add pseudo layout with function chunks size and basic stats table

Fix IDA 9.0 / idalib

.github/CONTRIBUTING.md

@@ -57,7 +57,7 @@ When we make a significant decision in how we maintain the project and what we c
  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.
+If it is *not* documented there, or you can't find an answer, please open an issue.
 We'll link to existing issues when appropriate to keep discussions in one place.
 
 ## How Can I Contribute?

.github/dependabot.yml

@@ -4,3 +4,6 @@ updates:
     directory: "/"
     schedule:
       interval: "weekly"
+    ignore:
+      - dependency-name: "*"
+        update-types: ["version-update:semver-patch"]

.github/flake8.ini

@@ -40,4 +40,4 @@ per-file-ignores =
 
 copyright-check = True
 copyright-min-file-size = 1 
-copyright-regexp = Copyright \(C\) 2023 Mandiant, Inc. All Rights Reserved.
+copyright-regexp = Copyright \(C\) \d{4} Mandiant, Inc. All Rights Reserved.

.github/mypy/mypy.ini

@@ -1,11 +1,5 @@
 [mypy]
 
-[mypy-halo.*]
-ignore_missing_imports = True
-
-[mypy-tqdm.*]
-ignore_missing_imports = True
-
 [mypy-ruamel.*]
 ignore_missing_imports = True
 

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

@@ -24,7 +24,7 @@ excludedimports = [
     "pyqtwebengine",
     # the above are imported by these viv modules.
     # so really, we'd want to exclude these submodules of viv.
-    # but i dont think this works.
+    # but i don't think this works.
     "vqt",
     "vdb.qt",
     "envi.qt",

.github/pyinstaller/pyinstaller.spec

@@ -1,10 +1,18 @@
 # -*- mode: python -*-
 # Copyright (C) 2020 Mandiant, Inc. All Rights Reserved.
-import os.path
-import subprocess
+import sys
 
-import wcwidth
+import capa.rules.cache
 
+from pathlib import Path
+
+# SPECPATH is a global variable which points to .spec file path
+capa_dir = Path(SPECPATH).parent.parent
+rules_dir = capa_dir / 'rules'
+cache_dir = capa_dir / 'cache'
+
+if not capa.rules.cache.generate_rule_cache(rules_dir, cache_dir):
+    sys.exit(-1)
 
 a = Analysis(
     # when invoking pyinstaller from the project root,
@@ -20,13 +28,6 @@ a = Analysis(
         ("../../rules", "rules"),
         ("../../sigs", "sigs"),
         ("../../cache", "cache"),
-        # capa.render.default uses tabulate that depends on wcwidth.
-        # it seems wcwidth uses a json file `version.json`
-        # and this doesn't get picked up by pyinstaller automatically.
-        # so we manually embed the wcwidth resources here.
-        #
-        # ref: https://stackoverflow.com/a/62278462/87207
-        (os.path.dirname(wcwidth.__file__), "wcwidth"),
     ],
     # when invoking pyinstaller from the project root,
     # this gets run from the project root.
@@ -39,11 +40,6 @@ a = Analysis(
         "tkinter",
         "_tkinter",
         "Tkinter",
-        # tqdm provides renderers for ipython,
-        # however, this drags in a lot of dependencies.
-        # since we don't spawn a notebook, we can safely remove these.
-        "IPython",
-        "ipywidgets",
         # these are pulled in by networkx
         # but we don't need to compute the strongly connected components.
         "numpy",
@@ -61,7 +57,10 @@ a = Analysis(
         "qt5",
         "pyqtwebengine",
         "pyasn1",
+        # don't pull in Binary Ninja/IDA bindings that should
+        # only be installed locally.
         "binaryninja",
+        "ida",
     ],
 )
 

.github/ruff.toml

@@ -1,16 +1,16 @@
 # Enable the pycodestyle (`E`) and Pyflakes (`F`) rules by default.
 # Unlike Flake8, Ruff doesn't enable pycodestyle warnings (`W`) or
 # McCabe complexity (`C901`) by default.
-select = ["E", "F"]
+lint.select = ["E", "F"]
 
 # Allow autofix for all enabled rules (when `--fix`) is provided.
-fixable = ["ALL"]
-unfixable = []
+lint.fixable = ["ALL"]
+lint.unfixable = []
 
 # E402 module level import not at top of file
 # E722 do not use bare 'except'
 # E501 line too long
-ignore = ["E402", "E722", "E501"]
+lint.ignore = ["E402", "E722", "E501"]
 
 line-length = 120
 

.github/workflows/build.yml

@@ -3,6 +3,10 @@ name: build
 on:
   pull_request:
     branches: [ master ]
+    paths-ignore:
+      - 'web/**'
+      - 'doc/**'
+      - '**.md'
   release:
     types: [edited, published]
 
@@ -26,24 +30,24 @@ jobs:
             python_version: 3.8
           - os: ubuntu-20.04
             artifact_name: capa
-            asset_name: linux-py311
-            python_version: 3.11
+            asset_name: linux-py312
+            python_version: 3.12
           - os: windows-2019
             artifact_name: capa.exe
             asset_name: windows
             python_version: 3.8
-          - os: macos-11
+          - os: macos-12
             # use older macOS for assumed better portability
             artifact_name: capa
             asset_name: macos
             python_version: 3.8
     steps:
       - name: Checkout capa
-        uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
+        uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
         with:
           submodules: true
       - name: Set up Python ${{ matrix.python_version }}
-        uses: actions/setup-python@d27e3f3d7c64b4bbf8e4abfb9b63b83e846e0435 # v4.5.0
+        uses: actions/setup-python@0a5c61591373683505ea898e09a3ea4f39ef2b9c # v5.0.0
         with:
           python-version: ${{ matrix.python_version }}
       - if: matrix.os == 'ubuntu-20.04'
@@ -51,9 +55,9 @@ jobs:
       - name: Upgrade pip, setuptools
         run: python -m pip install --upgrade pip setuptools
       - name: Install capa with build requirements
-        run: pip install -e .[build]
-      - name: Cache the rule set
-        run: python ./scripts/cache-ruleset.py ./rules/ ./cache/
+        run: |
+          pip install -r requirements.txt
+          pip install -e .[build]
       - name: Build standalone executable
         run: pyinstaller --log-level DEBUG .github/pyinstaller/pyinstaller.spec
       - name: Does it run (PE)?
@@ -66,7 +70,7 @@ jobs:
         run: |
           7z e "tests/data/dynamic/cape/v2.2/d46900384c78863420fb3e297d0a2f743cd2b6b3f7f82bf64059a168e07aceb7.json.gz"
           dist/capa -d "d46900384c78863420fb3e297d0a2f743cd2b6b3f7f82bf64059a168e07aceb7.json"
-      - uses: actions/upload-artifact@0b7f8abb1508181956e8e162db84b466c27e18ce # v3.1.2
+      - uses: actions/upload-artifact@5d5d22a31266ced268874388b861e4b58bb5c2f3 # v4.3.1
         with:
           name: ${{ matrix.asset_name }}
           path: dist/${{ matrix.artifact_name }}
@@ -84,13 +88,13 @@ jobs:
             asset_name: linux
           - os: ubuntu-22.04
             artifact_name: capa
-            asset_name: linux-py311
+            asset_name: linux-py312
           - os: windows-2022
             artifact_name: capa.exe
             asset_name: windows
     steps:
       - name: Download ${{ matrix.asset_name }}
-        uses: actions/download-artifact@9bc31d5ccc31df68ecc42ccf4149144866c47d8a # v3.0.2
+        uses: actions/download-artifact@eaceaf801fd36c7dee90939fad912460b18a1ffe # v4.1.2
         with:
           name: ${{ matrix.asset_name }}
       - name: Set executable flag
@@ -110,7 +114,7 @@ jobs:
         include:
           - asset_name: linux
             artifact_name: capa
-          - asset_name: linux-py311
+          - asset_name: linux-py312
             artifact_name: capa
           - asset_name: windows
             artifact_name: capa.exe
@@ -118,7 +122,7 @@ jobs:
             artifact_name: capa
     steps:
       - name: Download ${{ matrix.asset_name }}
-        uses: actions/download-artifact@9bc31d5ccc31df68ecc42ccf4149144866c47d8a # v3.0.2
+        uses: actions/download-artifact@eaceaf801fd36c7dee90939fad912460b18a1ffe # v4.1.2
         with:
           name: ${{ matrix.asset_name }}
       - name: Set executable flag

.github/workflows/changelog.yml

@@ -7,7 +7,8 @@ on:
   pull_request_target:
     types: [opened, edited, synchronize]
 
-permissions: read-all
+permissions:
+  pull-requests: write
 
 jobs:
   check_changelog:
@@ -19,7 +20,7 @@ jobs:
     steps:
     - name: Get changed files
       id: files
-      uses: Ana06/get-changed-files@e0c398b7065a8d84700c471b6afc4116d1ba4e96 # v2.2.0
+      uses: Ana06/get-changed-files@25f79e676e7ea1868813e21465014798211fad8c # v2.3.0
     - name: check changelog updated
       id: changelog_updated
       env:
@@ -29,14 +30,14 @@ jobs:
         echo $FILES | grep -qF 'CHANGELOG.md' || echo $PR_BODY | grep -qiF "$NO_CHANGELOG"
     - name: Reject pull request if no CHANGELOG update
       if: ${{ always() && steps.changelog_updated.outcome == 'failure' }}
-      uses: Ana06/automatic-pull-request-review@0cf4e8a17ba79344ed3fdd7fed6dd0311d08a9d4 # v0.1.0
+      uses: Ana06/automatic-pull-request-review@76aaf9b15b116a54e1da7a28a46f91fe089600bf # v0.2.0
       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@0cf4e8a17ba79344ed3fdd7fed6dd0311d08a9d4 # v0.1.0
+      uses: Ana06/automatic-pull-request-review@76aaf9b15b116a54e1da7a28a46f91fe089600bf # v0.2.0
       with:
         repo-token: ${{ secrets.GITHUB_TOKEN }}
         event: DISMISS

.github/workflows/publish.yml

@@ -17,20 +17,21 @@ jobs:
     permissions:
       id-token: write
     steps:
-      - uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
+      - uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
       - name: Set up Python
-        uses: actions/setup-python@d27e3f3d7c64b4bbf8e4abfb9b63b83e846e0435 # v4.5.0
+        uses: actions/setup-python@0a5c61591373683505ea898e09a3ea4f39ef2b9c # v5.0.0
         with:
           python-version: '3.8'
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip
+          pip install -r requirements.txt
           pip install -e .[build]
       - name: build package
         run: |
           python -m build
       - name: upload package artifacts
-        uses: actions/upload-artifact@0b7f8abb1508181956e8e162db84b466c27e18ce # v3.1.2
+        uses: actions/upload-artifact@5d5d22a31266ced268874388b861e4b58bb5c2f3 # v4.3.1
         with:
           path: dist/*
       - name: publish package

.github/workflows/scorecard.yml

@@ -32,12 +32,12 @@ jobs:
 
     steps:
       - name: "Checkout code"
-        uses: actions/checkout@93ea575cb5d8a053eaa0ac8fa3b40d7e05a33cc8 # v3.1.0
+        uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11  # v4.1.1
         with:
           persist-credentials: false
 
       - name: "Run analysis"
-        uses: ossf/scorecard-action@99c53751e09b9529366343771cc321ec74e9bd3d # v2.0.6
+        uses: ossf/scorecard-action@0864cf19026789058feabb7e87baa5f140aac736 # v2.3.1
         with:
           results_file: results.sarif
           results_format: sarif
@@ -59,7 +59,7 @@ jobs:
       # Upload the results as artifacts (optional). Commenting out will disable uploads of run results in SARIF
       # format to the repository Actions tab.
       - name: "Upload artifact"
-        uses: actions/upload-artifact@3cea5372237819ed00197afe530f5a7ea3e805c8 # v3.1.0
+        uses: actions/upload-artifact@5d5d22a31266ced268874388b861e4b58bb5c2f3 # v4.3.1
         with:
           name: SARIF file
           path: results.sarif
@@ -67,6 +67,6 @@ jobs:
 
       # Upload the results to GitHub's code scanning dashboard.
       - name: "Upload to code-scanning"
-        uses: github/codeql-action/upload-sarif@807578363a7869ca324a79039e6db9c843e0e100 # v2.1.27
+        uses: github/codeql-action/upload-sarif@8a470fddafa5cbb6266ee11b37ef4d8aae19c571  # v3.24.6
         with:
           sarif_file: results.sarif

.github/workflows/tag.yml

@@ -12,7 +12,7 @@ jobs:
     runs-on: ubuntu-20.04
     steps:
     - name: Checkout capa-rules
-      uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
+      uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
       with:
         repository: mandiant/capa-rules
         token: ${{ secrets.CAPA_TOKEN }}
@@ -25,7 +25,7 @@ jobs:
         git tag $name -m "https://github.com/mandiant/capa/releases/$name"
         # TODO update branch name-major=${name%%.*}
     - name: Push tag to capa-rules
-      uses: ad-m/github-push-action@0fafdd62b84042d49ec0cb92d9cac7f7ce4ec79e # master
+      uses: ad-m/github-push-action@d91a481090679876dfc4178fef17f286781251df # v0.8.0
       with:
         repository: mandiant/capa-rules
         github_token: ${{ secrets.CAPA_TOKEN }}

.github/workflows/tests.yml

@@ -1,10 +1,22 @@
 name: CI
 
+# tests.yml workflow will run for all changes except:
+# any file or directory under web/ or doc/
+# any Markdown (.md) file anywhere in the repository
+
 on:
   push:
     branches: [ master ]
+    paths-ignore:
+      - 'web/**'
+      - 'doc/**'
+      - '**.md'
   pull_request:
     branches: [ master ]
+    paths-ignore:
+      - 'web/**'
+      - 'doc/**'
+      - '**.md'
 
 permissions: read-all
 
@@ -17,7 +29,7 @@ jobs:
     runs-on: ubuntu-20.04
     steps:
     - name: Checkout capa
-      uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
+      uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
     # The sync GH action in capa-rules relies on a single '- *$' in the CHANGELOG file
     - name: Ensure CHANGELOG has '- *$'
       run: |
@@ -28,14 +40,16 @@ jobs:
     runs-on: ubuntu-20.04
     steps:
     - name: Checkout capa
-      uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
+      uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
     # use latest available python to take advantage of best performance
     - name: Set up Python 3.11
-      uses: actions/setup-python@d27e3f3d7c64b4bbf8e4abfb9b63b83e846e0435 # v4.5.0
+      uses: actions/setup-python@0a5c61591373683505ea898e09a3ea4f39ef2b9c # v5.0.0
       with:
         python-version: "3.11"
     - name: Install dependencies
-      run: pip install -e .[dev]
+      run: |
+        pip install -r requirements.txt
+        pip install -e .[dev,scripts]
     - name: Lint with ruff
       run: pre-commit run ruff
     - name: Lint with isort
@@ -45,21 +59,25 @@ jobs:
     - name: Lint with flake8
       run: pre-commit run flake8 --hook-stage manual
     - name: Check types with mypy
-      run:  pre-commit run mypy --hook-stage manual
+      run: pre-commit run mypy --hook-stage manual
+    - name: Check imports against dependencies
+      run: pre-commit run deptry --hook-stage manual
 
   rule_linter:
     runs-on: ubuntu-20.04
     steps:
     - name: Checkout capa with submodules
-      uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
+      uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
       with:
         submodules: recursive
     - name: Set up Python 3.11
-      uses: actions/setup-python@d27e3f3d7c64b4bbf8e4abfb9b63b83e846e0435 # v4.5.0
+      uses: actions/setup-python@0a5c61591373683505ea898e09a3ea4f39ef2b9c # v5.0.0
       with:
         python-version: "3.11"
     - name: Install capa
-      run: pip install -e .[dev]
+      run: |
+        pip install -r requirements.txt
+        pip install -e .[dev,scripts]
     - name: Run rule linter
       run: python scripts/lint.py rules/
 
@@ -70,7 +88,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-20.04, windows-2019, macos-11]
+        os: [ubuntu-20.04, windows-2019, macos-12]
         # across all operating systems
         python-version: ["3.8", "3.11"]
         include:
@@ -83,18 +101,20 @@ jobs:
             python-version: "3.10"
     steps:
     - name: Checkout capa with submodules
-      uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
+      uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
       with:
         submodules: recursive
     - name: Set up Python ${{ matrix.python-version }}
-      uses: actions/setup-python@d27e3f3d7c64b4bbf8e4abfb9b63b83e846e0435 # v4.5.0
+      uses: actions/setup-python@0a5c61591373683505ea898e09a3ea4f39ef2b9c # v5.0.0
       with:
         python-version: ${{ matrix.python-version }}
     - name: Install pyyaml
       if: matrix.os == 'ubuntu-20.04'
       run: sudo apt-get install -y libyaml-dev
     - name: Install capa
-      run: pip install -e .[dev]
+      run: |
+        pip install -r requirements.txt
+        pip install -e .[dev,scripts]
     - name: Run tests (fast)
       # this set of tests runs about 80% of the cases in 20% of the time,
       # and should catch most errors quickly.
@@ -106,22 +126,22 @@ jobs:
     name: Binary Ninja tests for ${{ matrix.python-version }}
     env:
       BN_SERIAL: ${{ secrets.BN_SERIAL }}
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-22.04
     needs: [tests]
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["3.8", "3.11"]
+        python-version: ["3.9", "3.11"]
     steps:
     - name: Checkout capa with submodules
       # do only run if BN_SERIAL is available, have to do this in every step, see https://github.com/orgs/community/discussions/26726#discussioncomment-3253118
       if: ${{ env.BN_SERIAL != 0 }}
-      uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
+      uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
       with:
         submodules: recursive
     - name: Set up Python ${{ matrix.python-version }}
       if: ${{ env.BN_SERIAL != 0 }}
-      uses: actions/setup-python@d27e3f3d7c64b4bbf8e4abfb9b63b83e846e0435 # v4.5.0
+      uses: actions/setup-python@0a5c61591373683505ea898e09a3ea4f39ef2b9c # v5.0.0
       with:
         python-version: ${{ matrix.python-version }}
     - name: Install pyyaml
@@ -129,7 +149,9 @@ jobs:
       run: sudo apt-get install -y libyaml-dev
     - name: Install capa
       if: ${{ env.BN_SERIAL != 0 }}
-      run: pip install -e .[dev]
+      run: |
+        pip install -r requirements.txt
+        pip install -e .[dev,scripts]
     - name: install Binary Ninja
       if: ${{ env.BN_SERIAL != 0 }}
       run: |
@@ -153,31 +175,23 @@ jobs:
       matrix:
         python-version: ["3.8", "3.11"]
         java-version: ["17"]
-        gradle-version: ["7.3"]
-        ghidra-version: ["10.3"]
-        public-version: ["PUBLIC_20230510"] # for ghidra releases
-        jep-version: ["4.1.1"]
-        ghidrathon-version: ["3.0.0"]
+        ghidra-version: ["11.0.1"]
+        public-version: ["PUBLIC_20240130"] # for ghidra releases
+        ghidrathon-version: ["4.0.0"] 
     steps:
     - name: Checkout capa with submodules
-      uses: actions/checkout@ac593985615ec2ede58e132d2e21d2b1cbd6127c # v3.3.0
+      uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
       with:
         submodules: true
     - name: Set up Python ${{ matrix.python-version }}
-      uses: actions/setup-python@d27e3f3d7c64b4bbf8e4abfb9b63b83e846e0435 # v4.5.0
+      uses: actions/setup-python@0a5c61591373683505ea898e09a3ea4f39ef2b9c # v5.0.0
       with:
         python-version: ${{ matrix.python-version }}
     - name: Set up Java ${{ matrix.java-version }}
-      uses: actions/setup-java@5ffc13f4174014e2d4d4572b3d74c3fa61aeb2c2 # v3
+      uses: actions/setup-java@387ac29b308b003ca37ba93a6cab5eb57c8f5f93 # v4.0.0
       with:
         distribution: 'temurin'
         java-version: ${{ matrix.java-version }}
-    - name: Set up Gradle ${{ matrix.gradle-version }} 
-      uses: gradle/gradle-build-action@40b6781dcdec2762ad36556682ac74e31030cfe2 # v2.5.1
-      with:
-        gradle-version: ${{ matrix.gradle-version }}
-    - name: Install Jep ${{ matrix.jep-version }} 
-      run : pip install jep==${{ matrix.jep-version }}
     - name: Install Ghidra ${{ matrix.ghidra-version }} 
       run: |
         mkdir ./.github/ghidra
@@ -186,14 +200,17 @@ jobs:
     - name: Install Ghidrathon
       run : |
         mkdir ./.github/ghidrathon
-        curl -o ./.github/ghidrathon/ghidrathon-${{ matrix.ghidrathon-version }}.zip "https://codeload.github.com/mandiant/Ghidrathon/zip/refs/tags/v${{ matrix.ghidrathon-version }}"
-        unzip .github/ghidrathon/ghidrathon-${{ matrix.ghidrathon-version }}.zip -d .github/ghidrathon/
-        gradle -p ./.github/ghidrathon/Ghidrathon-${{ matrix.ghidrathon-version }}/ -PGHIDRA_INSTALL_DIR=$(pwd)/.github/ghidra/ghidra_${{ matrix.ghidra-version }}_PUBLIC
-        unzip .github/ghidrathon/Ghidrathon-${{ matrix.ghidrathon-version }}/dist/*.zip -d .github/ghidra/ghidra_${{ matrix.ghidra-version }}_PUBLIC/Ghidra/Extensions
+        wget "https://github.com/mandiant/Ghidrathon/releases/download/v${{ matrix.ghidrathon-version }}/Ghidrathon-v${{ matrix.ghidrathon-version}}.zip" -O ./.github/ghidrathon/ghidrathon-v${{ matrix.ghidrathon-version }}.zip
+        unzip .github/ghidrathon/ghidrathon-v${{ matrix.ghidrathon-version }}.zip -d .github/ghidrathon/
+        python -m pip install -r .github/ghidrathon/requirements.txt
+        python .github/ghidrathon/ghidrathon_configure.py $(pwd)/.github/ghidra/ghidra_${{ matrix.ghidra-version }}_PUBLIC
+        unzip .github/ghidrathon/Ghidrathon-v${{ matrix.ghidrathon-version }}.zip -d .github/ghidra/ghidra_${{ matrix.ghidra-version }}_PUBLIC/Ghidra/Extensions
     - name: Install pyyaml
       run: sudo apt-get install -y libyaml-dev
     - name: Install capa
-      run: pip install -e .[dev] 
+      run: |
+        pip install -r requirements.txt
+        pip install -e .[dev,scripts]
     - name: Run tests
       run: | 
         mkdir ./.github/ghidra/project
@@ -201,4 +218,4 @@ jobs:
         cat ../output.log
         exit_code=$(cat ../output.log | grep exit | awk '{print $NF}')
         exit $exit_code
- 
+ 

