x64",
-            "offset/x32",
-            "offset/x64",
         ):
             # display instruction preview
             return CapaExplorerInstructionViewItem(parent, display, location)
@@ -583,11 +601,14 @@ class CapaExplorerDataModel(QtCore.QAbstractItemModel):
         if feature["type"] in ("string",):
             # display string preview
             return CapaExplorerStringViewItem(
-                parent, display, location, '"%s"' % capa.features.escape_string(feature[feature["type"]])
+                parent, display, location, '"%s"' % capa.features.common.escape_string(feature[feature["type"]])
             )
 
-        if feature["type"] in ("import", "export"):
+        if feature["type"] in ("import", "export", "function-name"):
             # display no preview
+            return CapaExplorerFeatureItem(parent, location=location, display=display)
+
+        if feature["type"] in ("arch", "os", "format"):
             return CapaExplorerFeatureItem(parent, display=display)
 
         raise RuntimeError("unexpected feature type: " + str(feature["type"]))

capa/ida/plugin/proxy.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 six
 from PyQt5 import QtCore
 from PyQt5.QtCore import Qt
 
@@ -208,7 +207,7 @@ class CapaExplorerSearchProxyModel(QtCore.QSortFilterProxyModel):
             if not data:
                 continue
 
-            if not isinstance(data, six.string_types):
+            if not isinstance(data, str):
                 # sanity check: should already be a string, but double check
                 continue
 

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
@@ -9,13 +9,16 @@ import re
 from collections import Counter
 
 import idc
+import idaapi
 from PyQt5 import QtGui, QtCore, QtWidgets
 
 import capa.rules
 import capa.engine
 import capa.ida.helpers
+import capa.features.common
 import capa.features.basicblock
 from capa.ida.plugin.item import CapaExplorerFunctionItem
+from capa.features.address import NO_ADDRESS
 from capa.ida.plugin.model import CapaExplorerDataModel
 
 MAX_SECTION_SIZE = 750
@@ -25,7 +28,7 @@ COLOR_GREEN_RGB = (79, 121, 66)
 COLOR_BLUE_RGB = (37, 147, 215)
 
 
-def calc_level_by_indent(line, prev_level=0):
+def calc_indent_from_line(line, prev_level=0):
     """ """
     if not len(line.strip()):
         # blank line, which may occur for comments so we simply use the last level
@@ -35,10 +38,13 @@ def calc_level_by_indent(line, prev_level=0):
         # need to adjust two spaces when encountering string description
         line = line[2:]
     # calc line level based on preceding whitespace
-    return len(line) - len(stripped)
+    indent = len(line) - len(stripped)
+
+    # round up to nearest even number; helps keep parsing more sane
+    return indent + (indent % 2)
 
 
-def parse_feature_for_node(feature):
+def parse_yaml_line(feature):
     """ """
     description = ""
     comment = ""
@@ -111,30 +117,6 @@ def parse_node_for_feature(feature, description, comment, depth):
     return display if display.endswith("\n") else display + "\n"
 
 
-def yaml_to_nodes(s):
-    level = 0
-    for line in s.splitlines():
-        feature, description, comment = parse_feature_for_node(line.strip())
-
-        o = QtWidgets.QTreeWidgetItem(None)
-
-        # set node attributes
-        setattr(o, "capa_level", calc_level_by_indent(line, level))
-
-        if feature.startswith(("- and:", "- or:", "- not:", "- basic block:", "- optional:")):
-            setattr(o, "capa_type", CapaExplorerRulgenEditor.get_node_type_expression())
-        elif feature.startswith("#"):
-            setattr(o, "capa_type", CapaExplorerRulgenEditor.get_node_type_comment())
-        else:
-            setattr(o, "capa_type", CapaExplorerRulgenEditor.get_node_type_feature())
-
-        # set node text
-        for (i, v) in enumerate((feature, description, comment)):
-            o.setText(i, v)
-
-        yield o
-
-
 def iterate_tree(o):
     """ """
     itr = QtWidgets.QTreeWidgetItemIterator(o)
@@ -143,6 +125,13 @@ def iterate_tree(o):
         itr += 1
 
 
+def expand_tree(root):
+    """ """
+    for node in iterate_tree(root):
+        if node.childCount() and not node.isExpanded():
+            node.setExpanded(True)
+
+
 def calc_item_depth(o):
     """ """
     depth = 0
@@ -177,6 +166,13 @@ def build_context_menu(o, actions):
     return menu
 
 
+def resize_columns_to_content(header):
+    """ """
+    header.resizeSections(QtWidgets.QHeaderView.ResizeToContents)
+    if header.sectionSize(0) > MAX_SECTION_SIZE:
+        header.resizeSection(0, MAX_SECTION_SIZE)
+
+
 class CapaExplorerRulgenPreview(QtWidgets.QTextEdit):
 
     INDENT = " " * 2
@@ -318,7 +314,6 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         self.preview = preview
 
         self.setHeaderLabels(["Feature", "Description", "Comment"])
-        self.header().setSectionResizeMode(QtWidgets.QHeaderView.ResizeToContents)
         self.header().setStretchLastSection(False)
         self.setExpandsOnDoubleClick(False)
         self.setEditTriggers(QtWidgets.QAbstractItemView.NoEditTriggers)
@@ -326,6 +321,10 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         self.setSelectionMode(QtWidgets.QAbstractItemView.ExtendedSelection)
         self.setStyleSheet("QTreeView::item {padding-right: 15 px;padding-bottom: 2 px;}")
 
+        # configure view columns to auto-resize
+        for idx in range(3):
+            self.header().setSectionResizeMode(idx, QtWidgets.QHeaderView.Interactive)
+
         # enable drag and drop
         self.setDragEnabled(True)
         self.setAcceptDrops(True)
@@ -335,8 +334,9 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
         self.itemChanged.connect(self.slot_item_changed)
         self.customContextMenuRequested.connect(self.slot_custom_context_menu_requested)
         self.itemDoubleClicked.connect(self.slot_item_double_clicked)
+        self.expanded.connect(self.slot_resize_columns_to_content)
+        self.collapsed.connect(self.slot_resize_columns_to_content)
 
-        self.root = None
         self.reset_view()
 
         self.is_editing = False
@@ -386,15 +386,17 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
 
         super(CapaExplorerRulgenEditor, self).dropEvent(e)
 
-        # self.prune_expressions()
         self.update_preview()
-        self.expandAll()
+        expand_tree(self.invisibleRootItem())
 
     def reset_view(self):
         """ """
-        self.root = None
         self.clear()
 
+    def slot_resize_columns_to_content(self):
+        """ """
+        resize_columns_to_content(self.header())
+
     def slot_item_changed(self, item, column):
         """ """
         if self.is_editing:
@@ -404,16 +406,21 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
     def slot_remove_selected(self, action):
         """ """
         for o in self.selectedItems():
-            if o == self.root:
+            if o.parent() is None:
+                # special handling for top-level items
                 self.takeTopLevelItem(self.indexOfTopLevelItem(o))
-                self.root = None
                 continue
             o.parent().removeChild(o)
 
     def slot_nest_features(self, action):
         """ """
-        # create a new parent under root node, by default; new node added last position in tree
-        new_parent = self.new_expression_node(self.root, (action.data()[0], ""))
+        # we don't want to add new features under the invisible root because capa rules should
+        # contain a single top-level node; this may not always be the case so we default to the last
+        # child node that was added to the invisible root
+        top_node = self.invisibleRootItem().child(self.invisibleRootItem().childCount() - 1)
+
+        # create a new parent under top-level node
+        new_parent = self.new_expression_node(top_node, (action.data()[0], ""))
 
         if "basic block" in action.data()[0]:
             # add default child expression when nesting under basic block
@@ -615,91 +622,138 @@ class CapaExplorerRulgenEditor(QtWidgets.QTreeWidget):
 
     def update_features(self, features):
         """ """
-        if not self.root:
-            # root node does not exist, create default node, set expanded
-            self.root = self.new_expression_node(self, ("- or:", ""))
+        if not self.invisibleRootItem().childCount():
+            # empty tree; add a default node
+            self.new_expression_node(self.invisibleRootItem(), ("- or:", ""))
+
+        # we don't want to add new features under the invisible root because capa rules should
+        # contain a single top-level node; this may not always be the case so we default to the last
+        # child node that was added to the invisible root
+        top_node = self.invisibleRootItem().child(self.invisibleRootItem().childCount() - 1)
 
         # build feature counts
         counted = list(zip(Counter(features).keys(), Counter(features).values()))
 
         # single features
         for (k, v) in filter(lambda t: t[1] == 1, counted):
-            if isinstance(k, (capa.features.String,)):
-                value = '"%s"' % capa.features.escape_string(k.get_value_str())
+            if isinstance(k, (capa.features.common.String,)):
+                value = '"%s"' % capa.features.common.escape_string(k.get_value_str())
             else:
                 value = k.get_value_str()
-            self.new_feature_node(self.root, ("- %s: %s" % (k.name.lower(), value), ""))
+            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):
             if k.value:
-                if isinstance(k, (capa.features.String,)):
-                    value = '"%s"' % capa.features.escape_string(k.get_value_str())
+                if isinstance(k, (capa.features.common.String,)):
+                    value = '"%s"' % capa.features.common.escape_string(k.get_value_str())
                 else:
                     value = k.get_value_str()
                 display = "- count(%s(%s)): %d" % (k.name.lower(), value, v)
             else:
                 display = "- count(%s): %d" % (k.name.lower(), v)
-            self.new_feature_node(self.root, (display, ""))
+            self.new_feature_node(top_node, (display, ""))
 
-        self.expandAll()
         self.update_preview()
+        expand_tree(self.invisibleRootItem())
+        resize_columns_to_content(self.header())
+
+    def make_child_node_from_feature(self, parent, feature):
+        """ """
+        feature, comment, description = feature
+
+        # we need special handling for the "description" tag; meaning we don't add a new node but simply
+        # set the "description" column for the appropriate parent node
+        if feature.startswith("description:"):
+            if not parent:
+                # we shouldn't have description without a parent; do nothing
+                return None
+
+            # we don't add a new node for description; either set description column of parent's last child
+            # or the parent itself
+            if parent.childCount():
+                parent.child(parent.childCount() - 1).setText(1, feature.lstrip("description:").lstrip())
+            else:
+                parent.setText(1, feature.lstrip("description:").lstrip())
+            return None
+        elif feature.startswith("- description:"):
+            if not parent:
+                # we shouldn't have a description without a parent; do nothing
+                return None
+
+            # we don't add a new node for description; set the description column of the parent instead
+            parent.setText(1, feature.lstrip("- description:").lstrip())
+            return None
+
+        node = QtWidgets.QTreeWidgetItem(parent)
+
+        # set node text to data parsed from feature
+        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())
+        elif feature.startswith("#"):
+            setattr(node, "capa_type", CapaExplorerRulgenEditor.get_node_type_comment())
+        else:
+            setattr(node, "capa_type", CapaExplorerRulgenEditor.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)
+
+        parent.addChild(node)
+
+        return node
 
     def load_features_from_yaml(self, rule_text, update_preview=False):
         """ """
-
-        def add_node(parent, node):
-            if node.text(0).startswith("description:"):
-                if parent.childCount():
-                    parent.child(parent.childCount() - 1).setText(1, node.text(0).lstrip("description:").lstrip())
-                else:
-                    parent.setText(1, node.text(0).lstrip("description:").lstrip())
-            elif node.text(0).startswith("- description:"):
-                parent.setText(1, node.text(0).lstrip("- description:").lstrip())
-            else:
-                parent.addChild(node)
-
-        def build(parent, nodes):
-            if nodes:
-                child_lvl = nodes[0].capa_level
-                while nodes:
-                    node = nodes.pop(0)
-                    if node.capa_level == child_lvl:
-                        add_node(parent, node)
-                    elif node.capa_level > child_lvl:
-                        nodes.insert(0, node)
-                        build(parent.child(parent.childCount() - 1), nodes)
-                    else:
-                        parent = parent.parent() if parent.parent() else parent
-                        add_node(parent, node)
-
         self.reset_view()
 