.github/workflows/web-deploy.yml

@@ -0,0 +1,134 @@
+name: deploy web to GitHub Pages
+
+on:
+  push:
+    branches: [ master ]
+    paths:
+      - 'web/**'
+
+  # Allows to run this workflow manually from the Actions tab
+  workflow_dispatch:
+
+# Sets the GITHUB_TOKEN permissions to allow deployment to GitHub Pages
+permissions:
+  contents: read
+  pages: write
+  id-token: write
+
+# Allow one concurrent deployment
+concurrency:
+  group: 'pages'
+  cancel-in-progress: true
+
+jobs:
+  build-landing-page:
+    name: Build landing page
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - uses: actions/upload-artifact@v4
+        with:
+          name: landing-page
+          path: './web/public'
+
+  build-explorer:
+    name: Build capa Explorer Web
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+        with:
+          submodules: 'recursive'
+          fetch-depth: 1
+          show-progress: true
+      - name: Set up Node
+        uses: actions/setup-node@0a44ba7841725637a19e28fa30b79a866c81b0a6 # v4.0.4
+        with:
+          node-version: 20
+          cache: 'npm'
+          cache-dependency-path: './web/explorer/package-lock.json'
+      - name: Install dependencies
+        run: npm ci
+        working-directory: ./web/explorer
+      - name: Generate release bundle
+        run: npm run build:bundle
+        working-directory: ./web/explorer
+      - name: Zip release bundle
+        run: zip -r public/capa-explorer-web.zip capa-explorer-web
+        working-directory: ./web/explorer
+      - name: Build
+        run: npm run build
+        working-directory: ./web/explorer
+      - uses: actions/upload-artifact@v4
+        with:
+          name: explorer
+          path: './web/explorer/dist'
+
+  build-rules:
+    name: Build rules site
+    runs-on: ubuntu-latest
+    steps:
+      - name: Check out the repository
+        uses: actions/checkout@v4
+        with:
+          submodules: 'recursive'
+          # full depth so that capa-rules has a full history
+          # and we can construct a timeline of rule updates.
+          fetch-depth: 0
+      - name: Set up Python
+        uses: actions/setup-python@0a5c61591373683505ea898e09a3ea4f39ef2b9c # v5.0.0
+        with:
+          python-version: '3.12'
+      - uses: extractions/setup-just@v2
+      - name: Install pagefind
+        uses: supplypike/setup-bin@v4
+        with:
+          uri: "https://github.com/CloudCannon/pagefind/releases/download/v1.1.0/pagefind-v1.1.0-x86_64-unknown-linux-musl.tar.gz"
+          name: "pagefind"
+          version: "1.1.0"
+      - name: Install dependencies
+        working-directory: ./web/rules
+        run: pip install -r requirements.txt
+      - name: Build the website
+        working-directory: ./web/rules
+        run: just build
+      - name: Index the website
+        working-directory: ./web/rules
+        run: pagefind --site "public"
+      # upload the build website to artifacts
+      # so that we can download and inspect, if desired.
+      - uses: actions/upload-artifact@v4
+        with:
+          name: rules
+          path: './web/rules/public'
+
+  deploy:
+    name: Deploy site to GitHub Pages
+    environment:
+      name: github-pages
+      url: ${{ steps.deployment.outputs.page_url }}
+    runs-on: ubuntu-latest
+    needs: [build-landing-page, build-explorer, build-rules]
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          name: landing-page
+          path: './public/'
+      - uses: actions/download-artifact@v4
+        with:
+          name: explorer
+          path: './public/explorer'
+      - uses: actions/download-artifact@v4
+        with:
+          name: rules
+          path: './public/rules'
+      - name: Setup Pages
+        uses: actions/configure-pages@v4
+      - name: Upload artifact
+        uses: actions/upload-pages-artifact@v3
+        with:
+          path: './public'
+      - name: Deploy to GitHub Pages
+        id: deployment
+        uses: actions/deploy-pages@v4

.github/workflows/web-tests.yml

@@ -0,0 +1,42 @@
+name: Capa Explorer Web tests 
+
+on:
+  pull_request:
+    branches: [ master ]
+    paths:
+      - 'web/explorer/**'
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+
+    steps:
+    - name: Checkout
+      uses: actions/checkout@v4
+      with:
+        submodules: 'recursive'
+        fetch-depth: 1
+        show-progress: true
+
+    - name: Set up Node
+      uses: actions/setup-node@0a44ba7841725637a19e28fa30b79a866c81b0a6 # v4.0.4
+      with:
+        node-version: 20
+        cache: 'npm'
+        cache-dependency-path: './web/explorer/package-lock.json'
+
+    - name: Install dependencies
+      run: npm ci
+      working-directory: ./web/explorer
+
+    - name: Lint
+      run: npm run lint
+      working-directory: ./web/explorer
+
+    - name: Format
+      run: npm run format:check
+      working-directory: ./web/explorer
+
+    - name: Run unit tests
+      run: npm run test
+      working-directory: ./web/explorer

.gitignore

@@ -126,3 +126,5 @@ Pipfile.lock
 .github/binja/binaryninja
 .github/binja/download_headless.py
 .github/binja/BinaryNinja-headless.zip
+justfile
+data/

.gitmodules

@@ -1,8 +1,6 @@
 [submodule "rules"]
 	path = rules
-	url = ../capa-rules.git
-	branch = dynamic-syntax
+	url = ../../mandiant/capa-rules.git
 [submodule "tests/data"]
 	path = tests/data
-	url = ../capa-testfiles.git
-	branch = dynamic-feature-extractor
+	url = ../../mandiant/capa-testfiles.git

.justfile

@@ -0,0 +1,25 @@
+@isort:
+    pre-commit run isort --show-diff-on-failure --all-files
+
+@black:
+    pre-commit run black --show-diff-on-failure --all-files
+
+@ruff:
+    pre-commit run ruff --all-files
+
+@flake8:
+    pre-commit run flake8 --hook-stage manual --all-files
+
+@mypy:
+    pre-commit run mypy --hook-stage manual --all-files
+
+@deptry:
+    pre-commit run deptry --hook-stage manual --all-files
+
+@lint:
+    -just isort
+    -just black
+    -just ruff
+    -just flake8
+    -just mypy
+    -just deptry

.pre-commit-config.yaml

@@ -38,6 +38,7 @@ repos:
         -   "capa/"
         -   "scripts/"
         -   "tests/"
+        -   "web/rules/scripts/"
         always_run: true
         pass_filenames: false
 
@@ -55,6 +56,7 @@ repos:
         -   "capa/"
         -   "scripts/"
         -   "tests/"
+        -   "web/rules/scripts/"
         always_run: true
         pass_filenames: false
 
@@ -72,6 +74,7 @@ repos:
         -   "capa/"
         -   "scripts/"
         -   "tests/"
+        -   "web/rules/scripts/"
         always_run: true
         pass_filenames: false
 
@@ -86,10 +89,11 @@ repos:
         -   "--config"
         -   ".github/flake8.ini"
         -   "--extend-exclude"
-        -   "capa/render/proto/capa_pb2.py"
+        -   "capa/render/proto/capa_pb2.py,capa/features/extractors/binexport2/binexport2_pb2.py"
         -   "capa/"
         -   "scripts/"
         -   "tests/"
+        -   "web/rules/scripts/"
         always_run: true
         pass_filenames: false
 
@@ -104,9 +108,21 @@ repos:
         -   "--check-untyped-defs"
         -   "--ignore-missing-imports"
         -   "--config-file=.github/mypy/mypy.ini"
+        -   "--enable-incomplete-feature=NewGenericSyntax"
         -   "capa/"
         -   "scripts/"
         -   "tests/"
+        -   "web/rules/scripts/"
+        always_run: true
+        pass_filenames: false
+
+-   repo: local
+    hooks:
+    -   id: deptry
+        name: deptry
+        stages: [push, manual]
+        language: system
+        entry: deptry .
         always_run: true
         pass_filenames: false
 
@@ -127,3 +143,4 @@ repos:
         -   "--ignore=tests/test_scripts.py"
         always_run: true
         pass_filenames: false
+

CITATION.cff

@@ -0,0 +1,8 @@
+cff-version: 1.2.0
+message: "If you use this software, please cite it as below."
+authors:
+  - name: "The FLARE Team"
+title: "capa, a tool to identify capabilities in programs and sandbox traces."
+date-released: 2020-07-16
+url: "https://github.com/mandiant/capa"
+

LICENSE.txt

@@ -187,7 +187,7 @@
       same "printed page" as the copyright notice for easier
       identification within third-party archives.
 
-   Copyright (C) 2023 Mandiant, 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

@@ -1,22 +1,40 @@
-![capa](https://github.com/mandiant/capa/blob/master/.github/logo.png)
+<br />
+<div align="center">
+<a href="https://mandiant.github.io/capa/" target="_blank">
+  <img src="https://github.com/mandiant/capa/blob/master/.github/logo.png">
+</a>
+<p align="center">
+  <a href="https://mandiant.github.io/capa/" target="_blank">Website</a>
+  |
+  <a href="https://github.com/mandiant/capa/releases/latest" target="_blank">Download</a>
+  |
+  <a href="https://mandiant.github.io/capa/explorer/" target="_blank">Web Interface</a>
+</p>
+<div align="center">
 
 [![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/mandiant/capa)](https://github.com/mandiant/capa/releases)
-[![Number of rules](https://img.shields.io/badge/rules-866-blue.svg)](https://github.com/mandiant/capa-rules)
+[![Number of rules](https://gist.githubusercontent.com/capa-bot/6d7960e911f48b3b74916df8988cf0f3/raw/rules_badge.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)
 
+</div>
+</div>
+
+---
+
 capa detects capabilities in executable files.
 You run it against a PE, ELF, .NET module, shellcode file, or a sandbox report 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.mandiant.com/resources/capa-automatically-identify-malware-capabilities)
-- the major version 2.0 updates described in our [second blog post](https://www.mandiant.com/resources/capa-2-better-stronger-faster)
-- the major version 3.0 (ELF support) described in the [third blog post](https://www.mandiant.com/resources/elfant-in-the-room-capa-v3)
-- the major version 4.0 (.NET support) described in the [fourth blog post](https://www.mandiant.com/resources/blog/capa-v4-casting-wider-net)
+To interactively inspect capa results in your browser use the [capa Explorer Web](https://mandiant.github.io/capa/explorer/).
 
+If you want to inspect or write capa rules, head on over to the [capa-rules repository](https://github.com/mandiant/capa-rules). Otherwise, keep reading.
+
+Below you find a list of [our capa blog posts with more details.](#blog-posts)
+
+# example capa output
 ```
 $ capa.exe suspicious.exe
 
@@ -71,16 +89,23 @@ Download stable releases of the standalone capa binaries [here](https://github.c
 
 To use capa as a library or integrate with another tool, see [doc/installation.md](https://github.com/mandiant/capa/blob/master/doc/installation.md) for further setup instructions.
 
-For more information about how to use capa, see [doc/usage.md](https://github.com/mandiant/capa/blob/master/doc/usage.md).
+# capa Explorer Web
+The [capa Explorer Web](https://mandiant.github.io/capa/explorer/) enables you to interactively explore capa results in your web browser. Besides the online version you can download a standalone HTML file for local offline usage.
+
+![capa Explorer Web screenshot](https://github.com/mandiant/capa/blob/master/doc/img/capa_web_explorer.png)
+
+More details on the web UI is available in the [capa Explorer Web README](https://github.com/mandiant/capa/blob/master/web/explorer/README.md).
 
 # example
 
-In the above sample output, we ran capa against an unknown binary (`suspicious.exe`),
-and the tool reported that the program can send HTTP requests, decode data via XOR and Base64,
+In the above sample output, we run capa against an unknown binary (`suspicious.exe`),
+and the tool reports that the program can send HTTP requests, decode data via XOR and Base64,
 install services, and spawn new processes.
 Taken together, this makes us think that `suspicious.exe` could be a persistent backdoor.
 Therefore, our next analysis step might be to run `suspicious.exe` in a sandbox and try to recover the command and control server.
 
+## detailed results
+
 By passing the `-vv` flag (for very verbose), capa reports exactly where it found evidence of these capabilities.
 This is useful for at least two reasons:
 
@@ -125,10 +150,15 @@ function @ 0x4011C0
 ...
 ```
 
-Additionally, capa also supports analyzing [CAPE](https://github.com/kevoreilly/CAPEv2) sandbox reports for dynamic capabilty extraction.
-In order to use this, you first submit your sample to CAPE for analysis, and then run capa against the generated report (JSON).
+capa also supports dynamic capabilities detection for multiple sandboxes including:
+* [CAPE](https://github.com/kevoreilly/CAPEv2) (supported report formats: `.json`, `.json_`, `.json.gz`)
+* [DRAKVUF](https://github.com/CERT-Polska/drakvuf-sandbox/) (supported report formats: `.log`, `.log.gz`)
+* [VMRay](https://www.vmray.com/) (supported report formats: analysis archive `.zip`)
 
-Here's an example of running capa against a packed binary, and then running capa against the CAPE report of that binary:
+
+To use this feature, submit your file to a supported sandbox and then download and run capa against the generated report file. This feature enables capa to match capabilities against dynamic and static features that the sandbox captured during execution.
+
+Here's an example of running capa against a packed file, and then running capa against the CAPE report generated for the same packed file:
 
 ```yaml
 $ capa 05be49819139a3fdcdbddbdefd298398779521f3d68daa25275cc77508e42310.exe
@@ -215,6 +245,7 @@ $ capa 05be49819139a3fdcdbddbdefd298398779521f3d68daa25275cc77508e42310.json
 ┕━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┷━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┙
 ```
 