         # check for lack of features block
         if -1 == rule_text.find("features:"):
             return
 
-        rule_features = rule_text[rule_text.find("features:") + len("features:") :].strip()
-        rule_nodes = list(yaml_to_nodes(rule_features))
+        rule_features = rule_text[rule_text.find("features:") + len("features:") :].strip("\n")
 
-        # check for lack of nodes
-        if not rule_nodes:
+        if not rule_features:
+            # no features; nothing to do
             return
 
-        for o in rule_nodes:
-            (self.set_expression_node, self.set_feature_node, self.set_comment_node)[o.capa_type](o)
+        # build tree from yaml text using stack-based algorithm to build parent -> child edges
+        stack = [self.invisibleRootItem()]
+        for line in rule_features.splitlines():
+            if not len(line.strip()):
+                continue
 
-        self.root = rule_nodes.pop(0)
-        self.addTopLevelItem(self.root)
+            indent = calc_indent_from_line(line)
+
+            # we need to grow our stack to ensure proper parent -> child edges
+            if indent > len(stack):
+                stack.extend([None] * (indent - len(stack)))
+
+            # shave the stack; divide by 2 because even indent, add 1 to avoid shaving root node
+            stack[indent // 2 + 1 :] = []
+
+            # find our parent; should be last node in stack not None
+            parent = None
+            for o in stack[::-1]:
+                if o:
+                    parent = o
+                    break
+
+            node = self.make_child_node_from_feature(parent, parse_yaml_line(line.strip()))
+
+            # append our new node in case its a parent for another node
+            if node:
+                stack.append(node)
 
         if update_preview:
             self.preview.blockSignals(True)
             self.preview.setPlainText(rule_text)
             self.preview.blockSignals(False)
 
-        build(self.root, rule_nodes)
-
-        self.expandAll()
+        expand_tree(self.invisibleRootItem())
 
     def get_features(self, selected=False, ignore=()):
         """ """
@@ -735,9 +789,12 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
         self.editor = editor
 
         self.setHeaderLabels(["Feature", "Virtual Address"])
-        self.header().setSectionResizeMode(QtWidgets.QHeaderView.ResizeToContents)
         self.setStyleSheet("QTreeView::item {padding-right: 15 px;padding-bottom: 2 px;}")
 
+        # configure view columns to auto-resize
+        for idx in range(2):
+            self.header().setSectionResizeMode(idx, QtWidgets.QHeaderView.Interactive)
+
         self.setExpandsOnDoubleClick(False)
         self.setContextMenuPolicy(QtCore.Qt.CustomContextMenu)
         self.setSelectionMode(QtWidgets.QAbstractItemView.ExtendedSelection)
@@ -745,6 +802,8 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
         # connect slots
         self.itemDoubleClicked.connect(self.slot_item_double_clicked)
         self.customContextMenuRequested.connect(self.slot_custom_context_menu_requested)
+        self.expanded.connect(self.slot_resize_columns_to_content)
+        self.collapsed.connect(self.slot_resize_columns_to_content)
 
         self.reset_view()
 
@@ -772,12 +831,24 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
         """ """
         self.clear()
 
+    def slot_resize_columns_to_content(self):
+        """ """
+        resize_columns_to_content(self.header())
+
     def slot_add_selected_features(self, action):
         """ """
         selected = [item.data(0, 0x100) for item in self.selectedItems()]
         if selected:
             self.editor.update_features(selected)
 
+    def slot_add_n_bytes_feature(self, action):
+        """ """
+        count = idaapi.ask_long(16, f"Enter number of bytes (1-{capa.features.common.MAX_BYTES_FEATURE_SIZE}):")
+        if count and 1 <= count <= capa.features.common.MAX_BYTES_FEATURE_SIZE:
+            item = self.selectedItems()[0].data(0, 0x100)
+            item.value = item.value[:count]
+            self.editor.update_features([item])
+
     def slot_custom_context_menu_requested(self, pos):
         """ """
         actions = []
@@ -789,6 +860,8 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
 
         if selected_items_count == 1:
             action_add_features_fmt = "Add feature"
+            if isinstance(self.selectedItems()[0].data(0, 0x100), capa.features.common.Bytes):
+                actions.append(("Add n bytes...", (), self.slot_add_n_bytes_feature))
         else:
             action_add_features_fmt = "Add %d features" % selected_items_count
 
@@ -818,13 +891,58 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
                 if data:
                     to_match = data.get_value_str()
                     if not to_match or text.lower() not in to_match.lower():
-                        o.setHidden(True)
+                        if not o.isHidden():
+                            o.setHidden(True)
                         continue
-                o.setHidden(False)
-                o.setExpanded(True)
+                if o.isHidden():
+                    o.setHidden(False)
+                if o.childCount() and not o.isExpanded():
+                    o.setExpanded(True)
         else:
             self.show_all_items()
 
+    def filter_items_by_ea(self, min_ea, max_ea=None):
+        """ """
+        visited = []
+
+        def show_item_and_parents(_o):
+            """iteratively show and expand an item and its' parents"""
+            while _o:
+                visited.append(_o)
+                if _o.isHidden():
+                    _o.setHidden(False)
+                if _o.childCount() and not _o.isExpanded():
+                    _o.setExpanded(True)
+                _o = _o.parent()
+
+        for o in iterate_tree(self):
+            if o in visited:
+                # save some cycles, only visit item once
+                continue
+
+            # read ea from "Address" column
+            o_ea = o.text(CapaExplorerRulegenFeatures.get_column_address_index())
+
+            if o_ea == "":
+                # ea may be empty, hide by default
+                if not o.isHidden():
+                    o.setHidden(True)
+                continue
+
+            o_ea = int(o_ea, 16)
+
+            if max_ea is not None and min_ea <= o_ea <= max_ea:
+                show_item_and_parents(o)
+            elif o_ea == min_ea:
+                show_item_and_parents(o)
+            else:
+                # made it here, hide by default
+                if not o.isHidden():
+                    o.setHidden(True)
+
+        # resize the view for UX
+        resize_columns_to_content(self.header())
+
     def style_parent_node(self, o):
         """ """
         font = QtGui.QFont()
@@ -886,23 +1004,26 @@ class CapaExplorerRulegenFeatures(QtWidgets.QTreeWidget):
         self.parse_features_for_tree(self.new_parent_node(self, ("File Scope",)), file_features)
         if func_features:
             self.parse_features_for_tree(self.new_parent_node(self, ("Function/Basic Block Scope",)), func_features)
+        resize_columns_to_content(self.header())
 
     def parse_features_for_tree(self, parent, features):
         """ """
         self.parent_items = {}
 
         def format_address(e):
+            if e == NO_ADDRESS:
+                return ""
             return "%X" % e if e else ""
 
         def format_feature(feature):
             """ """
             name = feature.name.lower()
             value = feature.get_value_str()
-            if isinstance(feature, (capa.features.String,)):
-                value = '"%s"' % capa.features.escape_string(value)
+            if isinstance(feature, (capa.features.common.String,)):
+                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
@@ -914,7 +1035,7 @@ 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
                     )
@@ -924,14 +1045,18 @@ 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:
-                ea = eas.pop()
-                for (i, v) in enumerate((format_feature(feature), format_address(ea))):
+                if addrs:
+                    addr = addrs.pop()
+                else:
+                    # some features may not have an address e.g. "format"
+                    addr = ""
+                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)
 
@@ -999,11 +1124,7 @@ class CapaExplorerQtreeView(QtWidgets.QTreeView):
     def slot_resize_columns_to_content(self):
         """reset view columns to contents"""
         if self.should_resize_columns:
-            self.header().resizeSections(QtWidgets.QHeaderView.ResizeToContents)
-
-            # limit size of first section
-            if self.header().sectionSize(0) > MAX_SECTION_SIZE:
-                self.header().resizeSection(0, MAX_SECTION_SIZE)
+            resize_columns_to_content(self.header())
 
     def map_index_to_source_item(self, model_index):
         """map proxy model index to source model item

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=lambda n: get_node_cost(n))
+        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 collections
+from typing import Dict
+
+# this structure is unstable and may change before the next major release.
+counters: Dict[str, int] = collections.Counter()
+
+
+def reset():
+    global counters
+    counters = collections.Counter()

capa/render/__init__.py

@@ -1,266 +0,0 @@
-# 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 six
-
-import capa.rules
-import capa.engine
-
-
-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
-
-
-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 == "regex":
-        result["match"] = feature.match
-    return result
-
-
-def convert_node_to_result_document(node):
-    """
-    "node": {
-        "type": "statement",
-        "statement": { ... }
-    },
-
-    "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.Feature):
-        return {
-            "type": "feature",
-            "feature": convert_feature_to_result_document(node),
-        }
-    else:
-        raise RuntimeError("unexpected match node type")
-
-
-def convert_match_to_result_document(rules, capabilities, result):
-    """
-    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.
-    """
-    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.Feature):
-        if bool(result.success):
-            doc["locations"] = result.locations
-    elif isinstance(result.statement, capa.rules.Range):
-        if bool(result.success):
-            doc["locations"] = result.locations
-
-    # if we have a `match` statement, then we're referencing another rule.
-    # 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"]
-    ):
-
-        rule_name = doc["node"]["feature"]["match"]
-        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,
-                },
-            }
-
-        for location in doc["locations"]:
-            doc["children"].append(convert_match_to_result_document(rules, capabilities, rule_matches[location]))
-
-    return doc
-
-
-def convert_capabilities_to_result_document(meta, rules, capabilities):
-    """
-    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 `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
-
-        doc["rules"][rule_name] = {
-            "meta": dict(rule.meta),
-            "source": rule.definition,
-            "matches": {
-                addr: convert_match_to_result_document(rules, capabilities, match) for (addr, match) in matches
-            },
-        }
-
-    return doc
-
-
-def render_vverbose(meta, rules, capabilities):
-    # there's an import loop here
-    # if capa.render imports capa.render.vverbose
-    # and capa.render.vverbose import capa.render (implicitly, as a submodule)
-    # so, defer the import until routine is called, breaking the import loop.
-    import capa.render.vverbose
-
-    doc = convert_capabilities_to_result_document(meta, rules, capabilities)
-    return capa.render.vverbose.render_vverbose(doc)
-
-
-def render_verbose(meta, rules, capabilities):
-    # break import loop
-    import capa.render.verbose
-
-    doc = convert_capabilities_to_result_document(meta, rules, capabilities)
-    return capa.render.verbose.render_verbose(doc)
-
-
-def render_default(meta, rules, capabilities):
-    # break import loop
-    import capa.render.default
-    import capa.render.verbose
-
-    doc = convert_capabilities_to_result_document(meta, rules, capabilities)
-    return capa.render.default.render_default(doc)
-
-
-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, six.string_types):
-            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_json(meta, rules, capabilities):
-    return json.dumps(
-        convert_capabilities_to_result_document(meta, rules, capabilities),
-        cls=CapaJsonObjectEncoder,
-        sort_keys=True,
-    )

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
@@ -8,15 +8,18 @@
 
 import collections
 
-import six
 import tabulate
 
 import capa.render.utils as rutils
+import capa.render.result_document
+from capa.rules import RuleSet
+from capa.engine import MatchResults
+from capa.render.utils import StringIO
 
 tabulate.PRESERVE_WHITESPACE = True
 
 
-def width(s, character_count):
+def width(s: str, character_count: int) -> str:
     """pad the given string to at least `character_count`"""
     if len(s) < character_count:
         return s + " " * (character_count - len(s))
@@ -24,11 +27,14 @@ def width(s, character_count):
         return s
 
 
-def render_meta(doc, ostream):
+def render_meta(doc, 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"]),
     ]
 
@@ -58,13 +64,13 @@ def find_subrule_matches(doc):
                 matches.add(node["node"]["feature"]["match"])
 
     for rule in rutils.capability_rules(doc):
-        for node in rule["matches"].values():
+        for address, node in rule["matches"]:
             rec(node)
 
     return matches
 
 
-def render_capabilities(doc, ostream):
+def render_capabilities(doc, ostream: StringIO):
     """
     example::
 
@@ -102,7 +108,7 @@ def render_capabilities(doc, ostream):
         ostream.writeln(rutils.bold("no capabilities found"))
 
 
-def render_attack(doc, ostream):
+def render_attack(doc, ostream: StringIO):
     """
     example::
 
@@ -124,27 +130,16 @@ def render_attack(doc, ostream):
             continue
 
         for attack in rule["meta"]["att&ck"]:
-            tactic, _, rest = attack.partition("::")
-            if "::" in rest:
-                technique, _, rest = rest.partition("::")
-                subtechnique, _, id = rest.rpartition(" ")
-                tactics[tactic].add((technique, subtechnique, id))
-            else:
-                technique, _, id = rest.rpartition(" ")
-                tactics[tactic].add((technique, id))
+            tactics[attack["tactic"]].add((attack["technique"], attack.get("subtechnique"), attack["id"]))
 
     rows = []
     for tactic, techniques in sorted(tactics.items()):
         inner_rows = []
-        for spec in sorted(techniques):
-            if len(spec) == 2:
-                technique, id = spec
+        for (technique, subtechnique, id) in sorted(techniques):
+            if subtechnique is None:
                 inner_rows.append("%s %s" % (rutils.bold(technique), id))
-            elif len(spec) == 3:
-                technique, subtechnique, id = spec
-                inner_rows.append("%s::%s %s" % (rutils.bold(technique), subtechnique, id))
             else:
-                raise RuntimeError("unexpected ATT&CK spec format")
+                inner_rows.append("%s::%s %s" % (rutils.bold(technique), subtechnique, id))
         rows.append(
             (
                 rutils.bold(tactic.upper()),
@@ -161,7 +156,7 @@ def render_attack(doc, ostream):
         ostream.write("\n")
 
 
-def render_mbc(doc, ostream):
+def render_mbc(doc, ostream: StringIO):
     """
     example::
 
@@ -180,32 +175,17 @@ def render_mbc(doc, ostream):
         if not rule["meta"].get("mbc"):
             continue
 
-        mbcs = rule["meta"]["mbc"]
-        if not isinstance(mbcs, list):
-            raise ValueError("invalid rule: MBC mapping is not a list")
-
-        for mbc in mbcs:
-            objective, _, rest = mbc.partition("::")
-            if "::" in rest:
-                behavior, _, rest = rest.partition("::")
-                method, _, id = rest.rpartition(" ")
-                objectives[objective].add((behavior, method, id))
-            else:
-                behavior, _, id = rest.rpartition(" ")
-                objectives[objective].add((behavior, id))
+        for mbc in rule["meta"]["mbc"]:
+            objectives[mbc["objective"]].add((mbc["behavior"], mbc.get("method"), mbc["id"]))
 
     rows = []
     for objective, behaviors in sorted(objectives.items()):
         inner_rows = []
-        for spec in sorted(behaviors):
-            if len(spec) == 2:
-                behavior, id = spec
-                inner_rows.append("%s %s" % (rutils.bold(behavior), id))
-            elif len(spec) == 3:
-                behavior, method, id = spec
-                inner_rows.append("%s::%s %s" % (rutils.bold(behavior), method, id))
+        for (behavior, method, id) in sorted(behaviors):
+            if method is None:
+                inner_rows.append("%s [%s]" % (rutils.bold(behavior), id))
             else:
-                raise RuntimeError("unexpected MBC spec format")
+                inner_rows.append("%s::%s [%s]" % (rutils.bold(behavior), method, id))
         rows.append(
             (
                 rutils.bold(objective.upper()),
@@ -232,3 +212,8 @@ def render_default(doc):
     render_capabilities(doc, ostream)
 
     return ostream.getvalue()
+
+
+def render(meta, rules: RuleSet, capabilities: MatchResults) -> str:
+    doc = capa.render.result_document.convert_capabilities_to_result_document(meta, rules, capabilities)
+    return render_default(doc)

capa/render/json.py

@@ -0,0 +1,33 @@
+# 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
+
+from capa.rules import RuleSet
+from capa.engine import MatchResults
+from capa.render.result_document import convert_capabilities_to_result_document
+
+
+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(
+        convert_capabilities_to_result_document(meta, rules, capabilities),
+        cls=CapaJsonObjectEncoder,
+        sort_keys=True,
+    )

capa/render/result_document.py

@@ -0,0 +1,390 @@
+# 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
+from typing import Any, List
+
+import capa.engine
+import capa.render.utils
+import capa.features.common
+import capa.features.address
+from capa.rules import RuleSet
+from capa.engine import MatchResults
+from capa.helpers import assert_never
+from capa.features.address import Address
+
+
+def serialize_address(a: Address) -> Any:
+    if isinstance(a, capa.features.address.AbsoluteVirtualAddress):
+        return ("absolute", int(a))
+
+    elif isinstance(a, capa.features.address.RelativeVirtualAddress):
+        return ("relative", int(a))
+
+    elif isinstance(a, capa.features.address.FileOffsetAddress):
+        return ("file", int(a))
+
+    elif isinstance(a, capa.features.address.DNTokenAddress):
+        return ("dn token", a.token)
+
+    elif isinstance(a, capa.features.address.DNTokenOffsetAddress):
+        return ("dn token offset", a.token, a.offset)
+
+    elif a == capa.features.address.NO_ADDRESS or isinstance(a, capa.features.address._NoAddress):
+        return ("no address",)
+
+    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 deserialize_address(doc: List[Any]) -> Address:
+    atype = doc[0]
+
+    if atype == "absolute":
+        return capa.features.address.AbsoluteVirtualAddress(doc[1])
+
+    elif atype == "relative":
+        return capa.features.address.RelativeVirtualAddress(doc[1])
+
+    elif atype == "file":
+        return capa.features.address.FileOffsetAddress(doc[1])
+
+    elif atype == "dn token":
+        return capa.features.address.DNTokenAddress(doc[1])
+
+    elif atype == "dn token offset":
+        return capa.features.address.DNTokenOffsetAddress(doc[1], doc[2])
+
+    elif atype == "no address":
+        return capa.features.address.NO_ADDRESS
+
+    else:
+        assert_never(atype)
+
+
+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
+
+
+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"):
+        if feature.matches:
+            # regex featur matches are a dict from the capture group to list of location addresses
+            result["matches"] = {k: list(map(serialize_address, vs)) for k, vs in feature.matches.items()}
+        else:
+            # there were no matches
+            pass
+
+    return result
+
+
+def convert_node_to_result_document(node):
+    """
+    "node": {
+        "type": "statement",
+        "statement": { ... }
+    },
+
+    "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),
+        }
+    else:
+        raise RuntimeError("unexpected match node type")
+
+
+def convert_match_to_result_document(rules, capabilities, result):
+    """
+    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.
+    """
+    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"] = list(map(serialize_address, result.locations))
+    elif isinstance(result.statement, capa.engine.Range):
+        if bool(result.success):
+            doc["locations"] = list(map(serialize_address, result.locations))
+
+    # 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"]
+    ):
+
+        name = doc["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 = doc["node"]["feature"]["match"]
+            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,
+                    },
+                }
+
+            for location in result.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]
+
+            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:
+                            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.
+    """
+    meta["analysis"]["base_address"] = serialize_address(meta["analysis"]["base_address"])
+
+    meta["analysis"]["feature_counts"]["functions"] = [
+        {"address": serialize_address(address), "count": count}
+        for address, count in meta["analysis"]["feature_counts"]["functions"].items()
+    ]
+
+    meta["analysis"]["library_functions"] = [
+        {"address": serialize_address(address), "name": name}
+        for address, name in meta["analysis"]["library_functions"].items()
+    ]
+
+    meta["analysis"]["layout"]["functions"] = [
+        {
+            "address": serialize_address(faddr),
+            "matched_basic_blocks": list({"address": serialize_address(bb)} for bb in f["matched_basic_blocks"]),
+        }
+        for faddr, f in meta["analysis"]["layout"]["functions"].items()
+    ]
+
+    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": [
+                [serialize_address(addr), convert_match_to_result_document(rules, capabilities, match)]
+                for (addr, match) in matches
+            ],
+        }
+
+    return doc

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
@@ -6,21 +6,22 @@
 #  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 six
+import io
+
 import termcolor
 
 
-def bold(s):
+def bold(s: str) -> str:
     """draw attention to the given string"""
     return termcolor.colored(s, "blue")
 
 
-def bold2(s):
+def bold2(s: str) -> str:
     """draw attention to the given string, within a `bold` section"""
     return termcolor.colored(s, "green")
 
 
-def hex(n):
+def hex(n: int) -> str:
     """render the given number using upper case hex, like: 0x123ABC"""
     if n < 0:
         return "-0x%X" % (-n)
@@ -28,6 +29,24 @@ def hex(n):
         return "0x%X" % n
 