+# capa rules
 capa uses a collection of rules to identify capabilities within a program.
 These rules are easy to write, even for those new to reverse engineering.
 By authoring rules, you can extend the capabilities that capa recognizes.
@@ -251,15 +282,27 @@ rule:
       - property/read: System.Net.Sockets.TcpClient::Client
 ```
 
-The [github.com/mandiant/capa-rules](https://github.com/mandiant/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 rules that are distributed with capa.
 Please learn to write rules and contribute new entries as you find interesting techniques in malware.
 
+# IDA Pro plugin: capa explorer
 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.
+It also uses your local changes to the .idb to extract better features, such as when you rename a global variable that contains a dynamically resolved API address.
 
 ![capa + IDA Pro integration](https://github.com/mandiant/capa/blob/master/doc/img/explorer_expanded.png)
 
-If you use Ghidra, you can use the Python 3 [Ghidra feature extractor](/capa/ghidra/). This integration enables capa to extract features directly from your Ghidra database, which can help you identify capabilities in programs that you analyze using Ghidra.
+# Ghidra integration
+If you use Ghidra, then you can use the [capa + Ghidra integration](/capa/ghidra/) to run capa's analysis directly on your Ghidra database and render the results in Ghidra's user interface.
+
+<img src="https://github.com/mandiant/capa/assets/66766340/eeae33f4-99d4-42dc-a5e8-4c1b8c661492" width=300>
+
+# blog posts
+- [Dynamic capa: Exploring Executable Run-Time Behavior with the CAPE Sandbox](https://www.mandiant.com/resources/blog/dynamic-capa-executable-behavior-cape-sandbox)
+- [capa v4: casting a wider .NET](https://www.mandiant.com/resources/blog/capa-v4-casting-wider-net) (.NET support)
+- [ELFant in the Room – capa v3](https://www.mandiant.com/resources/elfant-in-the-room-capa-v3) (ELF support)
+- [capa 2.0: Better, Stronger, Faster](https://www.mandiant.com/resources/capa-2-better-stronger-faster)
+- [capa: Automatically Identify Malware Capabilities](https://www.mandiant.com/resources/capa-automatically-identify-malware-capabilities)
 
 # further information
 ## capa

capa/analysis/flirt.py

@@ -0,0 +1,38 @@
+# Copyright (C) 2024 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 pydantic import BaseModel
+
+import capa.features.extractors.ida.idalib as idalib
+
+if not idalib.has_idalib():
+    raise RuntimeError("cannot find IDA idalib module.")
+
+if not idalib.load_idalib():
+    raise RuntimeError("failed to load IDA idalib module.")
+
+import idaapi
+import idautils
+
+
+class FunctionId(BaseModel):
+    va: int
+    is_library: bool
+    name: str
+
+
+def get_flirt_matches(lib_only=True):
+    for fva in idautils.Functions():
+        f = idaapi.get_func(fva)
+        is_lib = bool(f.flags & idaapi.FUNC_LIB)
+        fname = idaapi.get_func_name(fva)
+
+        if lib_only and not is_lib:
+            continue
+
+        yield FunctionId(va=fva, is_library=is_lib, name=fname)

capa/analysis/libraries.py

@@ -0,0 +1,335 @@
+# Copyright (C) 2024 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 sys
+import logging
+import argparse
+import tempfile
+import contextlib
+import collections
+from enum import Enum
+from typing import List, Iterable, Optional
+from pathlib import Path
+
+import rich
+from pydantic import BaseModel
+from rich.text import Text
+from rich.console import Console
+
+import capa.main
+import capa.helpers
+import capa.analysis.flirt
+import capa.analysis.strings
+import capa.features.extractors.ida.idalib as idalib
+
+if not idalib.has_idalib():
+    raise RuntimeError("cannot find IDA idalib module.")
+
+if not idalib.load_idalib():
+    raise RuntimeError("failed to load IDA idalib module.")
+
+import idaapi
+import idapro
+import ida_auto
+import idautils
+
+logger = logging.getLogger(__name__)
+
+
+class Classification(str, Enum):
+    USER = "user"
+    LIBRARY = "library"
+    UNKNOWN = "unknown"
+
+
+class Method(str, Enum):
+    FLIRT = "flirt"
+    STRINGS = "strings"
+    THUNK = "thunk"
+    ENTRYPOINT = "entrypoint"
+    CALLGRAPH = "callgraph"
+
+
+class FunctionClassification(BaseModel):
+    va: int
+    classification: Classification
+    # name per the disassembler/analysis tool
+    # may be combined with the recovered/suspected name TODO below
+    name: str
+
+    # if is library, this must be provided
+    method: Optional[Method]
+
+    # TODO if is library, recovered/suspected name?
+
+    # if is library, these can optionally be provided.
+    library_name: Optional[str] = None
+    library_version: Optional[str] = None
+
+    # additional note on the classification, TODO removeme if not useful beyond dev/debug
+    note: Optional[str] = None
+
+
+class BinaryLayout(BaseModel):
+    va: int
+    # size of the function chunks in bytes
+    size: int
+
+
+class FunctionIdResults(BaseModel):
+    function_classifications: List[FunctionClassification]
+    layout: List[BinaryLayout]
+
+
+@contextlib.contextmanager
+def ida_session(input_path: Path, use_temp_dir=True):
+    if use_temp_dir:
+        t = Path(tempfile.mkdtemp(prefix="ida-")) / input_path.name
+    else:
+        t = input_path
+
+    logger.debug("using %s", str(t))
+    # stderr=True is used here to redirect the spinner banner to stderr,
+    # so that users can redirect capa's output.
+    console = Console(stderr=True, quiet=False)
+
+    try:
+        if use_temp_dir:
+            t.write_bytes(input_path.read_bytes())
+
+        # idalib writes to stdout (ugh), so we have to capture that
+        # so as not to screw up structured output.
+        with capa.helpers.stdout_redirector(io.BytesIO()):
+            idapro.enable_console_messages(False)
+            with capa.main.timing("analyze program"):
+                with console.status("analyzing program...", spinner="dots"):
+                    if idapro.open_database(str(t.absolute()), run_auto_analysis=True):
+                        raise RuntimeError("failed to analyze input file")
+
+            logger.debug("idalib: waiting for analysis...")
+            ida_auto.auto_wait()
+            logger.debug("idalib: opened database.")
+
+        yield
+    finally:
+        idapro.close_database()
+        if use_temp_dir:
+            t.unlink()
+
+
+def get_library_called_functions(
+    function_classifications: list[FunctionClassification],
+) -> Iterable[FunctionClassification]:
+    MAX_PASSES = 10
+    classifications_by_va = capa.analysis.strings.create_index(function_classifications, "va")
+    for n in range(MAX_PASSES):
+        found_new_lib_func = False
+
+        for fva in idautils.Functions():
+            if classifications_by_va.get(fva):
+                # already classified
+                continue
+
+            for ref in idautils.CodeRefsTo(fva, True):
+                f: idaapi.func_t = idaapi.get_func(ref)
+                if not f:
+                    # no function associated with reference location
+                    continue
+
+                ref_fva = f.start_ea
+                fname = idaapi.get_func_name(ref_fva)
+                if fname in ("___tmainCRTStartup",):
+                    # ignore library functions, where we know that they call user-code
+                    # TODO(mr): extend this list
+                    continue
+
+                if classifications := classifications_by_va.get(ref_fva):
+                    for c in classifications:
+                        if c.classification == Classification.LIBRARY:
+                            fc = FunctionClassification(
+                                va=fva,
+                                name=idaapi.get_func_name(fva),
+                                classification=Classification.LIBRARY,
+                                method=Method.CALLGRAPH,
+                                note=f"called by 0x{ref_fva:x} ({c.method.value}{f', {c.library_name}@{c.library_version})' if c.library_name else ')'}",
+                            )
+                            classifications_by_va[fva].append(fc)
+                            yield fc
+                            found_new_lib_func = True
+                            break
+
+        if not found_new_lib_func:
+            logger.debug("no update in pass %d, done here", n)
+            return
+
+
+def is_thunk_function(fva):
+    f = idaapi.get_func(fva)
+    return bool(f.flags & idaapi.FUNC_THUNK)
+
+
+def get_function_size(fva):
+    f = idaapi.get_func(fva)
+    assert f.start_ea == fva
+    return sum([end_ea - start_ea for (start_ea, end_ea) in idautils.Chunks(fva)])
+
+
+def main(argv=None):
+    if argv is None:
+        argv = sys.argv[1:]
+
+    parser = argparse.ArgumentParser(description="Identify library functions using various strategies.")
+    capa.main.install_common_args(parser, wanted={"input_file"})
+    parser.add_argument("--store-idb", action="store_true", default=False, help="store IDA database file")
+    parser.add_argument("--min-string-length", type=int, default=8, help="minimum string length")
+    parser.add_argument("-j", "--json", action="store_true", help="emit JSON instead of text")
+    args = parser.parse_args(args=argv)
+
+    try:
+        capa.main.handle_common_args(args)
+    except capa.main.ShouldExitError as e:
+        return e.status_code
+
+    dbs = capa.analysis.strings.get_default_databases()
+    capa.analysis.strings.prune_databases(dbs, n=args.min_string_length)
+
+    function_classifications: List[FunctionClassification] = []
+    with ida_session(args.input_file, use_temp_dir=not args.store_idb):
+        with capa.main.timing("FLIRT-based library identification"):
+            # TODO: add more signature (files)
+            # TOOD: apply more signatures
+            for flirt_match in capa.analysis.flirt.get_flirt_matches():
+                function_classifications.append(
+                    FunctionClassification(
+                        va=flirt_match.va,
+                        name=flirt_match.name,
+                        classification=Classification.LIBRARY,
+                        method=Method.FLIRT,
+                        # note: we cannot currently include which signature matched per function via the IDA API
+                    )
+                )
+
+        # thunks
+        for fva in idautils.Functions():
+            if is_thunk_function(fva):
+                function_classifications.append(
+                    FunctionClassification(
+                        va=fva,
+                        name=idaapi.get_func_name(fva),
+                        classification=Classification.LIBRARY,
+                        method=Method.THUNK,
+                    )
+                )
+
+        with capa.main.timing("string-based library identification"):
+            for string_match in capa.analysis.strings.get_string_matches(dbs):
+                function_classifications.append(
+                    FunctionClassification(
+                        va=string_match.va,
+                        name=idaapi.get_func_name(string_match.va),
+                        classification=Classification.LIBRARY,
+                        method=Method.STRINGS,
+                        library_name=string_match.metadata.library_name,
+                        library_version=string_match.metadata.library_version,
+                    )
+                )
+
+        for va in idautils.Functions():
+            name = idaapi.get_func_name(va)
+            if name not in {
+                "WinMain",
+                "_main",
+                "main",
+            }:
+                continue
+
+            function_classifications.append(
+                FunctionClassification(
+                    va=va,
+                    name=name,
+                    classification=Classification.USER,
+                    method=Method.ENTRYPOINT,
+                )
+            )
+
+        with capa.main.timing("call graph based library identification"):
+            for fc in get_library_called_functions(function_classifications):
+                function_classifications.append(fc)
+
+        doc = FunctionIdResults(function_classifications=[], layout=[])
+        classifications_by_va = capa.analysis.strings.create_index(function_classifications, "va")
+        for va in idautils.Functions():
+            if classifications := classifications_by_va.get(va):
+                doc.function_classifications.extend(classifications)
+            else:
+                doc.function_classifications.append(
+                    FunctionClassification(
+                        va=va,
+                        name=idaapi.get_func_name(va),
+                        classification=Classification.UNKNOWN,
+                        method=None,
+                    )
+                )
+            doc.layout.append(
+                BinaryLayout(
+                    va=va,
+                    size=get_function_size(va),
+                )
+            )
+
+        if args.json:
+            print(doc.model_dump_json())  # noqa: T201 print found
+
+        else:
+            table = rich.table.Table(
+                "FVA",
+                "CLASSIFICATION",
+                "METHOD",
+                "FNAME",
+                "EXTRA",
+                "SIZE"
+            )
+
+            classifications_by_va = capa.analysis.strings.create_index(doc.function_classifications, "va", sorted_=True)
+            size_by_va = {layout.va: layout.size for layout in doc.layout}
+            size_by_classification = collections.defaultdict(int)
+            for va, classifications in classifications_by_va.items():
+                # TODO count of classifications if multiple?
+                name = ", ".join({c.name for c in classifications})
+                if "sub_" in name:
+                    name = Text(name, style="grey53")
+
+                classification = {c.classification for c in classifications}
+                method = {c.method for c in classifications if c.method}
+                extra = {f"{c.library_name}@{c.library_version}" for c in classifications if c.library_name}
+                note = {f"{c.note}" for c in classifications if c.note}
+
+                table.add_row(
+                    hex(va),
+                    ", ".join(classification) if classification != {"unknown"} else Text("unknown", style="grey53"),
+                    ", ".join(method),
+                    name,
+                    f"{', '.join(extra)} {', '.join(note)}",
+                    f"{size_by_va[va]}",
+                )
+
+                size_by_classification["-".join(classification)] += size_by_va[va]
+
+            rich.print(table)
+
+            stats_table = rich.table.Table(
+                "ID", rich.table.Column("SIZE", justify="right"), rich.table.Column("%", justify="right")
+            )
+            size_all = sum(size_by_classification.values())
+            for k, s in size_by_classification.items():
+                stats_table.add_row(k, f"{s:d}", f"{100 * s / size_all:.2f}")
+            rich.print(stats_table)
+
+
+if __name__ == "__main__":
+    sys.exit(main())

capa/analysis/requirements.txt

@@ -0,0 +1,2 @@
+# temporary extra file to track dependencies of the analysis directory
+nltk==3.9.1

capa/analysis/strings/__init__.py

@@ -0,0 +1,269 @@
+# Copyright (C) 2024 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.
+
+"""
+further requirements:
+  - nltk
+"""
+import gzip
+import logging
+import collections
+from typing import Any, Dict, Mapping
+from pathlib import Path
+from dataclasses import dataclass
+
+import msgspec
+
+import capa.features.extractors.strings
+
+logger = logging.getLogger(__name__)
+
+
+class LibraryString(msgspec.Struct):
+    string: str
+    library_name: str
+    library_version: str
+    file_path: str | None = None
+    function_name: str | None = None
+    line_number: int | None = None
+
+
+@dataclass
+class LibraryStringDatabase:
+    metadata_by_string: Dict[str, LibraryString]
+
+    def __len__(self) -> int:
+        return len(self.metadata_by_string)
+
+    @classmethod
+    def from_file(cls, path: Path) -> "LibraryStringDatabase":
+        metadata_by_string: Dict[str, LibraryString] = {}
+        decoder = msgspec.json.Decoder(type=LibraryString)
+        for line in gzip.decompress(path.read_bytes()).split(b"\n"):
+            if not line:
+                continue
+            s = decoder.decode(line)
+            metadata_by_string[s.string] = s
+
+        return cls(metadata_by_string=metadata_by_string)
+
+
+DEFAULT_FILENAMES = (
+    "brotli.jsonl.gz",
+    "bzip2.jsonl.gz",
+    "cryptopp.jsonl.gz",
+    "curl.jsonl.gz",
+    "detours.jsonl.gz",
+    "jemalloc.jsonl.gz",
+    "jsoncpp.jsonl.gz",
+    "kcp.jsonl.gz",
+    "liblzma.jsonl.gz",
+    "libsodium.jsonl.gz",
+    "libpcap.jsonl.gz",
+    "mbedtls.jsonl.gz",
+    "openssl.jsonl.gz",
+    "sqlite3.jsonl.gz",
+    "tomcrypt.jsonl.gz",
+    "wolfssl.jsonl.gz",
+    "zlib.jsonl.gz",
+)
+
+DEFAULT_PATHS = tuple(Path(__file__).parent / "data" / "oss" / filename for filename in DEFAULT_FILENAMES) + (
+    Path(__file__).parent / "data" / "crt" / "msvc_v143.jsonl.gz",
+)
+
+
+def get_default_databases() -> list[LibraryStringDatabase]:
+    return [LibraryStringDatabase.from_file(path) for path in DEFAULT_PATHS]
+
+
+@dataclass
+class WindowsApiStringDatabase:
+    dll_names: set[str]
+    api_names: set[str]
+
+    def __len__(self) -> int:
+        return len(self.dll_names) + len(self.api_names)
+
+    @classmethod
+    def from_dir(cls, path: Path) -> "WindowsApiStringDatabase":
+        dll_names: set[str] = set()
+        api_names: set[str] = set()
+
+        for line in gzip.decompress((path / "dlls.txt.gz").read_bytes()).decode("utf-8").splitlines():
+            if not line:
+                continue
+            dll_names.add(line)
+
+        for line in gzip.decompress((path / "apis.txt.gz").read_bytes()).decode("utf-8").splitlines():
+            if not line:
+                continue
+            api_names.add(line)
+
+        return cls(dll_names=dll_names, api_names=api_names)
+
+    @classmethod
+    def from_defaults(cls) -> "WindowsApiStringDatabase":
+        return cls.from_dir(Path(__file__).parent / "data" / "winapi")
+
+
+def extract_strings(buf, n=4):
+    yield from capa.features.extractors.strings.extract_ascii_strings(buf, n=n)
+    yield from capa.features.extractors.strings.extract_unicode_strings(buf, n=n)
+
+
+def prune_databases(dbs: list[LibraryStringDatabase], n=8):
+    """remove less trustyworthy database entries.
+
+    such as:
+      - those found in multiple databases
+      - those that are English words
+      - those that are too short
+      - Windows API and DLL names
+    """
+
+    # TODO: consider applying these filters directly to the persisted databases, not at load time.
+
+    winapi = WindowsApiStringDatabase.from_defaults()
+
+    try:
+        from nltk.corpus import words as nltk_words
+
+        nltk_words.words()
+    except (ImportError, LookupError):
+        # one-time download of dataset.
+        # this probably doesn't work well for embedded use.
+        import nltk
+
+        nltk.download("words")
+        from nltk.corpus import words as nltk_words
+    words = set(nltk_words.words())
+
+    counter: collections.Counter[str] = collections.Counter()
+    to_remove = set()
+    for db in dbs:
+        for string in db.metadata_by_string.keys():
+            counter[string] += 1
+
+            if string in words:
+                to_remove.add(string)
+                continue
+
+            if len(string) < n:
+                to_remove.add(string)
+                continue
+
+            if string in winapi.api_names:
+                to_remove.add(string)
+                continue
+
+            if string in winapi.dll_names:
+                to_remove.add(string)
+                continue
+
+    for string, count in counter.most_common():
+        if count <= 1:
+            break
+
+        # remove strings that are seen in more than one database
+        to_remove.add(string)
+
+    for db in dbs:
+        for string in to_remove:
+            if string in db.metadata_by_string:
+                del db.metadata_by_string[string]
+
+
+def get_function_strings():
+    import idaapi
+    import idautils
+
+    import capa.features.extractors.ida.helpers as ida_helpers
+
+    strings_by_function = collections.defaultdict(set)
+    for ea in idautils.Functions():
+        f = idaapi.get_func(ea)
+
+        # ignore library functions and thunk functions as identified by IDA
+        if f.flags & idaapi.FUNC_THUNK:
+            continue
+        if f.flags & idaapi.FUNC_LIB:
+            continue
+
+        for bb in ida_helpers.get_function_blocks(f):
+            for insn in ida_helpers.get_instructions_in_range(bb.start_ea, bb.end_ea):
+                ref = capa.features.extractors.ida.helpers.find_data_reference_from_insn(insn)
+                if ref == insn.ea:
+                    continue
+
+                string = capa.features.extractors.ida.helpers.find_string_at(ref)
+                if not string:
+                    continue
+