 
+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):
+    """
+    format canonical representation of ATT&CK/MBC parts and ID
+    """
+    return "%s [%s]" % ("::".join(data["parts"]), data["id"])
+
+
 def capability_rules(doc):
     """enumerate the rules in (namespace, name) order that are 'capability' rules (not lib/subscope/disposition/etc)."""
     for (_, _, rule) in sorted(
@@ -41,6 +60,8 @@ def capability_rules(doc):
             continue
         if rule["meta"].get("maec/analysis-conclusion-ov"):
             continue
+        if rule["meta"].get("maec/malware-family"):
+            continue
         if rule["meta"].get("maec/malware-category"):
             continue
         if rule["meta"].get("maec/malware-category-ov"):
@@ -49,7 +70,7 @@ def capability_rules(doc):
         yield rule
 
 
-class StringIO(six.StringIO):
+class StringIO(io.StringIO):
     def writeln(self, s):
         self.write(s)
         self.write("\n")

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
@@ -23,9 +23,42 @@ Unless required by applicable law or agreed to in writing, software distributed
 See the License for the specific language governing permissions and limitations under the License.
 """
 import tabulate
+import dnfile.mdtable
 
 import capa.rules
 import capa.render.utils as rutils
+import capa.render.result_document
+import capa.render.result_document as rd
+from capa.rules import RuleSet
+from capa.engine import MatchResults
+from capa.features.address import (
+    NO_ADDRESS,
+    Address,
+    DNTokenAddress,
+    FileOffsetAddress,
+    DNTokenOffsetAddress,
+    AbsoluteVirtualAddress,
+    RelativeVirtualAddress,
+)
+
+
+def format_address(address: Address) -> str:
+    if isinstance(address, AbsoluteVirtualAddress):
+        return rutils.hex(int(address))
+    elif isinstance(address, RelativeVirtualAddress):
+        return f"base address+{rutils.hex(int(address))}"
+    elif isinstance(address, FileOffsetAddress):
+        return f"file+{rutils.hex(int(address))}"
+    elif isinstance(address, DNTokenAddress):
+        return str(address)
+    elif isinstance(address, DNTokenOffsetAddress):
+        name = dnfile.mdtable.ClrMetaDataTableFactory._table_number_map[address.token.table].name
+        rid = address.token.rid
+        return f"{name}[{rid}]+{rutils.hex(int(address.offset))}"
+    elif address == NO_ADDRESS:
+        return "global"
+    else:
+        raise ValueError("unexpected address type")
 
 
 def render_meta(ostream, doc):
@@ -38,7 +71,9 @@ def render_meta(ostream, doc):
         path                 /tmp/suspicious.dll_
         timestamp            2020-07-03T10:17:05.796933
         capa version         0.0.0
-        format               auto
+        os                   windows
+        format               pe
+        arch                 amd64
         extractor            VivisectFeatureExtractor
         base address         0x10000000
         rules                (embedded rules)
@@ -52,17 +87,21 @@ def render_meta(ostream, doc):
         ("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"]),
+        ("base address", format_address(rd.deserialize_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()),
+            + sum(map(lambda f: f["count"], doc["meta"]["analysis"]["feature_counts"]["functions"])),
         ),
     ]
+
     ostream.writeln(tabulate.tabulate(rows, tablefmt="plain"))
 
 
@@ -99,8 +138,8 @@ def render_rules(ostream, doc):
             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))))
+            locations = list(map(lambda m: m[0], doc["rules"][rule["meta"]["name"]]["matches"]))
+            rows.append(("matches", "\n".join(map(lambda d: format_address(rd.deserialize_address(d)), locations))))
 
         ostream.writeln(tabulate.tabulate(rows, tablefmt="plain"))
         ostream.write("\n")
@@ -119,3 +158,8 @@ def render_verbose(doc):
     ostream.write("\n")
 
     return ostream.getvalue()
+
+
+def render(meta, rules: RuleSet, capabilities: MatchResults) -> str:
+    doc = rd.convert_capabilities_to_result_document(meta, rules, capabilities)
+    return render_verbose(doc)

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
@@ -6,38 +6,48 @@
 #  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 collections
-
 import tabulate
 
 import capa.rules
 import capa.render.utils as rutils
 import capa.render.verbose
+import capa.features.common
+import capa.render.result_document as rd
+from capa.rules import RuleSet
+from capa.engine import MatchResults
 
 
 def render_locations(ostream, match):
+    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", [])))
     if len(locations) == 1:
         ostream.write(" @ ")
-        ostream.write(rutils.hex(locations[0]))
+        ostream.write(v.format_address(rd.deserialize_address(locations[0])))
     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(", ".join(map(lambda d: v.format_address(rd.deserialize_address(d)), locations[0:4])))
             ostream.write(", and %d more..." % (len(locations) - 4))
         else:
-            ostream.write(", ".join(map(rutils.hex, locations)))
+            ostream.write(", ".join(map(lambda d: v.format_address(rd.deserialize_address(d)), locations)))
 
 
 def render_statement(ostream, match, statement, indent=0):
     ostream.write("  " * indent)
     if statement["type"] in ("and", "or", "optional", "not", "subscope"):
-        ostream.write(statement["type"])
+        if statement["type"] == "subscope":
+            # emit `basic block:`
+            # rather than `subscope:`
+            ostream.write(statement["subscope"])
+        else:
+            # emit `and:`
+            ostream.write(statement["type"])
         ostream.write(":")
         if statement.get("description"):
             ostream.write(" = %s" % statement["description"])
@@ -57,7 +67,7 @@ def render_statement(ostream, match, statement, indent=0):
 
         if child[child["type"]]:
             if child["type"] == "string":
-                value = '"%s"' % capa.features.escape_string(child[child["type"]])
+                value = '"%s"' % capa.features.common.escape_string(child[child["type"]])
             else:
                 value = child[child["type"]]
             value = rutils.bold2(value)
@@ -85,30 +95,51 @@ def render_statement(ostream, match, statement, indent=0):
         raise RuntimeError("unexpected match statement type: " + str(statement))
 
 
+def render_string_value(s):
+    return '"%s"' % capa.features.common.escape_string(s)
+
+
 def render_feature(ostream, match, feature, indent=0):
     ostream.write("  " * indent)
 
     key = feature["type"]
     value = feature[feature["type"]]
-    if key == "regex":
-        key = "string"  # render string for regex to mirror the rule source
-        value = feature["match"]  # the match provides more information than the value for regex
 
-    if key == "string":
-        value = '"%s"' % capa.features.escape_string(value)
+    if key not in ("regex", "substring"):
+        # like:
+        #   number: 10 = SOME_CONSTANT @ 0x401000
+        if key == "string":
+            value = render_string_value(value)
 
-    ostream.write(key)
-    ostream.write(": ")
+        ostream.write(key)
+        ostream.write(": ")
 
-    if value:
-        ostream.write(rutils.bold2(value))
+        if value:
+            ostream.write(rutils.bold2(value))
 
-        if "description" in feature:
-            ostream.write(capa.rules.DESCRIPTION_SEPARATOR)
-            ostream.write(feature["description"])
+            if "description" in feature:
+                ostream.write(capa.rules.DESCRIPTION_SEPARATOR)
+                ostream.write(feature["description"])
 
-    render_locations(ostream, match)
-    ostream.write("\n")
+        if key not in ("os", "arch"):
+            render_locations(ostream, match)
+        ostream.write("\n")
+    else:
+        # like:
+        #  regex: /blah/ = SOME_CONSTANT
+        #    - "foo blah baz" @ 0x401000
+        #    - "aaa blah bbb" @ 0x402000, 0x403400
+        ostream.write(key)
+        ostream.write(": ")
+        ostream.write(value)
+        ostream.write("\n")
+
+        for match, locations in sorted(feature["matches"].items(), key=lambda p: p[0]):
+            ostream.write("  " * (indent + 1))
+            ostream.write("- ")
+            ostream.write(rutils.bold2(render_string_value(match)))
+            render_locations(ostream, {"locations": locations})
+            ostream.write("\n")
 
 
 def render_node(ostream, match, node, indent=0):
@@ -170,18 +201,35 @@ def render_rules(ostream, doc):
         ## 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 = {}
+    for finfo in doc["meta"]["analysis"]["layout"]["functions"]:
+        faddress = rd.deserialize_address(finfo["address"])
+
+        for bb in finfo["matched_basic_blocks"]:
+            bbaddress = rd.deserialize_address(bb["address"])
+            functions_by_bb[bbaddress] = faddress
+
     had_match = False
-    for rule in rutils.capability_rules(doc):
+
+    for (_, _, rule) in sorted(
+        map(lambda rule: (rule["meta"].get("namespace", ""), 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"].get("capa/subscope"):
+            continue
+
         count = len(rule["matches"])
         if count == 1:
             capability = rutils.bold(rule["meta"]["name"])
@@ -196,7 +244,20 @@ def render_rules(ostream, doc):
             if key == "name" or key not in rule["meta"]:
                 continue
 
+            if key == "examples":
+                # I can't think of a reason that an analyst would pivot to the concrete example
+                # directly from the capa output.
+                # the more likely flow is to review the rule and go from there.
+                # so, don't make the output messy by showing the examples.
+                continue
+
             v = rule["meta"][key]
+            if not v:
+                continue
+
+            if key in ("att&ck", "mbc"):
+                v = [rutils.format_parts_id(vv) for vv in v]
+
             if isinstance(v, list) and len(v) == 1:
                 v = v[0]
             elif isinstance(v, list) and len(v) > 1:
@@ -206,19 +267,27 @@ def render_rules(ostream, doc):
         ostream.writeln(tabulate.tabulate(rows, tablefmt="plain"))
 
         if rule["meta"]["scope"] == capa.rules.FILE_SCOPE:
-            matches = list(doc["rules"][rule["meta"]["name"]]["matches"].values())
+            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()):
+            for location, match in sorted(doc["rules"][rule["meta"]["name"]]["matches"]):
+                location = rd.deserialize_address(location)
+
                 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(functions_by_bb[location]))
+
+                ostream.write("\n")
                 render_match(ostream, match, indent=1)
         ostream.write("\n")
 
@@ -236,3 +305,8 @@ def render_vverbose(doc):
     ostream.write("\n")
 
     return ostream.getvalue()
+
+
+def render(meta, rules: RuleSet, capabilities: MatchResults) -> str:
+    doc = rd.convert_capabilities_to_result_document(meta, rules, capabilities)
+    return render_vverbose(doc)

capa/version.py

@@ -1 +1 @@
-__version__ = "1.6.1"
+__version__ = "3.2.0"

doc/installation.md

@@ -1,8 +1,8 @@
 # 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.
 
@@ -26,7 +26,11 @@ 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 doesn't pull the default rule set, so you should check it out separately from [capa-rules](https://github.com/mandiant/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/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:
+
+    capa   -r /path/to/capa-rules   -s /path/to/capa-sigs  suspicious.exe
 Alternatively, see Method 3 below.
 