+                strings_by_function[ea].add(string)
+
+    return strings_by_function
+
+
+@dataclass
+class LibraryStringClassification:
+    va: int
+    string: str
+    library_name: str
+    metadata: LibraryString
+
+
+def create_index(s: list, k: str, sorted_: bool = False) -> Mapping[Any, list]:
+    """create an index of the elements in `s` using the key `k`, optionally sorted by `k`"""
+    if sorted_:
+        s = sorted(s, key=lambda x: getattr(x, k))
+
+    s_by_k = collections.defaultdict(list)
+    for v in s:
+        p = getattr(v, k)
+        s_by_k[p].append(v)
+    return s_by_k
+
+
+def get_string_matches(dbs: list[LibraryStringDatabase]) -> list[LibraryStringClassification]:
+    matches: list[LibraryStringClassification] = []
+
+    for function, strings in sorted(get_function_strings().items()):
+        for string in strings:
+            for db in dbs:
+                if metadata := db.metadata_by_string.get(string):
+                    matches.append(
+                        LibraryStringClassification(
+                            va=function,
+                            string=string,
+                            library_name=metadata.library_name,
+                            metadata=metadata,
+                        )
+                    )
+
+    # if there are less than N strings per library, ignore that library
+    matches_by_library = create_index(matches, "library_name")
+    for library_name, library_matches in matches_by_library.items():
+        if len(library_matches) > 5:
+            continue
+
+        logger.info("pruning library %s: only %d matched string", library_name, len(library_matches))
+        matches = [m for m in matches if m.library_name != library_name]
+
+    # if there are conflicts within a single function, don't label it
+    matches_by_function = create_index(matches, "va")
+    for va, function_matches in matches_by_function.items():
+        library_names = {m.library_name for m in function_matches}
+        if len(library_names) == 1:
+            continue
+
+        logger.info("conflicting matches: 0x%x: %s", va, sorted(library_names))
+        # this is potentially slow (O(n**2)) but hopefully fast enough in practice.
+        matches = [m for m in matches if m.va != va]
+
+    return matches

capa/analysis/strings/__main__.py

@@ -0,0 +1,130 @@
+# Copyright (C) 2024 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 logging
+import collections
+from pathlib import Path
+
+import rich
+from rich.text import Text
+
+import capa.analysis.strings
+import capa.features.extractors.strings
+import capa.features.extractors.ida.helpers as ida_helpers
+
+logger = logging.getLogger(__name__)
+
+
+def open_ida(input_path: Path):
+    import tempfile
+
+    import idapro
+
+    t = Path(tempfile.mkdtemp(prefix="ida-")) / input_path.name
+    t.write_bytes(input_path.read_bytes())
+    # resource leak: we should delete this upon exit
+
+    idapro.enable_console_messages(False)
+    idapro.open_database(str(t.absolute()), run_auto_analysis=True)
+
+    import ida_auto
+
+    ida_auto.auto_wait()
+
+
+def main():
+    logging.basicConfig(level=logging.DEBUG)
+
+    # use n=8 to ignore common words
+    N = 8
+
+    input_path = Path(sys.argv[1])
+
+    dbs = capa.analysis.strings.get_default_databases()
+    capa.analysis.strings.prune_databases(dbs, n=N)
+
+    strings_by_library = collections.defaultdict(set)
+    for string in capa.analysis.strings.extract_strings(input_path.read_bytes(), n=N):
+        for db in dbs:
+            if metadata := db.metadata_by_string.get(string.s):
+                strings_by_library[metadata.library_name].add(string.s)
+
+    console = rich.get_console()
+    console.print("found libraries:", style="bold")
+    for library, strings in sorted(strings_by_library.items(), key=lambda p: len(p[1]), reverse=True):
+        console.print(f"  - [b]{library}[/] ({len(strings)} strings)")
+
+        for string in sorted(strings)[:10]:
+            console.print(f"    - {string}", markup=False, style="grey37")
+
+        if len(strings) > 10:
+            console.print("    ...", style="grey37")
+
+    if not strings_by_library:
+        console.print("  (none)", style="grey37")
+        # since we're not going to find any strings
+        # return early and don't do IDA analysis
+        return
+
+    open_ida(input_path)
+
+    import idaapi
+    import idautils
+    import ida_funcs
+
+    strings_by_function = collections.defaultdict(set)
+    for ea in idautils.Functions():
+        f = idaapi.get_func(ea)
+
+        # ignore library functions and thunk functions as identified by IDA
+        if f.flags & idaapi.FUNC_THUNK:
+            continue
+        if f.flags & idaapi.FUNC_LIB:
+            continue
+
+        for bb in ida_helpers.get_function_blocks(f):
+            for insn in ida_helpers.get_instructions_in_range(bb.start_ea, bb.end_ea):
+                ref = capa.features.extractors.ida.helpers.find_data_reference_from_insn(insn)
+                if ref == insn.ea:
+                    continue
+
+                string = capa.features.extractors.ida.helpers.find_string_at(ref)
+                if not string:
+                    continue
+
+                for db in dbs:
+                    if metadata := db.metadata_by_string.get(string):
+                        strings_by_function[ea].add(string)
+
+    # ensure there are at least XXX functions renamed, or ignore those entries
+
+    console.print("functions:", style="bold")
+    for function, strings in sorted(strings_by_function.items()):
+        if strings:
+            name = ida_funcs.get_func_name(function)
+
+            console.print(f"  [b]{name}[/]@{function:08x}:")
+
+            for string in strings:
+                for db in dbs:
+                    if metadata := db.metadata_by_string.get(string):
+                        location = Text(
+                            f"{metadata.library_name}@{metadata.library_version}::{metadata.function_name}",
+                            style="grey37",
+                        )
+                        console.print("    - ", location, ": ", string.rstrip())
+
+    console.print()
+
+    console.print(
+        f"found {len(strings_by_function)} library functions across {len(list(idautils.Functions()))} functions"
+    )
+
+
+if __name__ == "__main__":
+    main()

capa/analysis/strings/data/oss/.gitignore

@@ -0,0 +1,3 @@
+*.csv
+*.jsonl
+*.jsonl.gz

capa/analysis/strings/data/oss/readme.md

@@ -0,0 +1,99 @@
+# Strings from Open Source libraries
+
+This directory contains databases of strings extracted from open soure software. 
+capa uses these databases to ignore functions that are likely library code.
+
+There is one file for each database. Each database is a gzip-compressed, JSONL (one JSON document per line) file.
+The JSON document looks like this:
+
+    string: "1.0.8, 13-Jul-2019"
+    library_name: "bzip2"
+    library_version: "1.0.8#3"
+    file_path: "CMakeFiles/bz2.dir/bzlib.c.obj"
+    function_name: "BZ2_bzlibVersion"
+    line_number: null
+
+The following databases were extracted via the vkpkg & jh technique:
+
+  - brotli 1.0.9#5
+  - bzip2 1.0.8#3
+  - cryptopp 8.7.0
+  - curl 7.86.0#1
+  - detours 4.0.1#7
+  - jemalloc 5.3.0#1
+  - jsoncpp 1.9.5
+  - kcp 1.7
+  - liblzma 5.2.5#6
+  - libsodium 1.0.18#8
+  - libpcap 1.10.1#3
+  - mbedtls 2.28.1
+  - openssl 3.0.7#1
+  - sqlite3 3.40.0#1
+  - tomcrypt 1.18.2#2
+  - wolfssl 5.5.0
+  - zlib 1.2.13
+
+This code was originally developed in FLOSS and imported into capa.
+
+## The vkpkg & jh technique
+
+Major steps:
+
+  1. build static libraries via vcpkg
+  2. extract features via jh
+  3. convert to JSONL format with `jh_to_qs.py`
+  4. compress with gzip
+
+### Build static libraries via vcpkg
+
+[vcpkg](https://vcpkg.io/en/) is a free C/C++ package manager for acquiring and managing libraries.
+We use it to easily build common open source libraries, like zlib.
+Use the triplet `x64-windows-static` to build static archives (.lib files that are AR archives containing COFF object files):
+
+```console
+PS > C:\vcpkg\vcpkg.exe install --triplet x64-windows-static zlib
+```
+
+### Extract features via jh
+
+[jh](https://github.com/williballenthin/lancelot/blob/master/bin/src/bin/jh.rs)
+is a lancelot-based utility that parses AR archives containing COFF object files,
+reconstructs their control flow, finds functions, and extracts features. 
+jh extracts numbers, API calls, and strings; we are only interested in the string features.
+
+For each feature, jh emits a CSV line with the fields 
+  - target triplet
+  - compiler 
+  - library
+  - version
+  - build profile
+  - path
+  - function
+  - feature type
+  - feature value
+
+For example:
+
+```csv
+x64-windows-static,msvc143,bzip2,1.0.8#3,release,CMakeFiles/bz2.dir/bzlib.c.obj,BZ2_bzBuffToBuffCompress,number,0x00000100
+```
+
+For example, to invoke jh:
+
+```console
+$ ~/lancelot/target/release/jh x64-windows-static msvc143 zlib 1.2.13 release /mnt/c/vcpkg/installed/x64-windows-static/lib/zlib.lib > ~/flare-floss/floss/qs/db/data/oss/zlib.csv
+```
+
+### Convert to OSS database format
+
+We use the script `jh_to_qs.py` to convert these CSV lines into JSONL file prepared for FLOSS:
+
+```console
+$ python3 jh_to_qs.py zlib.csv > zlib.jsonl
+```
+
+These files are then gzip'd:
+
+```console
+$  gzip -c zlib.jsonl > zlib.jsonl.gz
+```

capa/capabilities/dynamic.py

@@ -9,16 +9,13 @@
 import logging
 import itertools
 import collections
-from typing import Any, Tuple
-
-import tqdm
+from typing import Any, List, Tuple
 
 import capa.perf
 import capa.features.freeze as frz
 import capa.render.result_document as rdoc
 from capa.rules import Scope, RuleSet
 from capa.engine import FeatureSet, MatchResults
-from capa.helpers import redirecting_print_to_tqdm
 from capa.capabilities.common import find_file_capabilities
 from capa.features.extractors.base_extractor import CallHandle, ThreadHandle, ProcessHandle, DynamicFeatureExtractor
 
@@ -65,7 +62,7 @@ def find_thread_capabilities(
     features: FeatureSet = collections.defaultdict(set)
 
     # matches found at the call scope.
-    # might be found at different calls, thats ok.
+    # might be found at different calls, that's ok.
     call_matches: MatchResults = collections.defaultdict(list)
 
     for ch in extractor.get_calls(ph, th):
@@ -103,11 +100,11 @@ def find_process_capabilities(
     process_features: FeatureSet = collections.defaultdict(set)
 
     # matches found at the basic threads.
-    # might be found at different threads, thats ok.
+    # might be found at different threads, that's ok.
     thread_matches: MatchResults = collections.defaultdict(list)
 
     # matches found at the call scope.
-    # might be found at different calls, thats ok.
+    # might be found at different calls, that's ok.
     call_matches: MatchResults = collections.defaultdict(list)
 
     for th in extractor.get_threads(ph):
@@ -138,33 +135,30 @@ def find_dynamic_capabilities(
     feature_counts = rdoc.DynamicFeatureCounts(file=0, processes=())
 
     assert isinstance(extractor, DynamicFeatureExtractor)
-    with redirecting_print_to_tqdm(disable_progress):
-        with tqdm.contrib.logging.logging_redirect_tqdm():
-            pbar = tqdm.tqdm
-            if disable_progress:
-                # do not use tqdm to avoid unnecessary side effects when caller intends
-                # to disable progress completely
-                def pbar(s, *args, **kwargs):
-                    return s
+    processes: List[ProcessHandle] = list(extractor.get_processes())
+    n_processes: int = len(processes)
 
-            processes = list(extractor.get_processes())
+    with capa.helpers.CapaProgressBar(
+        console=capa.helpers.log_console, transient=True, disable=disable_progress
+    ) as pbar:
+        task = pbar.add_task("matching", total=n_processes, unit="processes")
+        for p in processes:
+            process_matches, thread_matches, call_matches, feature_count = find_process_capabilities(
+                ruleset, extractor, p
+            )
+            feature_counts.processes += (
+                rdoc.ProcessFeatureCount(address=frz.Address.from_capa(p.address), count=feature_count),
+            )
+            logger.debug("analyzed %s and extracted %d features", p.address, feature_count)
 
-            pb = pbar(processes, desc="matching", unit=" processes", leave=False)
-            for p in pb:
-                process_matches, thread_matches, call_matches, feature_count = find_process_capabilities(
-                    ruleset, extractor, p
-                )
-                feature_counts.processes += (
-                    rdoc.ProcessFeatureCount(address=frz.Address.from_capa(p.address), count=feature_count),
-                )
-                logger.debug("analyzed %s and extracted %d features", p.address, feature_count)
+            for rule_name, res in process_matches.items():
+                all_process_matches[rule_name].extend(res)
+            for rule_name, res in thread_matches.items():
+                all_thread_matches[rule_name].extend(res)
+            for rule_name, res in call_matches.items():
+                all_call_matches[rule_name].extend(res)
 
-                for rule_name, res in process_matches.items():
-                    all_process_matches[rule_name].extend(res)
-                for rule_name, res in thread_matches.items():
-                    all_thread_matches[rule_name].extend(res)
-                for rule_name, res in call_matches.items():
-                    all_call_matches[rule_name].extend(res)
+            pbar.advance(task)
 
     # collection of features that captures the rule matches within process and thread scopes.
     # mapping from feature (matched rule) to set of addresses at which it matched.

capa/capabilities/static.py

@@ -10,16 +10,14 @@ import time
 import logging
 import itertools
 import collections
-from typing import Any, Tuple
-
-import tqdm.contrib.logging
+from typing import Any, List, Tuple
 
 import capa.perf
+import capa.helpers
 import capa.features.freeze as frz
 import capa.render.result_document as rdoc
 from capa.rules import Scope, RuleSet
 from capa.engine import FeatureSet, MatchResults
-from capa.helpers import redirecting_print_to_tqdm
 from capa.capabilities.common import find_file_capabilities
 from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle, StaticFeatureExtractor
 
@@ -66,7 +64,7 @@ def find_basic_block_capabilities(
     features: FeatureSet = collections.defaultdict(set)
 
     # matches found at the instruction scope.
-    # might be found at different instructions, thats ok.
+    # might be found at different instructions, that's ok.
     insn_matches: MatchResults = collections.defaultdict(list)
 
     for insn in extractor.get_instructions(f, bb):
@@ -106,11 +104,11 @@ def find_code_capabilities(
     function_features: FeatureSet = collections.defaultdict(set)
 
     # matches found at the basic block scope.
-    # might be found at different basic blocks, thats ok.
+    # might be found at different basic blocks, that's ok.
     bb_matches: MatchResults = collections.defaultdict(list)
 
     # matches found at the instruction scope.
-    # might be found at different instructions, thats ok.
+    # might be found at different instructions, that's ok.
     insn_matches: MatchResults = collections.defaultdict(list)
 
     for bb in extractor.get_basic_blocks(fh):
@@ -142,63 +140,58 @@ def find_static_capabilities(
     library_functions: Tuple[rdoc.LibraryFunction, ...] = ()
 
     assert isinstance(extractor, StaticFeatureExtractor)
-    with redirecting_print_to_tqdm(disable_progress):
-        with tqdm.contrib.logging.logging_redirect_tqdm():
-            pbar = tqdm.tqdm
-            if capa.helpers.is_runtime_ghidra():
-                # Ghidrathon interpreter cannot properly handle
-                # the TMonitor thread that is created via a monitor_interval
-                # > 0
-                pbar.monitor_interval = 0
-            if disable_progress:
-                # do not use tqdm to avoid unnecessary side effects when caller intends
-                # to disable progress completely
-                def pbar(s, *args, **kwargs):
-                    return s
+    functions: List[FunctionHandle] = list(extractor.get_functions())
+    n_funcs: int = len(functions)
+    n_libs: int = 0
+    percentage: float = 0
 
-            functions = list(extractor.get_functions())
-            n_funcs = len(functions)
-
-            pb = pbar(functions, desc="matching", unit=" functions", postfix="skipped 0 library functions", leave=False)
-            for f in pb:
-                t0 = time.time()
-                if extractor.is_library_function(f.address):
-                    function_name = extractor.get_function_name(f.address)
-                    logger.debug("skipping library function 0x%x (%s)", f.address, function_name)
-                    library_functions += (
-                        rdoc.LibraryFunction(address=frz.Address.from_capa(f.address), name=function_name),
-                    )
-                    n_libs = len(library_functions)
-                    percentage = round(100 * (n_libs / n_funcs))
-                    if isinstance(pb, tqdm.tqdm):
-                        pb.set_postfix_str(f"skipped {n_libs} library functions ({percentage}%)")
-                    continue
-
-                function_matches, bb_matches, insn_matches, feature_count = find_code_capabilities(
-                    ruleset, extractor, f
+    with capa.helpers.CapaProgressBar(
+        console=capa.helpers.log_console, transient=True, disable=disable_progress
+    ) as pbar:
+        task = pbar.add_task(
+            "matching", total=n_funcs, unit="functions", postfix=f"skipped {n_libs} library functions, {percentage}%"
+        )
+        for f in functions:
+            t0 = time.time()
+            if extractor.is_library_function(f.address):
+                function_name = extractor.get_function_name(f.address)
+                logger.debug("skipping library function 0x%x (%s)", f.address, function_name)
+                library_functions += (
+                    rdoc.LibraryFunction(address=frz.Address.from_capa(f.address), name=function_name),
                 )
-                feature_counts.functions += (
-                    rdoc.FunctionFeatureCount(address=frz.Address.from_capa(f.address), count=feature_count),
-                )
-                t1 = time.time()
+                n_libs = len(library_functions)
+                percentage = round(100 * (n_libs / n_funcs))
+                pbar.update(task, postfix=f"skipped {n_libs} library functions, {percentage}%")
+                pbar.advance(task)
+                continue
 
-                match_count = sum(len(res) for res in function_matches.values())
-                match_count += sum(len(res) for res in bb_matches.values())
-                match_count += sum(len(res) for res in insn_matches.values())
-                logger.debug(
-                    "analyzed function 0x%x and extracted %d features, %d matches in %0.02fs",
-                    f.address,
-                    feature_count,
-                    match_count,
-                    t1 - t0,
-                )
+            function_matches, bb_matches, insn_matches, feature_count = find_code_capabilities(ruleset, extractor, f)
+            feature_counts.functions += (
+                rdoc.FunctionFeatureCount(address=frz.Address.from_capa(f.address), count=feature_count),
+            )
+            t1 = time.time()
 
-                for rule_name, res in function_matches.items():
-                    all_function_matches[rule_name].extend(res)
-                for rule_name, res in bb_matches.items():
-                    all_bb_matches[rule_name].extend(res)
-                for rule_name, res in insn_matches.items():
-                    all_insn_matches[rule_name].extend(res)
+            match_count = 0
+            for name, matches_ in itertools.chain(function_matches.items(), bb_matches.items(), insn_matches.items()):
+                if not ruleset.rules[name].is_subscope_rule():
+                    match_count += len(matches_)
+
+            logger.debug(
+                "analyzed function 0x%x and extracted %d features, %d matches in %0.02fs",
+                f.address,
+                feature_count,
+                match_count,
+                t1 - t0,
+            )
+
+            for rule_name, res in function_matches.items():
+                all_function_matches[rule_name].extend(res)
+            for rule_name, res in bb_matches.items():
+                all_bb_matches[rule_name].extend(res)
+            for rule_name, res in insn_matches.items():
+                all_insn_matches[rule_name].extend(res)
+
+            pbar.advance(task)
 
     # collection of features that captures the rule matches within function, BB, and instruction scopes.
     # mapping from feature (matched rule) to set of addresses at which it matched.
@@ -213,7 +206,7 @@ def find_static_capabilities(
     all_file_matches, feature_count = find_file_capabilities(ruleset, extractor, function_and_lower_features)
     feature_counts.file = feature_count
 
-    matches = dict(
+    matches: MatchResults = dict(
         itertools.chain(
             # each rule exists in exactly one scope,
             # so there won't be any overlap among these following MatchResults,

capa/engine.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, 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
@@ -102,14 +102,14 @@ class And(Statement):
         super().__init__(description=description)
         self.children = children
 
-    def evaluate(self, ctx, short_circuit=True):
+    def evaluate(self, features: FeatureSet, 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)
+                result = child.evaluate(features, short_circuit=short_circuit)
                 results.append(result)
                 if not result:
                     # short circuit
@@ -117,7 +117,7 @@ class And(Statement):
 
             return Result(True, self, results)
         else:
-            results = [child.evaluate(ctx, short_circuit=short_circuit) for child in self.children]
+            results = [child.evaluate(features, short_circuit=short_circuit) for child in self.children]
             success = all(results)
             return Result(success, self, results)
 
@@ -135,14 +135,14 @@ class Or(Statement):
         super().__init__(description=description)
         self.children = children
 
-    def evaluate(self, ctx, short_circuit=True):
+    def evaluate(self, features: FeatureSet, 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)
+                result = child.evaluate(features, short_circuit=short_circuit)
                 results.append(result)
                 if result:
                     # short circuit as soon as we hit one match
@@ -150,7 +150,7 @@ class Or(Statement):
 
             return Result(False, self, results)
         else:
-            results = [child.evaluate(ctx, short_circuit=short_circuit) for child in self.children]
+            results = [child.evaluate(features, short_circuit=short_circuit) for child in self.children]
             success = any(results)
             return Result(success, self, results)
 
@@ -162,11 +162,11 @@ class Not(Statement):
         super().__init__(description=description)
         self.child = child
 
-    def evaluate(self, ctx, short_circuit=True):
+    def evaluate(self, features: FeatureSet, 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)]
+        results = [self.child.evaluate(features, short_circuit=short_circuit)]
         success = not results[0]
         return Result(success, self, results)
 
@@ -185,7 +185,7 @@ class Some(Statement):
         self.count = count
         self.children = children
 
-    def evaluate(self, ctx, short_circuit=True):
+    def evaluate(self, features: FeatureSet, short_circuit=True):
         capa.perf.counters["evaluate.feature"] += 1
         capa.perf.counters["evaluate.feature.some"] += 1
 
@@ -193,7 +193,7 @@ class Some(Statement):
             results = []
             satisfied_children_count = 0
             for child in self.children:
-                result = child.evaluate(ctx, short_circuit=short_circuit)
+                result = child.evaluate(features, short_circuit=short_circuit)
                 results.append(result)
                 if result:
                     satisfied_children_count += 1
@@ -204,7 +204,7 @@ class Some(Statement):
 
             return Result(False, self, results)
         else:
-            results = [child.evaluate(ctx, short_circuit=short_circuit) for child in self.children]
+            results = [child.evaluate(features, 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.
             #
@@ -214,7 +214,7 @@ class Some(Statement):
 
 
 class Range(Statement):
-    """match if the child is contained in the ctx set with a count in the given range."""
+    """match if the child is contained in the feature set with a count in the given range."""
 
     def __init__(self, child, min=None, max=None, description=None):
         super().__init__(description=description)
@@ -222,15 +222,15 @@ 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, **kwargs):
+    def evaluate(self, features: FeatureSet, short_circuit=True):
         capa.perf.counters["evaluate.feature"] += 1
         capa.perf.counters["evaluate.feature.range"] += 1
 
-        count = len(ctx.get(self.child, []))
+        count = len(features.get(self.child, []))
         if self.min == 0 and count == 0:
             return Result(True, self, [])
 
-        return Result(self.min <= count <= self.max, self, [], locations=ctx.get(self.child))
+        return Result(self.min <= count <= self.max, self, [], locations=features.get(self.child))
 
     def __str__(self):
         if self.max == (1 << 64 - 1):
@@ -250,7 +250,7 @@ class Subscope(Statement):
         self.scope = scope
         self.child = child
 
-    def evaluate(self, ctx, **kwargs):
+    def evaluate(self, features: FeatureSet, short_circuit=True):
         raise ValueError("cannot evaluate a subscope directly!")
 
 
@@ -270,6 +270,14 @@ class Subscope(Statement):
 MatchResults = Mapping[str, List[Tuple[Address, Result]]]
 
 
+def get_rule_namespaces(rule: "capa.rules.Rule") -> Iterator[str]:
+    namespace = rule.meta.get("namespace")
+    if namespace:
+        while namespace:
+            yield namespace
+            namespace, _, _ = namespace.rpartition("/")
+
+
 def index_rule_matches(features: FeatureSet, rule: "capa.rules.Rule", locations: Iterable[Address]):
     """
     record into the given featureset that the given rule matched at the given locations.
@@ -280,11 +288,8 @@ def index_rule_matches(features: FeatureSet, rule: "capa.rules.Rule", locations:
     updates `features` in-place. doesn't modify the remaining arguments.
     """
     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("/")
+    for namespace in get_rule_namespaces(rule):
+        features[capa.features.common.MatchedRule(namespace)].update(locations)
 
 
 def match(rules: List["capa.rules.Rule"], features: FeatureSet, addr: Address) -> Tuple[FeatureSet, MatchResults]:

capa/exceptions.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2022 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
@@ -23,3 +23,15 @@ class UnsupportedOSError(ValueError):
 
 class EmptyReportError(ValueError):
     pass
+
+
+class InvalidArgument(ValueError):
+    pass
+
+
+class NonExistantFunctionError(ValueError):
+    pass
+
+
+class NonExistantProcessError(ValueError):
+    pass

capa/features/address.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2022 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
@@ -93,7 +93,7 @@ class ThreadAddress(Address):
 
 
 class DynamicCallAddress(Address):
-    """addesses a call in a dynamic execution trace"""
+    """addresses a call in a dynamic execution trace"""
 
     def __init__(self, thread: ThreadAddress, id: int):
         assert id >= 0

capa/features/basicblock.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, 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/com/__init__.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2024 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/com/classes.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2024 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/com/interfaces.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2024 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/common.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2021 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
@@ -128,7 +128,7 @@ class Feature(abc.ABC):  # noqa: B024
 
     def __lt__(self, other):
         # implementing sorting by serializing to JSON is a huge hack.
-        # its slow, inelegant, and probably doesn't work intuitively;
+        # it's slow, inelegant, and probably doesn't work intuitively;
         # however, we only use it for deterministic output, so it's good enough for now.
 
         # circular import
@@ -166,10 +166,10 @@ class Feature(abc.ABC):  # noqa: B024
     def __repr__(self):
         return str(self)
 
-    def evaluate(self, ctx: Dict["Feature", Set[Address]], **kwargs) -> Result:
+    def evaluate(self, features: "capa.engine.FeatureSet", short_circuit=True) -> 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()))
+        return Result(self in features, self, [], locations=features.get(self, set()))
 
 
 class MatchedRule(Feature):
@@ -207,7 +207,7 @@ class Substring(String):
         super().__init__(value, description=description)
         self.value = value
 
-    def evaluate(self, ctx, short_circuit=True):
+    def evaluate(self, features: "capa.engine.FeatureSet", short_circuit=True):
         capa.perf.counters["evaluate.feature"] += 1
         capa.perf.counters["evaluate.feature.substring"] += 1
 
@@ -216,7 +216,7 @@ class Substring(String):
         matches: typing.DefaultDict[str, Set[Address]] = collections.defaultdict(set)
 
         assert isinstance(self.value, str)
-        for feature, locations in ctx.items():
+        for feature, locations in features.items():
             if not isinstance(feature, (String,)):
                 continue
 
@@ -227,7 +227,7 @@ class Substring(String):
             if self.value in feature.value:
                 matches[feature.value].update(locations)
                 if short_circuit:
-                    # we found one matching string, thats sufficient to match.
+                    # we found one matching string, that's sufficient to match.
                     # don't collect other matching strings in this mode.
                     break
 
@@ -299,7 +299,7 @@ class Regex(String):
                 f"invalid regular expression: {value} it should use Python syntax, try it at https://pythex.org"
             ) from exc
 
-    def evaluate(self, ctx, short_circuit=True):
+    def evaluate(self, features: "capa.engine.FeatureSet", short_circuit=True):
         capa.perf.counters["evaluate.feature"] += 1
         capa.perf.counters["evaluate.feature.regex"] += 1
 
@@ -307,7 +307,7 @@ class Regex(String):
         # will unique the locations later on.
         matches: typing.DefaultDict[str, Set[Address]] = collections.defaultdict(set)
 
-        for feature, locations in ctx.items():
+        for feature, locations in features.items():
             if not isinstance(feature, (String,)):
                 continue
 
@@ -322,7 +322,7 @@ class Regex(String):
             if self.re.search(feature.value):
                 matches[feature.value].update(locations)
                 if short_circuit:
-                    # we found one matching string, thats sufficient to match.
+                    # we found one matching string, that's sufficient to match.
                     # don't collect other matching strings in this mode.
                     break
 
@@ -384,12 +384,14 @@ class Bytes(Feature):
         super().__init__(value, description=description)
         self.value = value
 
-    def evaluate(self, ctx, **kwargs):
+    def evaluate(self, features: "capa.engine.FeatureSet", short_circuit=True):
+        assert isinstance(self.value, bytes)
+
         capa.perf.counters["evaluate.feature"] += 1
         capa.perf.counters["evaluate.feature.bytes"] += 1
+        capa.perf.counters["evaluate.feature.bytes." + str(len(self.value))] += 1
 
-        assert isinstance(self.value, bytes)
-        for feature, locations in ctx.items():
+        for feature, locations in features.items():
             if not isinstance(feature, (Bytes,)):
                 continue
 
@@ -407,9 +409,10 @@ class Bytes(Feature):
 # other candidates here: https://docs.microsoft.com/en-us/windows/win32/debug/pe-format#machine-types
 ARCH_I386 = "i386"
 ARCH_AMD64 = "amd64"
+ARCH_AARCH64 = "aarch64"
 # dotnet
 ARCH_ANY = "any"
-VALID_ARCH = (ARCH_I386, ARCH_AMD64, ARCH_ANY)
+VALID_ARCH = (ARCH_I386, ARCH_AMD64, ARCH_AARCH64, ARCH_ANY)
 
 
 class Arch(Feature):
@@ -421,10 +424,11 @@ class Arch(Feature):
 OS_WINDOWS = "windows"
 OS_LINUX = "linux"
 OS_MACOS = "macos"
+OS_ANDROID = "android"
 # 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})
+VALID_OS.update({OS_WINDOWS, OS_LINUX, OS_MACOS, OS_ANY, OS_ANDROID})
 # internal only, not to be used in rules
 OS_AUTO = "auto"
 
@@ -434,11 +438,11 @@ class OS(Feature):
         super().__init__(value, description=description)
         self.name = "os"
 
-    def evaluate(self, ctx, **kwargs):
+    def evaluate(self, features: "capa.engine.FeatureSet", short_circuit=True):
         capa.perf.counters["evaluate.feature"] += 1
         capa.perf.counters["evaluate.feature." + self.name] += 1
 
-        for feature, locations in ctx.items():
+        for feature, locations in features.items():
             if not isinstance(feature, (OS,)):
                 continue
 
@@ -458,6 +462,9 @@ FORMAT_AUTO = "auto"
 FORMAT_SC32 = "sc32"
 FORMAT_SC64 = "sc64"
 FORMAT_CAPE = "cape"
+FORMAT_DRAKVUF = "drakvuf"
+FORMAT_VMRAY = "vmray"
+FORMAT_BINEXPORT2 = "binexport2"
 FORMAT_FREEZE = "freeze"
 FORMAT_RESULT = "result"
 STATIC_FORMATS = {
@@ -468,9 +475,12 @@ STATIC_FORMATS = {
     FORMAT_DOTNET,
     FORMAT_FREEZE,
     FORMAT_RESULT,
+    FORMAT_BINEXPORT2,
 }
 DYNAMIC_FORMATS = {
     FORMAT_CAPE,
+    FORMAT_DRAKVUF,
+    FORMAT_VMRAY,
     FORMAT_FREEZE,
     FORMAT_RESULT,
 }
@@ -486,6 +496,6 @@ class Format(Feature):
 def is_global_feature(feature):
     """
     is this a feature that is extracted at every scope?
-    today, these are OS and arch features.
+    today, these are OS, arch, and format features.
     """
-    return isinstance(feature, (OS, Arch))
+    return isinstance(feature, (OS, Arch, Format))

capa/features/extractors/base_extractor.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2021 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
@@ -9,7 +9,9 @@
 import abc
 import hashlib
 import dataclasses
-from typing import Any, Dict, Tuple, Union, Iterator
+from copy import copy
+from types import MethodType
+from typing import Any, Set, Dict, Tuple, Union, Iterator
 from dataclasses import dataclass
 
 # TODO(williballenthin): use typing.TypeAlias directly when Python 3.9 is deprecated
@@ -75,7 +77,7 @@ class BBHandle:
 
 @dataclass
 class InsnHandle:
-    """reference to a instruction recognized by a feature extractor.
+    """reference to an instruction recognized by a feature extractor.
 
     Attributes:
         address: the address of the instruction address.
@@ -296,6 +298,22 @@ class StaticFeatureExtractor:
         raise NotImplementedError()
 
 
+def FunctionFilter(extractor: StaticFeatureExtractor, functions: Set) -> StaticFeatureExtractor:
+    original_get_functions = extractor.get_functions
+
+    def filtered_get_functions(self):
+        yield from (f for f in original_get_functions() if f.address in functions)
+
+    # we make a copy of the original extractor object and then update its get_functions() method with the decorated filter one.
+    # this is in order to preserve the original extractor object's get_functions() method, in case it is used elsewhere in the code.
+    # an example where this is important is in our testfiles where we may use the same extractor object with different tests,
+    # with some of these tests needing to install a functions filter on the extractor object.
+    new_extractor = copy(extractor)
+    new_extractor.get_functions = MethodType(filtered_get_functions, extractor)  # type: ignore
+
+    return new_extractor
+
+
 @dataclass
 class ProcessHandle:
     """
@@ -467,4 +485,20 @@ class DynamicFeatureExtractor:
         raise NotImplementedError()
 
 
+def ProcessFilter(extractor: DynamicFeatureExtractor, processes: Set) -> DynamicFeatureExtractor:
+    original_get_processes = extractor.get_processes
+
+    def filtered_get_processes(self):
+        yield from (f for f in original_get_processes() if f.address.pid in processes)
+
+    # we make a copy of the original extractor object and then update its get_processes() method with the decorated filter one.
+    # this is in order to preserve the original extractor object's get_processes() method, in case it is used elsewhere in the code.
+    # an example where this is important is in our testfiles where we may use the same extractor object with different tests,
+    # with some of these tests needing to install a processes filter on the extractor object.
+    new_extractor = copy(extractor)
+    new_extractor.get_processes = MethodType(filtered_get_processes, extractor)  # type: ignore
+
+    return new_extractor
+
+
 FeatureExtractor: TypeAlias = Union[StaticFeatureExtractor, DynamicFeatureExtractor]

capa/features/extractors/binexport2/__init__.py

@@ -0,0 +1,416 @@
+# Copyright (C) 2023 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.
+"""
+Proto files generated via protobuf v24.4:
+
+    protoc --python_out=. --mypy_out=. binexport2.proto
+
+from BinExport2 at 6916731d5f6693c4a4f0a052501fd3bd92cfd08b
+https://github.com/google/binexport/blob/6916731/binexport2.proto
+"""
+import io
+import hashlib
+import logging
+import contextlib
+from typing import Set, Dict, List, Tuple, Iterator
+from pathlib import Path
+from collections import defaultdict
+from dataclasses import dataclass
+
+from pefile import PE
+from elftools.elf.elffile import ELFFile
+
+import capa.features.common
+import capa.features.extractors.common
+import capa.features.extractors.binexport2.helpers
+from capa.features.extractors.binexport2.binexport2_pb2 import BinExport2
+
+logger = logging.getLogger(__name__)
+
+
+def get_binexport2(sample: Path) -> BinExport2:
+    be2: BinExport2 = BinExport2()
+    be2.ParseFromString(sample.read_bytes())
+    return be2
+
+
+def compute_common_prefix_length(m: str, n: str) -> int:
+    # ensure #m < #n
+    if len(n) < len(m):
+        m, n = n, m
+
+    for i, c in enumerate(m):
+        if n[i] != c:
+            return i
+
+    return len(m)
+
+
+def get_sample_from_binexport2(input_file: Path, be2: BinExport2, search_paths: List[Path]) -> Path:
+    """attempt to find the sample file, given a BinExport2 file.
+
+    searches in the same directory as the BinExport2 file, and then in search_paths.
+    """
+
+    def filename_similarity_key(p: Path) -> Tuple[int, str]:
+        # note closure over input_file.
+        # sort first by length of common prefix, then by name (for stability)
+        return (compute_common_prefix_length(p.name, input_file.name), p.name)
+
+    wanted_sha256: str = be2.meta_information.executable_id.lower()
+
+    input_directory: Path = input_file.parent
+    siblings: List[Path] = [p for p in input_directory.iterdir() if p.is_file()]
+    siblings.sort(key=filename_similarity_key, reverse=True)
+    for sibling in siblings:
+        # e.g. with open IDA files in the same directory on Windows
+        with contextlib.suppress(PermissionError):
+            if hashlib.sha256(sibling.read_bytes()).hexdigest().lower() == wanted_sha256:
+                return sibling
+
+    for search_path in search_paths:
+        candidates: List[Path] = [p for p in search_path.iterdir() if p.is_file()]
+        candidates.sort(key=filename_similarity_key, reverse=True)
+        for candidate in candidates:
+            with contextlib.suppress(PermissionError):
+                if hashlib.sha256(candidate.read_bytes()).hexdigest().lower() == wanted_sha256:
+                    return candidate
+
+    raise ValueError("cannot find sample, you may specify the path using the CAPA_SAMPLES_DIR environment variable")
+
+
+class BinExport2Index:
+    def __init__(self, be2: BinExport2):
+        self.be2: BinExport2 = be2
+
+        self.callers_by_vertex_index: Dict[int, List[int]] = defaultdict(list)
+        self.callees_by_vertex_index: Dict[int, List[int]] = defaultdict(list)
+
+        # note: flow graph != call graph (vertex)
+        self.flow_graph_index_by_address: Dict[int, int] = {}
+        self.flow_graph_address_by_index: Dict[int, int] = {}
+
+        # edges that come from the given basic block
+        self.source_edges_by_basic_block_index: Dict[int, List[BinExport2.FlowGraph.Edge]] = defaultdict(list)
+        # edges that end up at the given basic block
+        self.target_edges_by_basic_block_index: Dict[int, List[BinExport2.FlowGraph.Edge]] = defaultdict(list)
+
+        self.vertex_index_by_address: Dict[int, int] = {}
+
+        self.data_reference_index_by_source_instruction_index: Dict[int, List[int]] = defaultdict(list)
+        self.data_reference_index_by_target_address: Dict[int, List[int]] = defaultdict(list)
+        self.string_reference_index_by_source_instruction_index: Dict[int, List[int]] = defaultdict(list)
+
+        self.insn_address_by_index: Dict[int, int] = {}
+        self.insn_index_by_address: Dict[int, int] = {}
+        self.insn_by_address: Dict[int, BinExport2.Instruction] = {}
+
+        # must index instructions first
+        self._index_insn_addresses()
+        self._index_vertex_edges()
+        self._index_flow_graph_nodes()
+        self._index_flow_graph_edges()
+        self._index_call_graph_vertices()
+        self._index_data_references()
+        self._index_string_references()
+
+    def get_insn_address(self, insn_index: int) -> int:
+        assert insn_index in self.insn_address_by_index, f"insn must be indexed, missing {insn_index}"