 ### 1. Install capa module
@@ -40,15 +44,16 @@ If you'd like to review and modify the capa source code, you'll need to check it
 
 ### 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).
+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 clone everything use the `--recurse-submodules` option:
-- `$ 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)
+- 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/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 (common), 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`
 - `$ git submodule update --init rules`
 
@@ -59,40 +64,57 @@ Use `pip` to install the source code in "editable" mode. This means that Python
 
 You'll find that the `capa.exe` (Windows) or `capa` (Linux/MacOS) executables in your path now invoke the capa binary from this directory.
 
+#### Development
+
+##### venv [optional]
+
+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 activate `capa-env` in Linux or MacOS, run:
+`$ source ../capa-env/bin/activate`
+
+To activate `capa-env` in Windows, run:
+`$ ..\capa-env\Scripts\activate.bat`
+
+For more details about creating and using virtual environments, check out the [venv documentation](https://docs.python.org/3/tutorial/venv.html).
+
+##### 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`
   - [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:
 
 `$ pip install -e /local/path/to/src[dev]`
 
-Note that some development dependencies (including the black code formatter) require Python 3.
-
 To check the code style, formatting and run the tests you can run the script `scripts/ci.sh`.
 You can run it with the argument `no_tests` to skip the tests and only run the code style and formatting: `scripts/ci.sh no_tests`
 
+##### Setup hooks [optional]
+
+If you plan to contribute to capa, you may want to setup the 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.
+- The `pre-push` hook runs checks before every `git push`.
+  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.
+
 ### 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.
 
 #### Install PyInstaller:
-For Python 2.7: `$ pip install 'pyinstaller==3.*'` (PyInstaller 4 doesn't support Python 2.7)
-
-For Python 3: `$ pip install 'pyinstaller`
+`$ pip install pyinstaller` (Python 3)
 
 #### Run Pyinstaller
 `$ pyinstaller .github/pyinstaller/pyinstaller.spec`
 
 You can find the compiled binary in the created directory `dist/`.
-
-### 4. Setup hooks [optional]
-If you plan to contribute to capa, you may want to setup the 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.
-  You can skip this check by using the `--no-verify` git option.
-- The `pre-push` hook runs checks before every `git push`.
-  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.

doc/limitations.md

@@ -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

@@ -1,22 +1,16 @@
 # 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 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).
 - [ ] 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
-    - New Features
-    - New Rules
-    - Bug Fixes
-    - Changes
-    - Development
-    - Raw diffs
+  - Sections: see template below
   - Update `Raw diffs` links
   - Create placeholder for `master (unreleased)` section
     ```
@@ -24,21 +18,26 @@
 
     ### New Features
 
-    ### New Rules
+    ### Breaking Changes
+
+    ### New Rules (0)
+
+    -
 
     ### Bug Fixes
 
-    ### Changes
+    ### capa explorer IDA Pro plugin
 
     ### 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) and [create a tag in capa rules](https://github.com/mandiant/capa-rules/tags) upon completion.
 - [ ] [Spread the word](https://twitter.com)
+- [ ] Update internal service
 

doc/usage.md

@@ -11,3 +11,9 @@ For example, `capa -t william.ballenthin@mandiant.com` runs rules that reference
 
 ### IDA Pro plugin: capa explorer
 Please check out the [capa explorer documentation](/capa/ida/plugin/README.md).
+
+### save time by reusing .viv files
+Set the environment variable `CAPA_SAVE_WORKSPACE` to instruct the underlying analysis engine to 
+cache its intermediate results to the file system. For example, vivisect will create `.viv` files.
+Subsequently, capa may run faster when reprocessing the same input file.
+This is particularly useful during rule development as you repeatedly test a rule against a known sample.

scripts/bulk-process.py

@@ -47,7 +47,7 @@ usage:
                             parallelism factor
       --no-mp               disable subprocesses
 
-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
@@ -55,6 +55,7 @@ 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 os
 import sys
 import json
 import logging
@@ -66,7 +67,7 @@ import multiprocessing.pool
 import capa
 import capa.main
 import capa.rules
-import capa.render
+import capa.render.json
 
 logger = logging.getLogger("capa")
 
@@ -77,6 +78,7 @@ def get_capa_results(args):
 
     args is a tuple, containing:
       rules (capa.rules.RuleSet): the rules to match
+      signatures (List[str]): list of file system paths to signature files
       format (str): the name of the sample file format
       path (str): the file system path to the sample to process
 
@@ -93,10 +95,13 @@ def get_capa_results(args):
       meta (dict): the meta analysis results
       capabilities (dict): the matched capabilities and their result objects
     """
-    rules, format, path = args
+    rules, sigpaths, format, path = args
+    should_save_workspace = os.environ.get("CAPA_SAVE_WORKSPACE") not in ("0", "no", "NO", "n", None)
     logger.info("computing capa results for: %s", path)
     try:
-        extractor = capa.main.get_extractor(path, format, capa.main.BACKEND_VIV, disable_progress=True)
+        extractor = capa.main.get_extractor(
+            path, format, capa.main.BACKEND_VIV, sigpaths, should_save_workspace, disable_progress=True
+        )
     except capa.main.UnsupportedFormatError:
         # i'm 100% sure if multiprocessing will reliably raise exceptions across process boundaries.
         # so instead, return an object with explicit success/failure status.
@@ -121,9 +126,10 @@ def get_capa_results(args):
             "error": "unexpected error: %s" % (e),
         }
 
-    meta = capa.main.collect_metadata("", path, "", format, extractor)
+    meta = capa.main.collect_metadata("", path, "", extractor)
     capabilities, counts = capa.main.find_capabilities(rules, extractor, disable_progress=True)
     meta["analysis"].update(counts)
+    meta["analysis"]["layout"] = capa.main.compute_layout(rules, extractor, capabilities)
 
     return {
         "path": path,
@@ -140,7 +146,7 @@ def main(argv=None):
         argv = sys.argv[1:]
 
         parser = argparse.ArgumentParser(description="detect capabilities in programs.")
-        capa.main.install_common_args(parser, wanted={"rules"})
+        capa.main.install_common_args(parser, wanted={"rules", "signatures"})
         parser.add_argument("input", type=str, help="Path to directory of files to recursively analyze")
         parser.add_argument(
             "-n", "--parallelism", type=int, default=multiprocessing.cpu_count(), help="parallelism factor"
@@ -149,14 +155,6 @@ def main(argv=None):
         args = parser.parse_args(args=argv)
         capa.main.handle_common_args(args)
 
-        if args.rules == "(embedded rules)":
-            logger.info("using default embedded rules")
-            logger.debug("detected running from source")
-            args.rules = os.path.join(os.path.dirname(__file__), "..", "rules")
-            logger.debug("default rule path (source method): %s", args.rules)
-        else:
-            logger.info("using rules path: %s", args.rules)
-
         try:
             rules = capa.main.get_rules(args.rules)
             rules = capa.rules.RuleSet(rules)
@@ -165,6 +163,12 @@ def main(argv=None):
             logger.error("%s", str(e))
             return -1
 
+        try:
+            sig_paths = capa.main.get_signatures(args.signatures)
+        except (IOError) as e:
+            logger.error("%s", str(e))
+            return -1
+
         samples = []
         for (base, directories, files) in os.walk(args.input):
             for file in files:
@@ -196,7 +200,7 @@ def main(argv=None):
 
         results = {}
         for result in mapper(
-            get_capa_results, [(rules, "pe", sample) for sample in samples], parallelism=args.parallelism
+            get_capa_results, [(rules, sig_paths, "pe", sample) for sample in samples], parallelism=args.parallelism
         ):
             if result["status"] == "error":
                 logger.warning(result["error"])
@@ -205,7 +209,7 @@ def main(argv=None):
                 capabilities = result["ok"]["capabilities"]
                 # our renderer expects to emit a json document for a single sample
                 # so we deserialize the json document, store it in a larger dict, and we'll subsequently re-encode.
-                results[result["path"]] = json.loads(capa.render.render_json(meta, rules, capabilities))
+                results[result["path"]] = json.loads(capa.render.json.render(meta, rules, capabilities))
             else:
                 raise ValueError("unexpected status: %s" % (result["status"]))
 

scripts/capa2yara.py

@@ -0,0 +1,760 @@
+"""
+Convert capa rules to YARA rules (where this is possible)
+
+    usage: capa2yara.py [-h] [--private] [--version] [-v] [-vv] [-d] [-q] [--color {auto,always,never}] [-t TAG] rules
+
+Capa to YARA rule converter
+
+positional arguments:
+  rules                 Path to rules
+
+optional arguments:
+  -h, --help            show this help message and exit
+  --private, -p         Create private rules
+  --version             show program's version number and exit
+  -v, --verbose         enable verbose result document (no effect with --json)
+  -vv, --vverbose       enable very verbose result document (no effect with --json)
+  -d, --debug           enable debugging output on STDERR
+  -q, --quiet           disable all output but errors
+  --color {auto,always,never}
+                        enable ANSI color codes in results, default: only during interactive session
+  -t TAG, --tag TAG     filter on rule meta field values
+
+
+Copyright (C) 2020, 2021 Arnim Rupp (@ruppde) and 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 re
+import sys
+import string
+import logging
+import argparse
+import datetime
+import itertools
+
+import capa.main
+import capa.rules
+import capa.engine
+import capa.features
+import capa.features.insn
+from capa.features.common import String
+
+logger = logging.getLogger("capa2yara")
+
+today = str(datetime.date.today())
+
+# create unique variable names for each rule in case somebody wants to move/copy stuff around later
+var_names = ["".join(letters) for letters in itertools.product(string.ascii_lowercase, repeat=3)]
+
+
+# this have to be the internal names used by capa.py which are sometimes different to the ones written out in the rules, e.g. "2 or more" is "Some", count is Range
+unsupported = ["characteristic", "mnemonic", "offset", "subscope", "Range"]
+# TODO shorten this list, possible stuff:
+# - 2 or more strings: e.g.
+# -- https://github.com/mandiant/capa-rules/blob/master/collection/file-managers/gather-direct-ftp-information.yml
+# -- https://github.com/mandiant/capa-rules/blob/master/collection/browser/gather-firefox-profile-information.yml
+# - count(string    (1 rule: /executable/subfile/pe/contain-an-embedded-pe-file.yml)
+# - count(match( could be done by creating the referenced rule a 2nd time with the condition, that it hits x times (only 1 rule: ./anti-analysis/anti-disasm/contain-anti-disasm-techniques.yml)
+# - it would be technically possible to get the "basic blocks" working, but the rules contain mostly other non supported statements in there => not worth the effort.
+
+# collect all converted rules to be able to check if we have needed sub rules for match:
+converted_rules = []
+count_incomplete = 0
+
+default_tags = "CAPA "
+
+# minimum number of rounds to do be able to convert rules which depend on referenced rules in several levels of depth
+min_rounds = 5
+
+unsupported_capa_rules = open("unsupported_capa_rules.yml", "wb")
+unsupported_capa_rules_names = open("unsupported_capa_rules.txt", "wb")
+unsupported_capa_rules_list = []
+
+condition_header = """
+    capa_pe_file and
+"""
+
+condition_rule = """
+private rule capa_pe_file : CAPA {
+    meta:
+        description = "match in PE files. used by all further CAPA rules"
+        author = "Arnim Rupp"
+    condition:
+        uint16be(0) == 0x4d5a
+        or uint16be(0) == 0x558b
+        or uint16be(0) == 0x5649
+}
+"""
+
+
+def check_feature(statement, rulename):
+    if statement in unsupported:
+        logger.info("unsupported: " + statement + " in rule: " + rulename)
+        return True
+    else:
+        return False
+
+
+def get_rule_url(path):
+    path = re.sub(r"\.\.\/", "", path)
+    path = re.sub(r"capa-rules\/", "", path)
+    return "https://github.com/mandiant/capa-rules/blob/master/" + path
+
+
+def convert_capa_number_to_yara_bytes(number):
+    if not number.startswith("0x"):
+        print("TODO: fix decimal")
+        sys.exit()
+
+    number = re.sub(r"^0[xX]", "", number)
+    logger.info("number ok: " + repr(number))
+
+    # include spaces every 2 hex
+    bytesv = re.sub(r"(..)", r"\1 ", number)
+
+    # reverse order
+    bytesl = bytesv.split(" ")
+    bytesl.reverse()
+    bytesv = " ".join(bytesl)
+
+    # fix spaces
+    bytesv = bytesv[1:] + " "
+
+    return bytesv
+
+
+def convert_rule_name(rule_name):
+
+    # yara rule names: "Identifiers must follow the same lexical conventions of the C programming language, they can contain any alphanumeric character and the underscore character, but the first character cannot be a digit. Rule identifiers are case sensitive and cannot exceed 128 characters." so we replace any non-alpanum with _
+    rule_name = re.sub(r"\W", "_", rule_name)
+    rule_name = "capa_" + rule_name
+
+    return rule_name
+
+
+def convert_description(statement):
+    try:
+        desc = statement.description
+        if desc:
+            yara_desc = " // " + desc
+            logger.info("using desc: " + repr(yara_desc))
+            return yara_desc
+    except:
+        # no description
+        pass
+
+    return ""
+
+
+def convert_rule(rule, rulename, cround, depth):
+
+    depth += 1
+    logger.info("recursion depth: " + str(depth))
+
+    global var_names
+
+    def do_statement(s_type, kid):
+        yara_strings = ""
+        yara_condition = ""
+        if check_feature(s_type, rulename):
+            return "BREAK", s_type
+        elif s_type == "string":
+            string = kid.value
+            logger.info("doing string: " + repr(string))
+            string = string.replace("\\", "\\\\")
+            string = string.replace("\n", "\\n")
+            string = string.replace("\t", "\\t")
+            var_name = "str_" + var_names.pop(0)
+            yara_strings += "\t$" + var_name + ' = "' + string + '" ascii wide' + convert_description(kid) + "\n"
+            yara_condition += "\t$" + var_name + " "
+        elif s_type == "api" or s_type == "import":
+            # TODO: is it possible in YARA to make a difference between api & import?
+
+            # https://github.com/mandiant/capa-rules/blob/master/doc/format.md#api
+            api = kid.value
+            logger.info("doing api: " + repr(api))
+
+            #    e.g. kernel32.CreateNamedPipe => look for kernel32.dll and CreateNamedPipe
+            if "." in api:
+                dll, api = api.split(".")
+
+                # usage of regex is needed and /i because string search for "CreateMutex" in imports() doesn't look for e.g. CreateMutexA
+                yara_condition += "\tpe.imports(/" + dll + "/i, /" + api + "/) "
+
+            else:
+                # e.g. - api: 'CallNextHookEx'
+                # (from user32.dll)
+
+                # even looking for empty string in dll_regex doesn't work for some files (list below) with pe.imports so do just a string search
+                # yara_condition += '\tpe.imports(/.{0,30}/i, /' + api + '/) '
+                # 5fbbfeed28b258c42e0cfeb16718b31c, 2D3EDC218A90F03089CC01715A9F047F, 7EFF498DE13CC734262F87E6B3EF38AB, C91887D861D9BD4A5872249B641BC9F9, a70052c45e907820187c7e6bcdc7ecca, 0596C4EA5AA8DEF47F22C85D75AACA95
+                var_name = "api_" + var_names.pop(0)
+
+                # limit regex with word boundary \b but also search for appended A and W
+                # TODO: better use something like /(\\x00|\\x01|\\x02|\\x03|\\x04)' + api + '(A|W)?\\x00/  ???
+                yara_strings += "\t$" + var_name + " = /\\b" + api + "(A|W)?\\b/ ascii wide\n"
+                yara_condition += "\t$" + var_name + " "
+
+        elif s_type == "export":
+            export = kid.value
+            logger.info("doing export: " + repr(export))
+
+            yara_condition += '\tpe.exports("' + export + '") '
+
+        elif s_type == "section":
+            # https://github.com/mandiant/capa-rules/blob/master/doc/format.md#section
+            section = kid.value
+            logger.info("doing section: " + repr(section))
+
+            # e.g. - section: .rsrc
+            var_name_sec = var_names.pop(0)
+            # yeah, it would be better to make one loop out of multiple sections but we're in POC-land (and I guess it's not much of a performance hit, loop over short array?)
+            yara_condition += (
+                "\tfor any " + var_name_sec + " in pe.sections : ( " + var_name_sec + '.name == "' + section + '" ) '
+            )
+
+        elif s_type == "match":
+            # https://github.com/mandiant/capa-rules/blob/master/doc/format.md#matching-prior-rule-matches-and-namespaces
+            match = kid.value
+            logger.info("doing match: " + repr(match))
+
+            # e.g. - match: create process
+            #      - match: host-interaction/file-system/write
+            match_rule_name = convert_rule_name(match)
+
+            if match.startswith(rulename + "/"):
+                logger.info("Depending on myself = basic block: " + match)
+                return "BREAK", "Depending on myself = basic block"
+
+            if match_rule_name in converted_rules:
+                yara_condition += "\t" + match_rule_name + "\n"
+            else:
+                # don't complain in the early rounds as there should be 3+ rounds (if all rules are converted)
+                if cround > min_rounds - 2:
+                    logger.info("needed sub-rule not converted (yet, maybe in next round): " + repr(match))
+                    return "BREAK", "needed sub-rule not converted"
+                else:
+                    return "BREAK", "NOLOG"
+
+        elif s_type == "bytes":
+            bytesv = kid.get_value_str()
+            logger.info("doing bytes: " + repr(bytesv))
+            var_name = var_names.pop(0)
+
+            yara_strings += "\t$" + var_name + " = { " + bytesv + " }" + convert_description(kid) + "\n"
+            yara_condition += "\t$" + var_name + " "
+
+        elif s_type == "number":
+            number = kid.get_value_str()
+            logger.info("doing number: " + repr(number))
+
+            if len(number) < 10:
+                logger.info("too short for byte search (until I figure out how to do it properly)" + repr(number))
+                return "BREAK", "Number too short"
+
+            # there's just one rule which contains 0xFFFFFFF but yara gives a warning if if used
+            if number == "0xFFFFFFFF":
+                return "BREAK", "slow byte pattern for YARA search"
+
+            logger.info("number ok: " + repr(number))
+            number = convert_capa_number_to_yara_bytes(number)
+            logger.info("number ok: " + repr(number))
+
+            var_name = "num_" + var_names.pop(0)
+            yara_strings += "\t$" + var_name + " = { " + number + "}" + convert_description(kid) + "\n"
+            yara_condition += "$" + var_name + " "
+
+        elif s_type == "regex":
+            regex = kid.get_value_str()
+            logger.info("doing regex: " + repr(regex))
+
+            # change capas /xxx/i to yaras /xxx/ nocase, count will be used later to decide appending 'nocase'
+            regex, count = re.subn(r"/i$", "/", regex)
+
+            # remove / in the begining and end
+            regex = regex[1:-1]
+
+            # all .* in the regexes of capa look like they should be maximum 100 chars so take 1000 to speed up rules and prevent yara warnings on poor performance
+            regex = regex.replace(".*", ".{,1000}")
+            # strange: capa accepts regexes with unsescaped / like - string: /com/exe4j/runtime/exe4jcontroller/i in capa-rules/compiler/exe4j/compiled-with-exe4j.yml, needs a fix for yara:
+            # would assume that get_value_str() gives the raw string
+            regex = re.sub(r"(?<!\\)/", r"\/", regex)
+
+            # capa uses python regex which accepts /reg(|.exe)/ but yaras regex engine doesn't not => fix it
+            # /reg(|.exe)/ => /reg(.exe)?/
+            regex = re.sub(r"\(\|([^\)]+)\)", r"(\1)?", regex)
+
+            # change begining of line to null byte, e.g. /^open => /\x00open (not word boundary because we're not looking for the begining of a word in a text but usually a function name if there's ^ in a capa rule)
+            regex = re.sub(r"^\^", r"\\x00", regex)
+
+            # regex = re.sub(r"^\^", r"\\b", regex)
+
+            regex = "/" + regex + "/"
+            if count:
+                regex += " nocase"
+
+            # strange: if statement.name == "string", the string is as it is, if statement.name == "regex", the string has // around it, e.g. /regex/
+            var_name = "re_" + var_names.pop(0)
+            yara_strings += "\t" + "$" + var_name + " = " + regex + " ascii wide " + convert_description(kid) + "\n"
+            yara_condition += "\t" + "$" + var_name + " "
+        elif s_type == "Not" or s_type == "And" or s_type == "Or":
+            pass
+        else:
+            logger.info("something unhandled: " + repr(s_type))
+            sys.exit()
+
+        return yara_strings, yara_condition
+
+    # end: def do_statement
+
+    yara_strings_list = []
+    yara_condition_list = []
+    rule_comment = ""
+    incomplete = 0
+
+    statement = rule.name
+
+    logger.info("doing statement: " + statement)
+
+    if check_feature(statement, rulename):
+        return "BREAK", statement, rule_comment, incomplete
+
+    if statement == "And" or statement == "Or":
+        desc = convert_description(rule)
+        if desc:
+            logger.info("description of bool statement: " + repr(desc))
+            yara_strings_list.append("\t" * depth + desc + "\n")
+    elif statement == "Not":
+        logger.info("one of those seldom nots: " + rule.name)
+
+    # check for nested statements
+    try:
+        kids = rule.children
+        num_kids = len(kids)
+        logger.info("kids: " + kids)
+    except:
+        logger.info("no kids in rule: " + rule.name)
+
+        try:
+            # maybe it's "Not" = only one child:
+            kid = rule.child
+            kids = [kid]
+            num_kids = 1
+            logger.info("kid: %s", kids)
+        except:
+            logger.info("no kid in rule: %s", rule.name)
+
+    # just a single statement without 'and' or 'or' before it in this rule
+    if "kids" not in locals().keys():
+        logger.info("no kids: " + rule.name)
+
+        yara_strings_sub, yara_condition_sub = do_statement(statement, rule)
+
+        if yara_strings_sub == "BREAK":
+            logger.info("Unknown feature at1: " + rule.name)
+            return "BREAK", yara_condition_sub, rule_comment, incomplete
+        yara_strings_list.append(yara_strings_sub)
+        yara_condition_list.append(yara_condition_sub)
+
+    else:
+        x = 0
+        logger.info("doing kids: %r - len: %s", kids, num_kids)
+        for kid in kids:
+            s_type = kid.name
+            logger.info("doing type: " + s_type + " kidnum: " + str(x))
+
+            if s_type == "Some":
+                cmin = kid.count
+                logger.info("Some type with mininum: " + str(cmin))
+
+                if not cmin:
+                    logger.info("this is optional: which means, we can just ignore it")
+                    x += 1
+                    continue
+                elif statement == "Or":
+                    logger.info("we're inside an OR, we can just ignore it")
+                    x += 1
+                    continue
+                else:
+                    # this is "x or more". could be coded for strings TODO
+                    return "BREAK", "Some aka x or more (TODO)", rule_comment, incomplete
+
+            if s_type == "And" or s_type == "Or" or s_type == "Not" and not kid.name == "Some":
+                logger.info("doing bool with recursion: " + repr(kid))
+                logger.info("kid coming: " + repr(kid.name))
+                # logger.info("grandchildren: " + repr(kid.children))
+
+                #
+                # here we go into RECURSION
+                #
+                yara_strings_sub, yara_condition_sub, rule_comment_sub, incomplete_sub = convert_rule(
+                    kid, rulename, cround, depth
+                )
+
+                logger.info("coming out of this recursion, depth: " + repr(depth) + " s_type: " + s_type)
+
+                if yara_strings_sub == "BREAK":
+                    logger.info(
+                        "Unknown feature at2: " + rule.name + " - s_type: " + s_type + " - depth: " + str(depth)