+        return self.insn_address_by_index[insn_index]
+
+    def get_basic_block_address(self, basic_block_index: int) -> int:
+        basic_block: BinExport2.BasicBlock = self.be2.basic_block[basic_block_index]
+        first_instruction_index: int = next(self.instruction_indices(basic_block))
+        return self.get_insn_address(first_instruction_index)
+
+    def _index_vertex_edges(self):
+        for edge in self.be2.call_graph.edge:
+            if not edge.source_vertex_index:
+                continue
+            if not edge.target_vertex_index:
+                continue
+
+            self.callers_by_vertex_index[edge.target_vertex_index].append(edge.source_vertex_index)
+            self.callees_by_vertex_index[edge.source_vertex_index].append(edge.target_vertex_index)
+
+    def _index_flow_graph_nodes(self):
+        for flow_graph_index, flow_graph in enumerate(self.be2.flow_graph):
+            function_address: int = self.get_basic_block_address(flow_graph.entry_basic_block_index)
+            self.flow_graph_index_by_address[function_address] = flow_graph_index
+            self.flow_graph_address_by_index[flow_graph_index] = function_address
+
+    def _index_flow_graph_edges(self):
+        for flow_graph in self.be2.flow_graph:
+            for edge in flow_graph.edge:
+                if not edge.HasField("source_basic_block_index") or not edge.HasField("target_basic_block_index"):
+                    continue
+
+                self.source_edges_by_basic_block_index[edge.source_basic_block_index].append(edge)
+                self.target_edges_by_basic_block_index[edge.target_basic_block_index].append(edge)
+
+    def _index_call_graph_vertices(self):
+        for vertex_index, vertex in enumerate(self.be2.call_graph.vertex):
+            if not vertex.HasField("address"):
+                continue
+
+            vertex_address: int = vertex.address
+            self.vertex_index_by_address[vertex_address] = vertex_index
+
+    def _index_data_references(self):
+        for data_reference_index, data_reference in enumerate(self.be2.data_reference):
+            self.data_reference_index_by_source_instruction_index[data_reference.instruction_index].append(
+                data_reference_index
+            )
+            self.data_reference_index_by_target_address[data_reference.address].append(data_reference_index)
+
+    def _index_string_references(self):
+        for string_reference_index, string_reference in enumerate(self.be2.string_reference):
+            self.string_reference_index_by_source_instruction_index[string_reference.instruction_index].append(
+                string_reference_index
+            )
+
+    def _index_insn_addresses(self):
+        # see https://github.com/google/binexport/blob/39f6445c232bb5caf5c4a2a996de91dfa20c48e8/binexport.cc#L45
+        if len(self.be2.instruction) == 0:
+            return
+
+        assert self.be2.instruction[0].HasField("address"), "first insn must have explicit address"
+
+        addr: int = 0
+        next_addr: int = 0
+        for idx, insn in enumerate(self.be2.instruction):
+            if insn.HasField("address"):
+                addr = insn.address
+                next_addr = addr + len(insn.raw_bytes)
+            else:
+                addr = next_addr
+                next_addr += len(insn.raw_bytes)
+            self.insn_address_by_index[idx] = addr
+            self.insn_index_by_address[addr] = idx
+            self.insn_by_address[addr] = insn
+
+    @staticmethod
+    def instruction_indices(basic_block: BinExport2.BasicBlock) -> Iterator[int]:
+        """
+        For a given basic block, enumerate the instruction indices.
+        """
+        for index_range in basic_block.instruction_index:
+            if not index_range.HasField("end_index"):
+                yield index_range.begin_index
+                continue
+            else:
+                yield from range(index_range.begin_index, index_range.end_index)
+
+    def basic_block_instructions(
+        self, basic_block: BinExport2.BasicBlock
+    ) -> Iterator[Tuple[int, BinExport2.Instruction, int]]:
+        """
+        For a given basic block, enumerate the instruction indices,
+        the instruction instances, and their addresses.
+        """
+        for instruction_index in self.instruction_indices(basic_block):
+            instruction: BinExport2.Instruction = self.be2.instruction[instruction_index]
+            instruction_address: int = self.get_insn_address(instruction_index)
+
+            yield instruction_index, instruction, instruction_address
+
+    def get_function_name_by_vertex(self, vertex_index: int) -> str:
+        vertex: BinExport2.CallGraph.Vertex = self.be2.call_graph.vertex[vertex_index]
+        name: str = f"sub_{vertex.address:x}"
+        if vertex.HasField("mangled_name"):
+            name = vertex.mangled_name
+
+        if vertex.HasField("demangled_name"):
+            name = vertex.demangled_name
+
+        if vertex.HasField("library_index"):
+            library: BinExport2.Library = self.be2.library[vertex.library_index]
+            if library.HasField("name"):
+                name = f"{library.name}!{name}"
+
+        return name
+
+    def get_function_name_by_address(self, address: int) -> str:
+        if address not in self.vertex_index_by_address:
+            return ""
+
+        vertex_index: int = self.vertex_index_by_address[address]
+        return self.get_function_name_by_vertex(vertex_index)
+
+    def get_instruction_by_address(self, address: int) -> BinExport2.Instruction:
+        assert address in self.insn_by_address, f"address must be indexed, missing {address:x}"
+        return self.insn_by_address[address]
+
+
+class BinExport2Analysis:
+    def __init__(self, be2: BinExport2, idx: BinExport2Index, buf: bytes):
+        self.be2: BinExport2 = be2
+        self.idx: BinExport2Index = idx
+        self.buf: bytes = buf
+        self.base_address: int = 0
+        self.thunks: Dict[int, int] = {}
+
+        self._find_base_address()
+        self._compute_thunks()
+
+    def _find_base_address(self):
+        sections_with_perms: Iterator[BinExport2.Section] = filter(
+            lambda s: s.flag_r or s.flag_w or s.flag_x, self.be2.section
+        )
+        # assume the lowest address is the base address.
+        # this works as long as BinExport doesn't record other
+        # libraries mapped into memory.
+        self.base_address = min(s.address for s in sections_with_perms)
+
+        logger.debug("found base address: %x", self.base_address)
+
+    def _compute_thunks(self):
+        for addr, idx in self.idx.vertex_index_by_address.items():
+            vertex: BinExport2.CallGraph.Vertex = self.be2.call_graph.vertex[idx]
+            if not capa.features.extractors.binexport2.helpers.is_vertex_type(
+                vertex, BinExport2.CallGraph.Vertex.Type.THUNK
+            ):
+                continue
+
+            curr_idx: int = idx
+            for _ in range(capa.features.common.THUNK_CHAIN_DEPTH_DELTA):
+                thunk_callees: List[int] = self.idx.callees_by_vertex_index[curr_idx]
+                # if this doesn't hold, then it doesn't seem like this is a thunk,
+                # because either, len is:
+                #    0 and the thunk doesn't point to anything, or
+                #   >1 and the thunk may end up at many functions.
+                assert len(thunk_callees) == 1, f"thunk @ {hex(addr)} failed"
+
+                thunked_idx: int = thunk_callees[0]
+                thunked_vertex: BinExport2.CallGraph.Vertex = self.be2.call_graph.vertex[thunked_idx]
+
+                if not capa.features.extractors.binexport2.helpers.is_vertex_type(
+                    thunked_vertex, BinExport2.CallGraph.Vertex.Type.THUNK
+                ):
+                    assert thunked_vertex.HasField("address")
+
+                    self.thunks[addr] = thunked_vertex.address
+                    break
+
+                curr_idx = thunked_idx
+
+
+@dataclass
+class MemoryRegion:
+    # location of the bytes, potentially relative to a base address
+    address: int
+    buf: bytes
+
+    @property
+    def end(self) -> int:
+        return self.address + len(self.buf)
+
+    def contains(self, address: int) -> bool:
+        # note: address must be relative to any base address
+        return self.address <= address < self.end
+
+
+class ReadMemoryError(ValueError): ...
+
+
+class AddressNotMappedError(ReadMemoryError): ...
+
+
+@dataclass
+class AddressSpace:
+    base_address: int
+    memory_regions: Tuple[MemoryRegion, ...]
+
+    def read_memory(self, address: int, length: int) -> bytes:
+        rva: int = address - self.base_address
+        for region in self.memory_regions:
+            if region.contains(rva):
+                offset: int = rva - region.address
+                return region.buf[offset : offset + length]
+
+        raise AddressNotMappedError(address)
+
+    @classmethod
+    def from_pe(cls, pe: PE, base_address: int):
+        regions: List[MemoryRegion] = []
+        for section in pe.sections:
+            address: int = section.VirtualAddress
+            size: int = section.Misc_VirtualSize
+            buf: bytes = section.get_data()
+
+            if len(buf) != size:
+                # pad the section with NULLs
+                # assume page alignment is already handled.
+                # might need more hardening here.
+                buf += b"\x00" * (size - len(buf))
+
+            regions.append(MemoryRegion(address, buf))
+
+        return cls(base_address, tuple(regions))
+
+    @classmethod
+    def from_elf(cls, elf: ELFFile, base_address: int):
+        regions: List[MemoryRegion] = []
+
+        # ELF segments are for runtime data,
+        # ELF sections are for link-time data.
+        for segment in elf.iter_segments():
+            # assume p_align is consistent with addresses here.
+            # otherwise, should harden this loader.
+            segment_rva: int = segment.header.p_vaddr
+            segment_size: int = segment.header.p_memsz
+            segment_data: bytes = segment.data()
+
+            if len(segment_data) < segment_size:
+                # pad the section with NULLs
+                # assume page alignment is already handled.
+                # might need more hardening here.
+                segment_data += b"\x00" * (segment_size - len(segment_data))
+
+            regions.append(MemoryRegion(segment_rva, segment_data))
+
+        return cls(base_address, tuple(regions))
+
+    @classmethod
+    def from_buf(cls, buf: bytes, base_address: int):
+        if buf.startswith(capa.features.extractors.common.MATCH_PE):
+            pe: PE = PE(data=buf)
+            return cls.from_pe(pe, base_address)
+        elif buf.startswith(capa.features.extractors.common.MATCH_ELF):
+            elf: ELFFile = ELFFile(io.BytesIO(buf))
+            return cls.from_elf(elf, base_address)
+        else:
+            raise NotImplementedError("file format address space")
+
+
+@dataclass
+class AnalysisContext:
+    sample_bytes: bytes
+    be2: BinExport2
+    idx: BinExport2Index
+    analysis: BinExport2Analysis
+    address_space: AddressSpace
+
+
+@dataclass
+class FunctionContext:
+    ctx: AnalysisContext
+    flow_graph_index: int
+    format: Set[str]
+    os: Set[str]
+    arch: Set[str]
+
+
+@dataclass
+class BasicBlockContext:
+    basic_block_index: int
+
+
+@dataclass
+class InstructionContext:
+    instruction_index: int

capa/features/extractors/binexport2/arch/arm/helpers.py

@@ -0,0 +1,15 @@
+# Copyright (C) 2024 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 capa.features.extractors.binexport2.binexport2_pb2 import BinExport2
+
+
+def is_stack_register_expression(be2: BinExport2, expression: BinExport2.Expression) -> bool:
+    return bool(
+        expression and expression.type == BinExport2.Expression.REGISTER and expression.symbol.lower().endswith("sp")
+    )

capa/features/extractors/binexport2/arch/arm/insn.py

@@ -0,0 +1,155 @@
+# Copyright (C) 2024 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, Optional
+
+import capa.features.extractors.binexport2.helpers
+from capa.features.insn import MAX_STRUCTURE_SIZE, Number, Offset, OperandNumber, OperandOffset
+from capa.features.common import Feature, Characteristic
+from capa.features.address import Address
+from capa.features.extractors.binexport2 import FunctionContext, InstructionContext
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle
+from capa.features.extractors.binexport2.helpers import (
+    BinExport2InstructionPatternMatcher,
+    mask_immediate,
+    is_address_mapped,
+    get_instruction_mnemonic,
+    get_operand_register_expression,
+    get_operand_immediate_expression,
+)
+from capa.features.extractors.binexport2.binexport2_pb2 import BinExport2
+from capa.features.extractors.binexport2.arch.arm.helpers import is_stack_register_expression
+
+logger = logging.getLogger(__name__)
+
+
+def extract_insn_number_features(
+    fh: FunctionHandle, _bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+
+    be2: BinExport2 = fhi.ctx.be2
+
+    instruction_index: int = ii.instruction_index
+    instruction: BinExport2.Instruction = be2.instruction[instruction_index]
+
+    if len(instruction.operand_index) == 0:
+        # skip things like:
+        #   .text:0040116e leave
+        return
+
+    mnemonic: str = get_instruction_mnemonic(be2, instruction)
+
+    if mnemonic in ("add", "sub"):
+        assert len(instruction.operand_index) == 3
+
+        operand1_expression: Optional[BinExport2.Expression] = get_operand_register_expression(
+            be2, be2.operand[instruction.operand_index[1]]
+        )
+        if operand1_expression and is_stack_register_expression(be2, operand1_expression):
+            # skip things like:
+            # add x0,sp,#0x8
+            return
+
+    for i, operand_index in enumerate(instruction.operand_index):
+        operand: BinExport2.Operand = be2.operand[operand_index]
+
+        immediate_expression: Optional[BinExport2.Expression] = get_operand_immediate_expression(be2, operand)
+        if not immediate_expression:
+            continue
+
+        value: int = mask_immediate(fhi.arch, immediate_expression.immediate)
+        if is_address_mapped(be2, value):
+            continue
+
+        yield Number(value), ih.address
+        yield OperandNumber(i, value), ih.address
+
+        if mnemonic == "add" and i == 2:
+            if 0 < value < MAX_STRUCTURE_SIZE:
+                yield Offset(value), ih.address
+                yield OperandOffset(i, value), ih.address
+
+
+OFFSET_PATTERNS = BinExport2InstructionPatternMatcher.from_str(
+    """
+    ldr|ldrb|ldrh|ldrsb|ldrsh|ldrex|ldrd|str|strb|strh|strex|strd reg, [reg(not-stack),  #int]                                 ; capture #int
+    ldr|ldrb|ldrh|ldrsb|ldrsh|ldrex|ldrd|str|strb|strh|strex|strd reg, [reg(not-stack),  #int]!                                ; capture #int
+    ldr|ldrb|ldrh|ldrsb|ldrsh|ldrex|ldrd|str|strb|strh|strex|strd reg, [reg(not-stack)],        #int                           ; capture #int
+    ldp|ldpd|stp|stpd                                             reg, reg,                     [reg(not-stack), #int]         ; capture #int
+    ldp|ldpd|stp|stpd                                             reg, reg,                     [reg(not-stack), #int]!        ; capture #int
+    ldp|ldpd|stp|stpd                                             reg, reg,                     [reg(not-stack)],       #int   ; capture #int
+    """
+)
+
+
+def extract_insn_offset_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+
+    be2: BinExport2 = fhi.ctx.be2
+
+    match = OFFSET_PATTERNS.match_with_be2(be2, ii.instruction_index)
+    if not match:
+        return
+
+    value = match.expression.immediate
+
+    value = mask_immediate(fhi.arch, value)
+    if not is_address_mapped(be2, value):
+        value = capa.features.extractors.binexport2.helpers.twos_complement(fhi.arch, value)
+        yield Offset(value), ih.address
+        yield OperandOffset(match.operand_index, value), ih.address
+
+
+NZXOR_PATTERNS = BinExport2InstructionPatternMatcher.from_str(
+    """
+    eor reg, reg, reg
+    eor reg, reg, #int
+    """
+)
+
+
+def extract_insn_nzxor_characteristic_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+    be2: BinExport2 = fhi.ctx.be2
+
+    if NZXOR_PATTERNS.match_with_be2(be2, ii.instruction_index) is None:
+        return
+
+    instruction: BinExport2.Instruction = be2.instruction[ii.instruction_index]
+    # guaranteed to be simple int/reg operands
+    # so we don't have to realize the tree/list.
+    operands: List[BinExport2.Operand] = [be2.operand[operand_index] for operand_index in instruction.operand_index]
+
+    if operands[1] != operands[2]:
+        yield Characteristic("nzxor"), ih.address
+
+
+INDIRECT_CALL_PATTERNS = BinExport2InstructionPatternMatcher.from_str(
+    """
+    blx|bx|blr reg
+    """
+)
+
+
+def extract_function_indirect_call_characteristic_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+    be2: BinExport2 = fhi.ctx.be2
+
+    if INDIRECT_CALL_PATTERNS.match_with_be2(be2, ii.instruction_index) is not None:
+        yield Characteristic("indirect call"), ih.address

capa/features/extractors/binexport2/arch/intel/helpers.py

@@ -0,0 +1,135 @@
+# Copyright (C) 2024 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, Optional
+from dataclasses import dataclass
+
+from capa.features.extractors.binexport2.helpers import get_operand_expressions
+from capa.features.extractors.binexport2.binexport2_pb2 import BinExport2
+
+# 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: int = 0x40
+
+
+@dataclass
+class OperandPhraseInfo:
+    scale: Optional[BinExport2.Expression] = None
+    index: Optional[BinExport2.Expression] = None
+    base: Optional[BinExport2.Expression] = None
+    displacement: Optional[BinExport2.Expression] = None
+
+
+def get_operand_phrase_info(be2: BinExport2, operand: BinExport2.Operand) -> Optional[OperandPhraseInfo]:
+    # assume the following (see https://blog.yossarian.net/2020/06/13/How-x86_64-addresses-memory):
+    #
+    # Scale: A 2-bit constant factor
+    # Index: Any general purpose register
+    # Base: Any general purpose register
+    # Displacement: An integral offset
+
+    expressions: List[BinExport2.Expression] = get_operand_expressions(be2, operand)
+
+    # skip expression up to and including BinExport2.Expression.DEREFERENCE, assume caller
+    # has checked for BinExport2.Expression.DEREFERENCE
+    for i, expression in enumerate(expressions):
+        if expression.type == BinExport2.Expression.DEREFERENCE:
+            expressions = expressions[i + 1 :]
+            break
+
+    expression0: BinExport2.Expression
+    expression1: BinExport2.Expression
+    expression2: BinExport2.Expression
+    expression3: BinExport2.Expression
+    expression4: BinExport2.Expression
+
+    if len(expressions) == 1:
+        expression0 = expressions[0]
+
+        assert (
+            expression0.type == BinExport2.Expression.IMMEDIATE_INT
+            or expression0.type == BinExport2.Expression.REGISTER
+        )
+
+        if expression0.type == BinExport2.Expression.IMMEDIATE_INT:
+            # Displacement
+            return OperandPhraseInfo(displacement=expression0)
+        elif expression0.type == BinExport2.Expression.REGISTER:
+            # Base
+            return OperandPhraseInfo(base=expression0)
+
+    elif len(expressions) == 3:
+        expression0 = expressions[0]
+        expression1 = expressions[1]
+        expression2 = expressions[2]
+
+        assert expression0.type == BinExport2.Expression.REGISTER
+        assert expression1.type == BinExport2.Expression.OPERATOR
+        assert (
+            expression2.type == BinExport2.Expression.IMMEDIATE_INT
+            or expression2.type == BinExport2.Expression.REGISTER
+        )
+
+        if expression2.type == BinExport2.Expression.REGISTER:
+            # Base + Index
+            return OperandPhraseInfo(base=expression0, index=expression2)
+        elif expression2.type == BinExport2.Expression.IMMEDIATE_INT:
+            # Base + Displacement
+            return OperandPhraseInfo(base=expression0, displacement=expression2)
+
+    elif len(expressions) == 5:
+        expression0 = expressions[0]
+        expression1 = expressions[1]
+        expression2 = expressions[2]
+        expression3 = expressions[3]
+        expression4 = expressions[4]
+
+        assert expression0.type == BinExport2.Expression.REGISTER
+        assert expression1.type == BinExport2.Expression.OPERATOR
+        assert (
+            expression2.type == BinExport2.Expression.REGISTER
+            or expression2.type == BinExport2.Expression.IMMEDIATE_INT
+        )
+        assert expression3.type == BinExport2.Expression.OPERATOR
+        assert expression4.type == BinExport2.Expression.IMMEDIATE_INT
+
+        if expression1.symbol == "+" and expression3.symbol == "+":
+            # Base + Index + Displacement
+            return OperandPhraseInfo(base=expression0, index=expression2, displacement=expression4)