+                    )
+
+                    # luckily this is only a killer, if we're inside an 'And', inside 'Or' we're just missing some coverage
+                    # only accept incomplete rules in rounds > 3 because the reason might be a reference to another rule not converted yet because of missing dependencies
+                    logger.info("rule.name,  depth,  cround: " + rule.name + ", " + str(depth) + ", " + str(cround))
+                    if rule.name == "Or" and depth == 1 and cround > min_rounds - 1:
+                        logger.info(
+                            "Unknown feature, just ignore this branch and keep the rest bec we're in Or (1): "
+                            + s_type
+                            + " - depth: "
+                            + str(depth)
+                        )
+                        # remove last 'or'
+                        # yara_condition = re.sub(r'\sor $', ' ', yara_condition)
+                        rule_comment += "This rule is incomplete because a branch inside an Or-statement had an unsupported feature and was skipped => coverage is reduced compared to the original capa rule. "
+                        x += 1
+                        incomplete = 1
+                        continue
+                    else:
+                        return "BREAK", yara_condition_sub, rule_comment, incomplete
+
+                rule_comment += rule_comment_sub
+                yara_strings_list.append(yara_strings_sub)
+                yara_condition_list.append(yara_condition_sub)
+
+                incomplete = incomplete or incomplete_sub
+
+            yara_strings_sub, yara_condition_sub = do_statement(s_type, kid)
+
+            if yara_strings_sub == "BREAK":
+                logger.info("Unknown feature at3: " + rule.name)
+                logger.info("rule.name,  depth,  cround: " + rule.name + ", " + str(depth) + ", " + str(cround))
+                if rule.name == "Or" and depth == 1 and cround > min_rounds - 1:
+                    logger.info(
+                        "Unknown feature, just ignore this branch and keep the rest bec we're in Or (2): "
+                        + s_type
+                        + " - depth: "
+                        + str(depth)
+                    )
+
+                    rule_comment += "This rule is incomplete because a branch inside an Or-statement had an unsupported feature and was skipped => coverage is reduced compared to the original capa rule. "
+                    x += 1
+                    incomplete = 1
+                    continue
+                else:
+                    return "BREAK", yara_condition_sub, rule_comment, incomplete
+
+            # don't append And or Or if we got no condition back from this kid from e.g. match in myself or unsupported feature inside 'Or'
+            if not yara_condition_sub:
+                continue
+
+            yara_strings_list.append(yara_strings_sub)
+            yara_condition_list.append(yara_condition_sub)
+            x += 1
+
+    # this might happen, if all conditions are inside "or" and none of them was supported
+    if not yara_condition_list:
+        return (
+            "BREAK",
+            'Multiple statements inside "- or:" where all unsupported, the last one was "' + s_type + '"',
+            rule_comment,
+            incomplete,
+        )
+
+    if statement == "And" or statement == "Or":
+        if yara_strings_list:
+            yara_strings = "".join(yara_strings_list)
+        else:
+            yara_strings = ""
+
+        yara_condition = " (\n\t\t" + ("\n\t\t" + statement.lower() + " ").join(yara_condition_list) + " \n\t) "
+
+    elif statement == "Some":
+        cmin = rule.count
+        logger.info("Some type with mininum at2: " + str(cmin))
+
+        if not cmin:
+            logger.info("this is optional: which means, we can just ignore it")
+        else:
+            # this is "x or more". could be coded for strings TODO
+            return "BREAK", "Some aka x or more (TODO)", rule_comment, incomplete
+    elif statement == "Not":
+        logger.info("Not")
+        yara_strings = "".join(yara_strings_list)
+        yara_condition = "not " + "".join(yara_condition_list) + " "
+    else:
+        if len(yara_condition_list) != 1:
+            logger.info("something wrong around here" + repr(yara_condition_list) + " - " + statement)
+            sys.exit()
+
+        # strings might be empty with only conditions
+        if yara_strings_list:
+            yara_strings = "\n\t" + yara_strings_list[0]
+
+        yara_condition = "\n\t" + yara_condition_list[0]
+
+    logger.info(f"# end of convert_rule() #strings: {len(yara_strings_list)} #conditions: {len(yara_condition_list)}")
+    logger.info(f"strings: {yara_strings} conditions: {yara_condition}")
+
+    return yara_strings, yara_condition, rule_comment, incomplete
+
+
+def output_yar(yara):
+    print(yara + "\n")
+
+
+def output_unsupported_capa_rules(yaml, capa_rulename, url, reason):
+
+    if reason != "NOLOG":
+        if capa_rulename not in unsupported_capa_rules_list:
+            logger.info("unsupported: " + capa_rulename + " - reason: " + reason + " - url: " + url)
+
+            unsupported_capa_rules_list.append(capa_rulename)
+            unsupported_capa_rules.write(yaml.encode("utf-8") + b"\n")
+            unsupported_capa_rules.write(
+                (
+                    "Reason: "
+                    + reason
+                    + " (there might be multiple unsupported things in this rule, this is the 1st one encountered)"
+                ).encode("utf-8")
+                + b"\n"
+            )
+            unsupported_capa_rules.write(url.encode("utf-8") + b"\n----------------------------------------------\n")
+            unsupported_capa_rules_names.write(capa_rulename.encode("utf-8") + b":")
+            unsupported_capa_rules_names.write(reason.encode("utf-8") + b":")
+            unsupported_capa_rules_names.write(url.encode("utf-8") + b"\n")
+
+
+def convert_rules(rules, namespaces, cround):
+    for rule in rules.rules.values():
+
+        rule_name = convert_rule_name(rule.name)
+
+        if rule.meta.get("capa/subscope-rule", False):
+            logger.info("skipping sub scope rule capa: " + rule.name)
+            continue
+
+        if rule_name in converted_rules:
+            logger.info("skipping already converted rule capa: " + rule.name + " - yara rule: " + rule_name)
+            continue
+
+        logger.info("-------------------------- DOING RULE CAPA: " + rule.name + " - yara rule: " + rule_name)
+        if "capa/path" in rule.meta:
+            url = get_rule_url(rule.meta["capa/path"])
+        else:
+            url = "no url"
+
+        logger.info("URL: " + url)
+        logger.info("statements: " + repr(rule.statement))
+
+        # don't really know what that passed empty string is good for :)
+        dependencies = rule.get_dependencies(namespaces)
+
+        if len(dependencies):
+            logger.info("Dependencies at4: " + rule.name + " - dep: " + str(dependencies))
+
+            for dep in dependencies:
+                logger.info("Dependencies at44: " + dep)
+                if not dep.startswith(rule.name + "/"):
+                    logger.info("Depending on another rule: " + dep)
+                    continue
+
+        yara_strings, yara_condition, rule_comment, incomplete = convert_rule(rule.statement, rule.name, cround, 0)
+
+        if yara_strings == "BREAK":
+            # only give up if in final extra round #9000
+            if cround == 9000:
+                output_unsupported_capa_rules(rule.to_yaml(), rule.name, url, yara_condition)
+            logger.info("Unknown feature at5: " + rule.name)
+        else:
+
+            yara_meta = ""
+            metas = rule.meta
+            rule_tags = ""
+
+            for meta in metas:
+                meta_name = meta
+                # e.g. 'examples:' can be a list
+                seen_hashes = []
+                if isinstance(metas[meta], list):
+                    if meta_name == "examples":
+                        meta_name = "hash"
+                    if meta_name == "att&ck":
+                        meta_name = "attack"
+                        for attack in list(metas[meta]):
+                            logger.info("attack:" + attack)
+                            # cut out tag in square brackets, e.g. Defense Evasion::Obfuscated Files or Information [T1027] => T1027
+                            r = re.search(r"\[(T[^\]]*)", attack)
+                            if r:
+                                tag = r.group(1)
+                                logger.info("attack tag:" + tag)
+                                tag = re.sub(r"\W", "_", tag)
+                                rule_tags += tag + " "
+                                # also add a line "attack = ..." to yaras 'meta:' to keep the long description:
+                                yara_meta += '\tattack = "' + attack + '"\n'
+                    elif meta_name == "mbc":
+                        for mbc in list(metas[meta]):
+                            logger.info("mbc:" + mbc)
+                            # cut out tag in square brackets, e.g. Cryptography::Encrypt Data::RC6 [C0027.010] => C0027.010
+                            r = re.search(r"\[(.[^\]]*)", mbc)
+                            if r:
+                                tag = r.group(1)
+                                logger.info("mbc tag:" + tag)
+                                tag = re.sub(r"\W", "_", tag)
+                                rule_tags += tag + " "
+
+                                # also add a line "mbc = ..." to yaras 'meta:' to keep the long description:
+                                yara_meta += '\tmbc = "' + mbc + '"\n'
+
+                    for value in metas[meta]:
+                        if meta_name == "hash":
+                            value = re.sub(r"^([0-9a-f]{20,64}):0x[0-9a-f]{1,10}$", r"\1", value, flags=re.IGNORECASE)
+
+                            # examples in capa can contain the same hash several times with different offset, so check if it's already there:
+                            # (keeping the offset might be interessting for some but breaks yara-ci for checking of the final rules
+                            if value not in seen_hashes:
+                                yara_meta += "\t" + meta_name + ' = "' + value + '"\n'
+                                seen_hashes.append(value)
+
+                else:
+                    # no list:
+                    if meta == "capa/path":
+                        url = get_rule_url(metas[meta])
+                        meta_name = "reference"
+                        meta_value = "This YARA rule converted from capa rule: " + url
+                    else:
+                        meta_value = metas[meta]
+
+                    if meta_name == "name":
+                        meta_name = "description"
+                        meta_value += " (converted from capa rule)"
+                    elif meta_name == "lib":
+                        meta_value = str(meta_value)
+                    elif meta_name == "capa/nursery":
+                        meta_name = "capa_nursery"
+                        meta_value = str(meta_value)
+
+                    # for the rest of the maec/malware-category names:
+                    meta_name = re.sub(r"\W", "_", meta_name)
+
+                    if meta_name and meta_value:
+                        yara_meta += "\t" + meta_name + ' = "' + meta_value + '"\n'
+
+            rule_name_bonus = ""
+            if rule_comment:
+                yara_meta += '\tcomment = "' + rule_comment + '"\n'
+            yara_meta += '\tdate = "' + today + '"\n'
+            yara_meta += '\tminimum_yara = "3.8"\n'
+            yara_meta += '\tlicense = "Apache-2.0 License"\n'
+
+            # check if there's some beef in condition:
+            tmp_yc = re.sub(r"(and|or|not)", "", yara_condition)
+            if re.search(r"\w", tmp_yc):
+
+                yara = ""
+                if make_priv:
+                    yara = "private "
+
+                # put yara rule tags here:
+                rule_tags = default_tags + rule_tags
+                yara += "rule " + rule_name + " : " + rule_tags + " { \n  meta: \n " + yara_meta + "\n"
+
+                if "$" in yara_strings:
+                    yara += "  strings: \n " + yara_strings + " \n"
+
+                yara += "  condition:" + condition_header + yara_condition + "\n}"
+
+                # TODO: now the rule is finished and could be automatically checked with the capa-testfile(s) named in meta (doing it for all of them using yara-ci upload at the moment)
+                output_yar(yara)
+                converted_rules.append(rule_name)
+                global count_incomplete
+                count_incomplete += incomplete
+            else:
+                output_unsupported_capa_rules(rule.to_yaml(), rule.name, url, yara_condition)
+                pass
+
+
+def main(argv=None):
+    if argv is None:
+        argv = sys.argv[1:]
+
+    parser = argparse.ArgumentParser(description="Capa to YARA rule converter")
+    parser.add_argument("rules", type=str, help="Path to rules")
+    parser.add_argument("--private", "-p", action="store_true", help="Create private rules", default=False)
+    capa.main.install_common_args(parser, wanted={"tag"})
+
+    args = parser.parse_args(args=argv)
+    global make_priv
+    make_priv = args.private
+
+    if args.verbose:
+        level = logging.DEBUG
+    elif args.quiet:
+        level = logging.ERROR
+    else:
+        level = logging.INFO
+
+    logging.basicConfig(level=level)
+    logging.getLogger("capa2yara").setLevel(level)
+
+    try:
+        rules = capa.main.get_rules([args.rules], disable_progress=True)
+        namespaces = capa.rules.index_rules_by_namespace(list(rules))
+        rules = capa.rules.RuleSet(rules)
+        logger.info("successfully loaded %s rules (including subscope rules which will be ignored)", len(rules))
+        if args.tag:
+            rules = rules.filter_rules_by_meta(args.tag)
+            logger.debug("selected %s rules", len(rules))
+            for i, r in enumerate(rules.rules, 1):
+                logger.debug(" %d. %s", i, r)
+    except (IOError, capa.rules.InvalidRule, capa.rules.InvalidRuleSet) as e:
+        logger.error("%s", str(e))
+        return -1
+
+    output_yar(
+        "// Rules from Mandiant's https://github.com/mandiant/capa-rules converted to YARA using https://github.com/mandiant/capa/blob/master/scripts/capa2yara.py by Arnim Rupp"
+    )
+    output_yar(
+        "// Beware: These are less rules than capa (because not all fit into YARA, stats at EOF) and is less precise because e.g. capas function scopes are applied to the whole file"
+    )
+    output_yar(
+        '// Beware: Some rules are incomplete because an optional branch was not supported by YARA. These rules are marked in a comment in meta: (search for "incomplete")'
+    )
+    output_yar("// Rule authors and license stay the same")
+    output_yar(
+        '// att&ck and MBC tags are put into YARA rule tags. All rules are tagged with "CAPA" for easy filtering'
+    )
+    output_yar("// The date = in meta: is the date of converting (there is no date in capa rules)")
+    output_yar("// Minimum YARA version is 3.8.0 plus PE module")
+    output_yar('\nimport "pe"')
+
+    output_yar(condition_rule)
+
+    # do several rounds of converting rules because some rules for match: might not be converted in the 1st run
+    num_rules = 9999999
+    cround = 0
+    while num_rules != len(converted_rules) or cround < min_rounds:
+        cround += 1
+        logger.info("doing convert_rules(), round: " + str(cround))
+        num_rules = len(converted_rules)
+        convert_rules(rules, namespaces, cround)
+
+    # one last round to collect all unconverted rules
+    convert_rules(rules, namespaces, 9000)
+
+    stats = "\n// converted rules              : " + str(len(converted_rules))
+    stats += "\n//   among those are incomplete : " + str(count_incomplete)
+    stats += "\n// unconverted rules            : " + str(len(unsupported_capa_rules_list)) + "\n"
+    logger.info(stats)
+    output_yar(stats)
+
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

scripts/capa_as_library.py

@@ -6,18 +6,18 @@ import collections
 import capa.main
 import capa.rules
 import capa.engine
-import capa.render
 import capa.features
+import capa.render.json
 import capa.render.utils as rutils
+import capa.render.default
+import capa.render.result_document
 from capa.engine import *
-from capa.render import convert_capabilities_to_result_document
 
 # edit this to set the path for file to analyze and rule directory
 RULES_PATH = "/tmp/capa/rules/"
 
 # load rules from disk
-rules = capa.main.get_rules(RULES_PATH, disable_progress=True)
-rules = capa.rules.RuleSet(rules)
+rules = capa.rules.RuleSet(capa.main.get_rules([RULES_PATH], disable_progress=True))
 
 # == Render ddictionary helpers
 def render_meta(doc, ostream):
@@ -104,28 +104,16 @@ def render_attack(doc, ostream):
     for rule in rutils.capability_rules(doc):
         if not rule["meta"].get("att&ck"):
             continue
-
         for attack in rule["meta"]["att&ck"]:
-            tactic, _, rest = attack.partition("::")
-            if "::" in rest:
-                technique, _, rest = rest.partition("::")
-                subtechnique, _, id = rest.rpartition(" ")
-                tactics[tactic].add((technique, subtechnique, id))
-            else:
-                technique, _, id = rest.rpartition(" ")
-                tactics[tactic].add((technique, id))
+            tactics[attack["tactic"]].add((attack["technique"], attack.get("subtechnique"), attack["id"]))
 
     for tactic, techniques in sorted(tactics.items()):
         inner_rows = []
-        for spec in sorted(techniques):
-            if len(spec) == 2:
-                technique, id = spec
+        for (technique, subtechnique, id) in sorted(techniques):
+            if subtechnique is None:
                 inner_rows.append("%s %s" % (technique, id))
-            elif len(spec) == 3:
-                technique, subtechnique, id = spec
-                inner_rows.append("%s::%s %s" % (technique, subtechnique, id))
             else:
-                raise RuntimeError("unexpected ATT&CK spec format")
+                inner_rows.append("%s::%s %s" % (technique, subtechnique, id))
         ostream["ATTCK"].setdefault(tactic.upper(), inner_rows)
 
 
@@ -150,31 +138,16 @@ def render_mbc(doc, ostream):
         if not rule["meta"].get("mbc"):
             continue
 
-        mbcs = rule["meta"]["mbc"]
-        if not isinstance(mbcs, list):
-            raise ValueError("invalid rule: MBC mapping is not a list")
-
-        for mbc in mbcs:
-            objective, _, rest = mbc.partition("::")
-            if "::" in rest:
-                behavior, _, rest = rest.partition("::")
-                method, _, id = rest.rpartition(" ")
-                objectives[objective].add((behavior, method, id))
-            else:
-                behavior, _, id = rest.rpartition(" ")
-                objectives[objective].add((behavior, id))
+        for mbc in rule["meta"]["mbc"]:
+            objectives[mbc["objective"]].add((mbc["behavior"], mbc.get("method"), mbc["id"]))
 
     for objective, behaviors in sorted(objectives.items()):
         inner_rows = []
-        for spec in sorted(behaviors):
-            if len(spec) == 2:
-                behavior, id = spec
-                inner_rows.append("%s %s" % (behavior, id))
-            elif len(spec) == 3:
-                behavior, method, id = spec
-                inner_rows.append("%s::%s %s" % (behavior, method, id))
+        for (behavior, method, id) in sorted(behaviors):
+            if method is None:
+                inner_rows.append("%s [%s]" % (behavior, id))
             else:
-                raise RuntimeError("unexpected MBC spec format")
+                inner_rows.append("%s::%s [%s]" % (behavior, method, id))
         ostream["MBC"].setdefault(objective.upper(), inner_rows)
 
 
@@ -190,26 +163,26 @@ def render_dictionary(doc):
 
 # ==== render dictionary helpers
 def capa_details(file_path, output_format="dictionary"):
-
     # extract features and find capabilities
-    extractor = capa.main.get_extractor(file_path, "auto", capa.main.BACKEND_VIV, disable_progress=True)
+    extractor = capa.main.get_extractor(file_path, "auto", capa.main.BACKEND_VIV, [], False, disable_progress=True)
     capabilities, counts = capa.main.find_capabilities(rules, extractor, disable_progress=True)
 
     # collect metadata (used only to make rendering more complete)
-    meta = capa.main.collect_metadata("", file_path, RULES_PATH, "auto", extractor)
+    meta = capa.main.collect_metadata("", file_path, RULES_PATH, extractor)
     meta["analysis"].update(counts)
+    meta["analysis"]["layout"] = capa.main.compute_layout(rules, extractor, capabilities)
 
     capa_output = False
     if output_format == "dictionary":
         # ...as python dictionary, simplified as textable but in dictionary
-        doc = convert_capabilities_to_result_document(meta, rules, capabilities)
+        doc = capa.render.result_document.convert_capabilities_to_result_document(meta, rules, capabilities)
         capa_output = render_dictionary(doc)
     elif output_format == "json":
         # render results
         # ...as json
-        capa_output = json.loads(capa.render.render_json(meta, rules, capabilities))
+        capa_output = json.loads(capa.render.json.render(meta, rules, capabilities))
     elif output_format == "texttable":
         # ...as human readable text table
-        capa_output = capa.render.render_default(meta, rules, capabilities)
+        capa_output = capa.render.default.render(meta, rules, capabilities)
 
     return capa_output