+        elif expression1.symbol == "+" and expression3.symbol == "*":
+            # Base + (Index * Scale)
+            return OperandPhraseInfo(base=expression0, index=expression2, scale=expression3)
+        elif expression1.symbol == "*" and expression3.symbol == "+":
+            # (Index * Scale) + Displacement
+            return OperandPhraseInfo(index=expression0, scale=expression2, displacement=expression3)
+        else:
+            raise NotImplementedError(expression1.symbol, expression3.symbol)
+
+    elif len(expressions) == 7:
+        expression0 = expressions[0]
+        expression1 = expressions[1]
+        expression2 = expressions[2]
+        expression3 = expressions[3]
+        expression4 = expressions[4]
+        expression5 = expressions[5]
+        expression6 = expressions[6]
+
+        assert expression0.type == BinExport2.Expression.REGISTER
+        assert expression1.type == BinExport2.Expression.OPERATOR
+        assert expression2.type == BinExport2.Expression.REGISTER
+        assert expression3.type == BinExport2.Expression.OPERATOR
+        assert expression4.type == BinExport2.Expression.IMMEDIATE_INT
+        assert expression5.type == BinExport2.Expression.OPERATOR
+        assert expression6.type == BinExport2.Expression.IMMEDIATE_INT
+
+        # Base + (Index * Scale) + Displacement
+        return OperandPhraseInfo(base=expression0, index=expression2, scale=expression4, displacement=expression6)
+
+    else:
+        raise NotImplementedError(len(expressions))
+
+    return None

capa/features/extractors/binexport2/arch/intel/insn.py

@@ -0,0 +1,248 @@
+# Copyright (C) 2024 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 capa.features.extractors.strings
+import capa.features.extractors.binexport2.helpers
+from capa.features.insn import MAX_STRUCTURE_SIZE, Number, Offset, OperandNumber, OperandOffset
+from capa.features.common import Feature, Characteristic
+from capa.features.address import Address
+from capa.features.extractors.binexport2 import BinExport2Index, FunctionContext, BasicBlockContext, InstructionContext
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle
+from capa.features.extractors.binexport2.helpers import (
+    BinExport2InstructionPatternMatcher,
+    mask_immediate,
+    is_address_mapped,
+    get_instruction_mnemonic,
+)
+from capa.features.extractors.binexport2.binexport2_pb2 import BinExport2
+from capa.features.extractors.binexport2.arch.intel.helpers import SECURITY_COOKIE_BYTES_DELTA
+
+logger = logging.getLogger(__name__)
+
+
+IGNORE_NUMBER_PATTERNS = BinExport2InstructionPatternMatcher.from_str(
+    """
+    ret  #int
+    retn #int
+    add  reg(stack), #int
+    sub  reg(stack), #int
+    """
+)
+
+NUMBER_PATTERNS = BinExport2InstructionPatternMatcher.from_str(
+    """
+    push #int0            ; capture #int0
+
+    # its a little tedious to enumerate all the address forms
+    # but at least we are explicit
+    cmp|and|or|test|mov|add|adc|sub|shl|shr|sal|sar  reg,                      #int0  ; capture #int0
+    cmp|and|or|test|mov|add|adc|sub|shl|shr|sal|sar [reg],                     #int0  ; capture #int0
+    cmp|and|or|test|mov|add|adc|sub|shl|shr|sal|sar [#int],                    #int0  ; capture #int0
+    cmp|and|or|test|mov|add|adc|sub|shl|shr|sal|sar [reg + #int],              #int0  ; capture #int0
+    cmp|and|or|test|mov|add|adc|sub|shl|shr|sal|sar [reg + reg + #int],        #int0  ; capture #int0
+    cmp|and|or|test|mov|add|adc|sub|shl|shr|sal|sar [reg + reg * #int],        #int0  ; capture #int0
+    cmp|and|or|test|mov|add|adc|sub|shl|shr|sal|sar [reg + reg * #int + #int], #int0  ; capture #int0
+
+    imul reg, reg, #int  ; capture #int
+    # note that int is first
+    cmp|test #int0, reg   ; capture #int0
+
+    # imagine reg is zero'd out, then this is like `mov reg, #int`
+    # which is not uncommon.
+    lea reg, [reg + #int]  ; capture #int
+    """
+)
+
+
+def extract_insn_number_features(
+    fh: FunctionHandle, _bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+
+    be2: BinExport2 = fhi.ctx.be2
+
+    if IGNORE_NUMBER_PATTERNS.match_with_be2(be2, ii.instruction_index):
+        return
+
+    match = NUMBER_PATTERNS.match_with_be2(be2, ii.instruction_index)
+    if not match:
+        return
+
+    value: int = mask_immediate(fhi.arch, match.expression.immediate)
+    if is_address_mapped(be2, value):
+        return
+
+    yield Number(value), ih.address
+    yield OperandNumber(match.operand_index, value), ih.address
+
+    instruction_index: int = ii.instruction_index
+    instruction: BinExport2.Instruction = be2.instruction[instruction_index]
+
+    mnemonic: str = get_instruction_mnemonic(be2, instruction)
+    if mnemonic.startswith("add"):
+        if 0 < value < MAX_STRUCTURE_SIZE:
+            yield Offset(value), ih.address
+            yield OperandOffset(match.operand_index, value), ih.address
+
+
+OFFSET_PATTERNS = BinExport2InstructionPatternMatcher.from_str(
+    """
+    mov|movzx|movsb|cmp [reg            +  reg * #int + #int0], #int  ; capture #int0
+    mov|movzx|movsb|cmp [reg            * #int + #int0],        #int  ; capture #int0
+    mov|movzx|movsb|cmp [reg            +  reg + #int0],        #int  ; capture #int0
+    mov|movzx|movsb|cmp [reg(not-stack) + #int0],               #int  ; capture #int0
+    mov|movzx|movsb|cmp [reg            +  reg * #int + #int0], reg   ; capture #int0
+    mov|movzx|movsb|cmp [reg            * #int + #int0],        reg   ; capture #int0
+    mov|movzx|movsb|cmp [reg            +  reg + #int0],        reg   ; capture #int0
+    mov|movzx|movsb|cmp [reg(not-stack) + #int0],               reg   ; capture #int0
+    mov|movzx|movsb|cmp|lea reg, [reg            +  reg * #int + #int0]  ; capture #int0
+    mov|movzx|movsb|cmp|lea reg, [reg            * #int + #int0]         ; capture #int0
+    mov|movzx|movsb|cmp|lea reg, [reg            +  reg + #int0]         ; capture #int0
+    mov|movzx|movsb|cmp|lea reg, [reg(not-stack) + #int0]                ; capture #int0
+    """
+)
+
+# these are patterns that access offset 0 from some pointer
+# (pointer is not the stack pointer).
+OFFSET_ZERO_PATTERNS = BinExport2InstructionPatternMatcher.from_str(
+    """
+    mov|movzx|movsb [reg(not-stack)], reg
+    mov|movzx|movsb [reg(not-stack)], #int
+    lea             reg,              [reg(not-stack)]
+    """
+)
+
+
+def extract_insn_offset_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+
+    be2: BinExport2 = fhi.ctx.be2
+
+    match = OFFSET_PATTERNS.match_with_be2(be2, ii.instruction_index)
+    if not match:
+        match = OFFSET_ZERO_PATTERNS.match_with_be2(be2, ii.instruction_index)
+        if not match:
+            return
+
+        yield Offset(0), ih.address
+        yield OperandOffset(match.operand_index, 0), ih.address
+
+    value = mask_immediate(fhi.arch, match.expression.immediate)
+    if is_address_mapped(be2, value):
+        return
+
+    value = capa.features.extractors.binexport2.helpers.twos_complement(fhi.arch, value, 32)
+    yield Offset(value), ih.address
+    yield OperandOffset(match.operand_index, value), ih.address
+
+
+def is_security_cookie(
+    fhi: FunctionContext,
+    bbi: BasicBlockContext,
+    instruction_address: int,
+    instruction: BinExport2.Instruction,
+) -> bool:
+    """
+    check if an instruction is related to security cookie checks.
+    """
+    be2: BinExport2 = fhi.ctx.be2
+    idx: BinExport2Index = fhi.ctx.idx
+
+    # security cookie check should use SP or BP
+    op1: BinExport2.Operand = be2.operand[instruction.operand_index[1]]
+    op1_exprs: List[BinExport2.Expression] = [be2.expression[expr_i] for expr_i in op1.expression_index]
+    if all(expr.symbol.lower() not in ("bp", "esp", "ebp", "rbp", "rsp") for expr in op1_exprs):
+        return False
+
+    # check_nzxor_security_cookie_delta
+    # if insn falls at the start of first entry block of the parent function.
+    flow_graph: BinExport2.FlowGraph = be2.flow_graph[fhi.flow_graph_index]
+    basic_block_index: int = bbi.basic_block_index
+    bb: BinExport2.BasicBlock = be2.basic_block[basic_block_index]
+    if flow_graph.entry_basic_block_index == basic_block_index:
+        first_addr: int = min((idx.insn_address_by_index[ir.begin_index] for ir in bb.instruction_index))
+        if instruction_address < first_addr + SECURITY_COOKIE_BYTES_DELTA:
+            return True
+    # or insn falls at the end before return in a terminal basic block.
+    if basic_block_index not in (e.source_basic_block_index for e in flow_graph.edge):
+        last_addr: int = max((idx.insn_address_by_index[ir.end_index - 1] for ir in bb.instruction_index))
+        if instruction_address > last_addr - SECURITY_COOKIE_BYTES_DELTA:
+            return True
+    return False
+
+
+NZXOR_PATTERNS = BinExport2InstructionPatternMatcher.from_str(
+    """
+    xor|xorpd|xorps|pxor reg, reg
+    xor|xorpd|xorps|pxor reg, #int
+    """
+)
+
+
+def extract_insn_nzxor_characteristic_features(
+    fh: FunctionHandle, bbh: 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.
+    """
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+
+    be2: BinExport2 = fhi.ctx.be2
+    idx: BinExport2Index = fhi.ctx.idx
+
+    if NZXOR_PATTERNS.match_with_be2(be2, ii.instruction_index) is None:
+        return
+
+    instruction: BinExport2.Instruction = be2.instruction[ii.instruction_index]
+    # guaranteed to be simple int/reg operands
+    # so we don't have to realize the tree/list.
+    operands: List[BinExport2.Operand] = [be2.operand[operand_index] for operand_index in instruction.operand_index]
+
+    if operands[0] == operands[1]:
+        return
+
+    instruction_address: int = idx.insn_address_by_index[ii.instruction_index]
+    if is_security_cookie(fhi, bbh.inner, instruction_address, instruction):
+        return
+
+    yield Characteristic("nzxor"), ih.address
+
+
+INDIRECT_CALL_PATTERNS = BinExport2InstructionPatternMatcher.from_str(
+    """
+    call|jmp reg0
+    call|jmp [reg + reg * #int + #int]
+    call|jmp [reg + reg * #int]
+    call|jmp [reg * #int + #int]
+    call|jmp [reg + reg + #int]
+    call|jmp [reg + #int]
+    call|jmp [reg]
+    """
+)
+
+
+def extract_function_indirect_call_characteristic_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+    be2: BinExport2 = fhi.ctx.be2
+
+    match = INDIRECT_CALL_PATTERNS.match_with_be2(be2, ii.instruction_index)
+    if match is None:
+        return
+
+    yield Characteristic("indirect call"), ih.address

capa/features/extractors/binexport2/basicblock.py

@@ -0,0 +1,40 @@
+# Copyright (C) 2023 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
+
+from capa.features.common import Feature, Characteristic
+from capa.features.address import Address, AbsoluteVirtualAddress
+from capa.features.basicblock import BasicBlock
+from capa.features.extractors.binexport2 import FunctionContext, BasicBlockContext
+from capa.features.extractors.base_extractor import BBHandle, FunctionHandle
+from capa.features.extractors.binexport2.binexport2_pb2 import BinExport2
+
+
+def extract_bb_tight_loop(fh: FunctionHandle, bbh: BBHandle) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+    bbi: BasicBlockContext = bbh.inner
+
+    idx = fhi.ctx.idx
+
+    basic_block_index: int = bbi.basic_block_index
+    target_edges: List[BinExport2.FlowGraph.Edge] = idx.target_edges_by_basic_block_index[basic_block_index]
+    if basic_block_index in (e.source_basic_block_index for e in target_edges):
+        basic_block_address: int = idx.get_basic_block_address(basic_block_index)
+        yield Characteristic("tight loop"), AbsoluteVirtualAddress(basic_block_address)
+
+
+def extract_features(fh: FunctionHandle, bbh: BBHandle) -> Iterator[Tuple[Feature, Address]]:
+    """extract basic block features"""
+    for bb_handler in BASIC_BLOCK_HANDLERS:
+        for feature, addr in bb_handler(fh, bbh):
+            yield feature, addr
+    yield BasicBlock(), bbh.address
+
+
+BASIC_BLOCK_HANDLERS = (extract_bb_tight_loop,)

capa/features/extractors/binexport2/binexport2_pb2.pyi

@@ -0,0 +1,784 @@
+"""
+@generated by mypy-protobuf.  Do not edit manually!
+isort:skip_file
+The representation is generic to accommodate various source architectures.
+In particular 32 and 64 bit versions of x86, ARM, PowerPC and MIPS have been
+tested.
+
+Multiple levels of deduping have been applied to make the format more compact
+and avoid redundant data duplication. Some of this due to hard-earned
+experience trying to cope with intentionally obfuscated malicious binaries.
+Note in particular that the same instruction may occur in multiple basic
+blocks and the same basic block in multiple functions (instruction and basic
+block sharing). Implemented naively, malware can use this to cause
+combinatorial explosion in memory usage, DOSing the analyst. This format
+should store every unique expression, mnemonic, operand, instruction and
+basic block only once instead of duplicating the information for every
+instance of it.
+
+This format does _not_ try to be 100% backwards compatible with the old
+version. In particular, we do not store IDA's comment types, making lossless
+porting of IDA comments impossible. We do however, store comments and
+expression substitutions, so porting the actual data is possible, just not
+the exact IDA type.
+
+While it would be more natural to use addresses when defining call graph and
+flow graph edges and other such references, it is more efficient to employ
+one more level of indirection and use indices into the basic block or
+function arrays instead. This is because addresses will usually use most of
+the available 64 bit space while indices will be much smaller and compress
+much better (less randomly distributed).
+
+We omit all fields that are set to their default value anyways. Note that
+this has two side effects:
+  - changing the defaults in this proto file will, in effect, change what's
+    read from disk
+  - the generated code has_* methods are somewhat less useful
+WARNING: We omit the defaults manually in the code writing the data. Do not
+         change the defaults here without changing the code!
+
+TODO(cblichmann): Link flow graphs to call graph nodes. The connection is
+                  there via the address, but tricky to extract.
+"""
+import builtins
+import collections.abc
+import google.protobuf.descriptor
+import google.protobuf.internal.containers
+import google.protobuf.internal.enum_type_wrapper
+import google.protobuf.message
+import sys
+import typing
+
+if sys.version_info >= (3, 10):
+    import typing as typing_extensions
+else:
+    import typing_extensions
+
+DESCRIPTOR: google.protobuf.descriptor.FileDescriptor
+
+@typing_extensions.final
+class BinExport2(google.protobuf.message.Message):
+    DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+    @typing_extensions.final
+    class Meta(google.protobuf.message.Message):
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        EXECUTABLE_NAME_FIELD_NUMBER: builtins.int
+        EXECUTABLE_ID_FIELD_NUMBER: builtins.int
+        ARCHITECTURE_NAME_FIELD_NUMBER: builtins.int
+        TIMESTAMP_FIELD_NUMBER: builtins.int
+        executable_name: builtins.str
+        """Input binary filename including file extension but excluding file path.
+        example: "insider_gcc.exe"
+        """
+        executable_id: builtins.str
+        """Application defined executable id. Often the SHA256 hash of the input
+        binary.
+        """
+        architecture_name: builtins.str
+        """Input architecture name, e.g. x86-32."""
+        timestamp: builtins.int
+        """When did this file get created? Unix time. This may be used for some
+        primitive versioning in case the file format ever changes.
+        """
+        def __init__(
+            self,
+            *,
+            executable_name: builtins.str | None = ...,
+            executable_id: builtins.str | None = ...,
+            architecture_name: builtins.str | None = ...,
+            timestamp: builtins.int | None = ...,
+        ) -> None: ...
+        def HasField(self, field_name: typing_extensions.Literal["architecture_name", b"architecture_name", "executable_id", b"executable_id", "executable_name", b"executable_name", "timestamp", b"timestamp"]) -> builtins.bool: ...
+        def ClearField(self, field_name: typing_extensions.Literal["architecture_name", b"architecture_name", "executable_id", b"executable_id", "executable_name", b"executable_name", "timestamp", b"timestamp"]) -> None: ...
+
+    @typing_extensions.final
+    class CallGraph(google.protobuf.message.Message):
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        @typing_extensions.final
+        class Vertex(google.protobuf.message.Message):
+            DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+            class _Type:
+                ValueType = typing.NewType("ValueType", builtins.int)
+                V: typing_extensions.TypeAlias = ValueType
+
+            class _TypeEnumTypeWrapper(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[BinExport2.CallGraph.Vertex._Type.ValueType], builtins.type):
+                DESCRIPTOR: google.protobuf.descriptor.EnumDescriptor
+                NORMAL: BinExport2.CallGraph.Vertex._Type.ValueType  # 0
+                """Regular function with full disassembly."""
+                LIBRARY: BinExport2.CallGraph.Vertex._Type.ValueType  # 1
+                """This function is a well known library function."""
+                IMPORTED: BinExport2.CallGraph.Vertex._Type.ValueType  # 2
+                """Imported from a dynamic link library (e.g. dll)."""
+                THUNK: BinExport2.CallGraph.Vertex._Type.ValueType  # 3
+                """A thunk function, forwarding its work via an unconditional jump."""
+                INVALID: BinExport2.CallGraph.Vertex._Type.ValueType  # 4
+                """An invalid function (a function that contained invalid code or was
+                considered invalid by some heuristics).
+                """
+
+            class Type(_Type, metaclass=_TypeEnumTypeWrapper): ...
+            NORMAL: BinExport2.CallGraph.Vertex.Type.ValueType  # 0
+            """Regular function with full disassembly."""
+            LIBRARY: BinExport2.CallGraph.Vertex.Type.ValueType  # 1
+            """This function is a well known library function."""
+            IMPORTED: BinExport2.CallGraph.Vertex.Type.ValueType  # 2
+            """Imported from a dynamic link library (e.g. dll)."""
+            THUNK: BinExport2.CallGraph.Vertex.Type.ValueType  # 3
+            """A thunk function, forwarding its work via an unconditional jump."""
+            INVALID: BinExport2.CallGraph.Vertex.Type.ValueType  # 4
+            """An invalid function (a function that contained invalid code or was
+            considered invalid by some heuristics).
+            """
+
+            ADDRESS_FIELD_NUMBER: builtins.int
+            TYPE_FIELD_NUMBER: builtins.int
+            MANGLED_NAME_FIELD_NUMBER: builtins.int
+            DEMANGLED_NAME_FIELD_NUMBER: builtins.int
+            LIBRARY_INDEX_FIELD_NUMBER: builtins.int
+            MODULE_INDEX_FIELD_NUMBER: builtins.int
+            address: builtins.int
+            """The function's entry point address. Messages need to be sorted, see
+            comment below on `vertex`.
+            """
+            type: global___BinExport2.CallGraph.Vertex.Type.ValueType
+            mangled_name: builtins.str
+            """If the function has a user defined, real name it will be given here.
+            main() is a proper name, sub_BAADF00D is not (auto generated dummy
+            name).
+            """
+            demangled_name: builtins.str
+            """Demangled name if the function is a mangled C++ function and we could
+            demangle it.
+            """
+            library_index: builtins.int
+            """If this is a library function, what is its index in library arrays."""
+            module_index: builtins.int
+            """If module name, such as class name for DEX files, is present - index in
+            module table.
+            """
+            def __init__(
+                self,
+                *,
+                address: builtins.int | None = ...,
+                type: global___BinExport2.CallGraph.Vertex.Type.ValueType | None = ...,
+                mangled_name: builtins.str | None = ...,
+                demangled_name: builtins.str | None = ...,
+                library_index: builtins.int | None = ...,
+                module_index: builtins.int | None = ...,
+            ) -> None: ...
+            def HasField(self, field_name: typing_extensions.Literal["address", b"address", "demangled_name", b"demangled_name", "library_index", b"library_index", "mangled_name", b"mangled_name", "module_index", b"module_index", "type", b"type"]) -> builtins.bool: ...
+            def ClearField(self, field_name: typing_extensions.Literal["address", b"address", "demangled_name", b"demangled_name", "library_index", b"library_index", "mangled_name", b"mangled_name", "module_index", b"module_index", "type", b"type"]) -> None: ...
+
+        @typing_extensions.final
+        class Edge(google.protobuf.message.Message):
+            DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+            SOURCE_VERTEX_INDEX_FIELD_NUMBER: builtins.int
+            TARGET_VERTEX_INDEX_FIELD_NUMBER: builtins.int
+            source_vertex_index: builtins.int
+            """source and target index into the vertex repeated field."""
+            target_vertex_index: builtins.int
+            def __init__(
+                self,
+                *,
+                source_vertex_index: builtins.int | None = ...,
+                target_vertex_index: builtins.int | None = ...,
+            ) -> None: ...
+            def HasField(self, field_name: typing_extensions.Literal["source_vertex_index", b"source_vertex_index", "target_vertex_index", b"target_vertex_index"]) -> builtins.bool: ...
+            def ClearField(self, field_name: typing_extensions.Literal["source_vertex_index", b"source_vertex_index", "target_vertex_index", b"target_vertex_index"]) -> None: ...
+
+        VERTEX_FIELD_NUMBER: builtins.int
+        EDGE_FIELD_NUMBER: builtins.int
+        @property
+        def vertex(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.CallGraph.Vertex]:
+            """vertices == functions in the call graph.
+            Important: Most downstream tooling (notably BinDiff), need these to be
+                       sorted by `Vertex::address` (ascending). For C++, the
+                       `BinExport2Writer` class enforces this invariant.
+            """
+        @property
+        def edge(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.CallGraph.Edge]:
+            """edges == calls in the call graph."""
+        def __init__(
+            self,
+            *,
+            vertex: collections.abc.Iterable[global___BinExport2.CallGraph.Vertex] | None = ...,
+            edge: collections.abc.Iterable[global___BinExport2.CallGraph.Edge] | None = ...,
+        ) -> None: ...
+        def ClearField(self, field_name: typing_extensions.Literal["edge", b"edge", "vertex", b"vertex"]) -> None: ...
+
+    @typing_extensions.final
+    class Expression(google.protobuf.message.Message):
+        """An operand consists of 1 or more expressions, linked together as a tree."""
+
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        class _Type:
+            ValueType = typing.NewType("ValueType", builtins.int)
+            V: typing_extensions.TypeAlias = ValueType
+
+        class _TypeEnumTypeWrapper(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[BinExport2.Expression._Type.ValueType], builtins.type):
+            DESCRIPTOR: google.protobuf.descriptor.EnumDescriptor
+            SYMBOL: BinExport2.Expression._Type.ValueType  # 1
+            IMMEDIATE_INT: BinExport2.Expression._Type.ValueType  # 2
+            IMMEDIATE_FLOAT: BinExport2.Expression._Type.ValueType  # 3
+            OPERATOR: BinExport2.Expression._Type.ValueType  # 4
+            REGISTER: BinExport2.Expression._Type.ValueType  # 5
+            SIZE_PREFIX: BinExport2.Expression._Type.ValueType  # 6
+            DEREFERENCE: BinExport2.Expression._Type.ValueType  # 7
+
+        class Type(_Type, metaclass=_TypeEnumTypeWrapper): ...
+        SYMBOL: BinExport2.Expression.Type.ValueType  # 1
+        IMMEDIATE_INT: BinExport2.Expression.Type.ValueType  # 2
+        IMMEDIATE_FLOAT: BinExport2.Expression.Type.ValueType  # 3
+        OPERATOR: BinExport2.Expression.Type.ValueType  # 4
+        REGISTER: BinExport2.Expression.Type.ValueType  # 5
+        SIZE_PREFIX: BinExport2.Expression.Type.ValueType  # 6
+        DEREFERENCE: BinExport2.Expression.Type.ValueType  # 7
+
+        TYPE_FIELD_NUMBER: builtins.int
+        SYMBOL_FIELD_NUMBER: builtins.int
+        IMMEDIATE_FIELD_NUMBER: builtins.int
+        PARENT_INDEX_FIELD_NUMBER: builtins.int
+        IS_RELOCATION_FIELD_NUMBER: builtins.int
+        type: global___BinExport2.Expression.Type.ValueType
+        """IMMEDIATE_INT is by far the most common type and thus we can save some
+        space by omitting it as the default.
+        """
+        symbol: builtins.str
+        """Symbol for this expression. Interpretation depends on type. Examples
+        include: "eax", "[", "+"
+        """
+        immediate: builtins.int
+        """If the expression can be interpreted as an integer value (IMMEDIATE_INT)
+        the value is given here.
+        """
+        parent_index: builtins.int
+        """The parent expression. Example expression tree for the second operand of:
+        mov eax, b4 [ebx + 12]
+        "b4" --- "[" --- "+" --- "ebx"
+                              \\  "12"
+        """
+        is_relocation: builtins.bool
+        """true if the expression has entry in relocation table"""
+        def __init__(
+            self,
+            *,
+            type: global___BinExport2.Expression.Type.ValueType | None = ...,
+            symbol: builtins.str | None = ...,
+            immediate: builtins.int | None = ...,
+            parent_index: builtins.int | None = ...,
+            is_relocation: builtins.bool | None = ...,
+        ) -> None: ...
+        def HasField(self, field_name: typing_extensions.Literal["immediate", b"immediate", "is_relocation", b"is_relocation", "parent_index", b"parent_index", "symbol", b"symbol", "type", b"type"]) -> builtins.bool: ...
+        def ClearField(self, field_name: typing_extensions.Literal["immediate", b"immediate", "is_relocation", b"is_relocation", "parent_index", b"parent_index", "symbol", b"symbol", "type", b"type"]) -> None: ...
+
+    @typing_extensions.final
+    class Operand(google.protobuf.message.Message):
+        """An instruction may have 0 or more operands."""
+
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        EXPRESSION_INDEX_FIELD_NUMBER: builtins.int
+        @property
+        def expression_index(self) -> google.protobuf.internal.containers.RepeatedScalarFieldContainer[builtins.int]:
+            """Contains all expressions constituting this operand. All expressions
+            should be linked into a single tree, i.e. there should only be one
+            expression in this list with parent_index == NULL and all others should
+            descend from that. Rendering order for expressions on the same tree level
+            (siblings) is implicitly given by the order they are referenced in this
+            repeated field.
+            Implicit: expression sequence
+            """
+        def __init__(
+            self,
+            *,
+            expression_index: collections.abc.Iterable[builtins.int] | None = ...,
+        ) -> None: ...
+        def ClearField(self, field_name: typing_extensions.Literal["expression_index", b"expression_index"]) -> None: ...
+
+    @typing_extensions.final
+    class Mnemonic(google.protobuf.message.Message):
+        """An instruction has exactly 1 mnemonic."""
+
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        NAME_FIELD_NUMBER: builtins.int
+        name: builtins.str
+        """Literal representation of the mnemonic, e.g.: "mov"."""
+        def __init__(
+            self,
+            *,
+            name: builtins.str | None = ...,
+        ) -> None: ...
+        def HasField(self, field_name: typing_extensions.Literal["name", b"name"]) -> builtins.bool: ...
+        def ClearField(self, field_name: typing_extensions.Literal["name", b"name"]) -> None: ...
+
+    @typing_extensions.final
+    class Instruction(google.protobuf.message.Message):
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        ADDRESS_FIELD_NUMBER: builtins.int
+        CALL_TARGET_FIELD_NUMBER: builtins.int
+        MNEMONIC_INDEX_FIELD_NUMBER: builtins.int
+        OPERAND_INDEX_FIELD_NUMBER: builtins.int
+        RAW_BYTES_FIELD_NUMBER: builtins.int
+        COMMENT_INDEX_FIELD_NUMBER: builtins.int
+        address: builtins.int
+        """This will only be filled for instructions that do not just flow from the
+        immediately preceding instruction. Regular instructions will have to
+        calculate their own address by adding raw_bytes.size() to the previous
+        instruction's address.
+        """
+        @property
+        def call_target(self) -> google.protobuf.internal.containers.RepeatedScalarFieldContainer[builtins.int]:
+            """If this is a call instruction and call targets could be determined
+            they'll be given here. Note that we may or may not have a flow graph for
+            the target and thus cannot use an index into the flow graph table here.
+            We could potentially use call graph nodes, but linking instructions to
+            the call graph directly does not seem a good choice.
+            """
+        mnemonic_index: builtins.int
+        """Index into the mnemonic array of strings. Used for de-duping the data.
+        The default value is used for the most common mnemonic in the executable.
+        """
+        @property
+        def operand_index(self) -> google.protobuf.internal.containers.RepeatedScalarFieldContainer[builtins.int]:
+            """Indices into the operand tree. On X86 this can be 0, 1 or 2 elements
+            long, 3 elements with VEX/EVEX.
+            Implicit: operand sequence
+            """
+        raw_bytes: builtins.bytes
+        """The unmodified input bytes corresponding to this instruction."""
+        @property
+        def comment_index(self) -> google.protobuf.internal.containers.RepeatedScalarFieldContainer[builtins.int]:
+            """Implicit: comment sequence"""
+        def __init__(
+            self,
+            *,
+            address: builtins.int | None = ...,
+            call_target: collections.abc.Iterable[builtins.int] | None = ...,
+            mnemonic_index: builtins.int | None = ...,
+            operand_index: collections.abc.Iterable[builtins.int] | None = ...,
+            raw_bytes: builtins.bytes | None = ...,
+            comment_index: collections.abc.Iterable[builtins.int] | None = ...,
+        ) -> None: ...
+        def HasField(self, field_name: typing_extensions.Literal["address", b"address", "mnemonic_index", b"mnemonic_index", "raw_bytes", b"raw_bytes"]) -> builtins.bool: ...
+        def ClearField(self, field_name: typing_extensions.Literal["address", b"address", "call_target", b"call_target", "comment_index", b"comment_index", "mnemonic_index", b"mnemonic_index", "operand_index", b"operand_index", "raw_bytes", b"raw_bytes"]) -> None: ...
+
+    @typing_extensions.final
+    class BasicBlock(google.protobuf.message.Message):
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        @typing_extensions.final
+        class IndexRange(google.protobuf.message.Message):
+            """This is a space optimization. The instructions for an individual basic
+            block will usually be in a continuous index range. Thus it is more
+            efficient to store the range instead of individual indices. However, this
+            does not hold true for all basic blocks, so we need to be able to store
+            multiple index ranges per block.
+            """
+
+            DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+            BEGIN_INDEX_FIELD_NUMBER: builtins.int
+            END_INDEX_FIELD_NUMBER: builtins.int
+            begin_index: builtins.int
+            """These work like begin and end iterators, i.e. the sequence is
+            [begin_index, end_index). If the sequence only contains a single
+            element end_index will be omitted.
+            """
+            end_index: builtins.int
+            def __init__(
+                self,
+                *,
+                begin_index: builtins.int | None = ...,
+                end_index: builtins.int | None = ...,
+            ) -> None: ...
+            def HasField(self, field_name: typing_extensions.Literal["begin_index", b"begin_index", "end_index", b"end_index"]) -> builtins.bool: ...
+            def ClearField(self, field_name: typing_extensions.Literal["begin_index", b"begin_index", "end_index", b"end_index"]) -> None: ...
+
+        INSTRUCTION_INDEX_FIELD_NUMBER: builtins.int
+        @property
+        def instruction_index(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.BasicBlock.IndexRange]:
+            """Implicit: instruction sequence"""
+        def __init__(
+            self,
+            *,
+            instruction_index: collections.abc.Iterable[global___BinExport2.BasicBlock.IndexRange] | None = ...,
+        ) -> None: ...
+        def ClearField(self, field_name: typing_extensions.Literal["instruction_index", b"instruction_index"]) -> None: ...
+
+    @typing_extensions.final
+    class FlowGraph(google.protobuf.message.Message):
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        @typing_extensions.final
+        class Edge(google.protobuf.message.Message):
+            DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+            class _Type:
+                ValueType = typing.NewType("ValueType", builtins.int)
+                V: typing_extensions.TypeAlias = ValueType
+
+            class _TypeEnumTypeWrapper(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[BinExport2.FlowGraph.Edge._Type.ValueType], builtins.type):
+                DESCRIPTOR: google.protobuf.descriptor.EnumDescriptor
+                CONDITION_TRUE: BinExport2.FlowGraph.Edge._Type.ValueType  # 1
+                CONDITION_FALSE: BinExport2.FlowGraph.Edge._Type.ValueType  # 2
+                UNCONDITIONAL: BinExport2.FlowGraph.Edge._Type.ValueType  # 3
+                SWITCH: BinExport2.FlowGraph.Edge._Type.ValueType  # 4
+
+            class Type(_Type, metaclass=_TypeEnumTypeWrapper): ...
+            CONDITION_TRUE: BinExport2.FlowGraph.Edge.Type.ValueType  # 1
+            CONDITION_FALSE: BinExport2.FlowGraph.Edge.Type.ValueType  # 2
+            UNCONDITIONAL: BinExport2.FlowGraph.Edge.Type.ValueType  # 3
+            SWITCH: BinExport2.FlowGraph.Edge.Type.ValueType  # 4
+
+            SOURCE_BASIC_BLOCK_INDEX_FIELD_NUMBER: builtins.int
+            TARGET_BASIC_BLOCK_INDEX_FIELD_NUMBER: builtins.int
+            TYPE_FIELD_NUMBER: builtins.int
+            IS_BACK_EDGE_FIELD_NUMBER: builtins.int
+            source_basic_block_index: builtins.int
+            """Source instruction will always be the last instruction of the source
+            basic block, target instruction the first instruction of the target
+            basic block.
+            """
+            target_basic_block_index: builtins.int
+            type: global___BinExport2.FlowGraph.Edge.Type.ValueType
+            is_back_edge: builtins.bool
+            """Indicates whether this is a loop edge as determined by Lengauer-Tarjan."""
+            def __init__(
+                self,
+                *,
+                source_basic_block_index: builtins.int | None = ...,
+                target_basic_block_index: builtins.int | None = ...,
+                type: global___BinExport2.FlowGraph.Edge.Type.ValueType | None = ...,
+                is_back_edge: builtins.bool | None = ...,
+            ) -> None: ...
+            def HasField(self, field_name: typing_extensions.Literal["is_back_edge", b"is_back_edge", "source_basic_block_index", b"source_basic_block_index", "target_basic_block_index", b"target_basic_block_index", "type", b"type"]) -> builtins.bool: ...
+            def ClearField(self, field_name: typing_extensions.Literal["is_back_edge", b"is_back_edge", "source_basic_block_index", b"source_basic_block_index", "target_basic_block_index", b"target_basic_block_index", "type", b"type"]) -> None: ...
+
+        BASIC_BLOCK_INDEX_FIELD_NUMBER: builtins.int
+        ENTRY_BASIC_BLOCK_INDEX_FIELD_NUMBER: builtins.int
+        EDGE_FIELD_NUMBER: builtins.int
+        @property
+        def basic_block_index(self) -> google.protobuf.internal.containers.RepeatedScalarFieldContainer[builtins.int]:
+            """Basic blocks are sorted by address."""
+        entry_basic_block_index: builtins.int
+        """The flow graph's entry point address is the first instruction of the
+        entry_basic_block.
+        """
+        @property
+        def edge(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.FlowGraph.Edge]: ...
+        def __init__(
+            self,
+            *,
+            basic_block_index: collections.abc.Iterable[builtins.int] | None = ...,
+            entry_basic_block_index: builtins.int | None = ...,
+            edge: collections.abc.Iterable[global___BinExport2.FlowGraph.Edge] | None = ...,
+        ) -> None: ...
+        def HasField(self, field_name: typing_extensions.Literal["entry_basic_block_index", b"entry_basic_block_index"]) -> builtins.bool: ...
+        def ClearField(self, field_name: typing_extensions.Literal["basic_block_index", b"basic_block_index", "edge", b"edge", "entry_basic_block_index", b"entry_basic_block_index"]) -> None: ...
+
+    @typing_extensions.final
+    class Reference(google.protobuf.message.Message):
+        """Generic reference class used for address comments (deprecated), string
+        references and expression substitutions. It allows referencing from an
+        instruction, operand, expression subtree tuple to a de-duped string in the
+        string table.
+        """
+
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        INSTRUCTION_INDEX_FIELD_NUMBER: builtins.int
+        INSTRUCTION_OPERAND_INDEX_FIELD_NUMBER: builtins.int
+        OPERAND_EXPRESSION_INDEX_FIELD_NUMBER: builtins.int
+        STRING_TABLE_INDEX_FIELD_NUMBER: builtins.int
+        instruction_index: builtins.int
+        """Index into the global instruction table."""
+        instruction_operand_index: builtins.int
+        """Index into the operand array local to an instruction."""
+        operand_expression_index: builtins.int
+        """Index into the expression array local to an operand."""
+        string_table_index: builtins.int
+        """Index into the global string table."""
+        def __init__(
+            self,
+            *,
+            instruction_index: builtins.int | None = ...,
+            instruction_operand_index: builtins.int | None = ...,
+            operand_expression_index: builtins.int | None = ...,
+            string_table_index: builtins.int | None = ...,
+        ) -> None: ...
+        def HasField(self, field_name: typing_extensions.Literal["instruction_index", b"instruction_index", "instruction_operand_index", b"instruction_operand_index", "operand_expression_index", b"operand_expression_index", "string_table_index", b"string_table_index"]) -> builtins.bool: ...
+        def ClearField(self, field_name: typing_extensions.Literal["instruction_index", b"instruction_index", "instruction_operand_index", b"instruction_operand_index", "operand_expression_index", b"operand_expression_index", "string_table_index", b"string_table_index"]) -> None: ...
+
+    @typing_extensions.final
+    class DataReference(google.protobuf.message.Message):
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        INSTRUCTION_INDEX_FIELD_NUMBER: builtins.int
+        ADDRESS_FIELD_NUMBER: builtins.int
+        instruction_index: builtins.int
+        """Index into the global instruction table."""
+        address: builtins.int
+        """Address being referred."""
+        def __init__(
+            self,
+            *,
+            instruction_index: builtins.int | None = ...,
+            address: builtins.int | None = ...,
+        ) -> None: ...
+        def HasField(self, field_name: typing_extensions.Literal["address", b"address", "instruction_index", b"instruction_index"]) -> builtins.bool: ...
+        def ClearField(self, field_name: typing_extensions.Literal["address", b"address", "instruction_index", b"instruction_index"]) -> None: ...
+
+    @typing_extensions.final
+    class Comment(google.protobuf.message.Message):
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        class _Type:
+            ValueType = typing.NewType("ValueType", builtins.int)
+            V: typing_extensions.TypeAlias = ValueType
+
+        class _TypeEnumTypeWrapper(google.protobuf.internal.enum_type_wrapper._EnumTypeWrapper[BinExport2.Comment._Type.ValueType], builtins.type):
+            DESCRIPTOR: google.protobuf.descriptor.EnumDescriptor
+            DEFAULT: BinExport2.Comment._Type.ValueType  # 0
+            """A regular instruction comment. Typically displayed next to the
+            instruction disassembly.
+            """
+            ANTERIOR: BinExport2.Comment._Type.ValueType  # 1
+            """A comment line that is typically displayed before (above) the
+            instruction it refers to.
+            """
+            POSTERIOR: BinExport2.Comment._Type.ValueType  # 2
+            """Like ANTERIOR, but a typically displayed after (below)."""
+            FUNCTION: BinExport2.Comment._Type.ValueType  # 3
+            """Similar to an ANTERIOR comment, but applies to the beginning of an
+            identified function. Programs displaying the proto may choose to render
+            these differently (e.g. above an inferred function signature).
+            """
+            ENUM: BinExport2.Comment._Type.ValueType  # 4
+            """Named constants, bitfields and similar."""
+            LOCATION: BinExport2.Comment._Type.ValueType  # 5
+            """Named locations, usually the target of a jump."""
+            GLOBAL_REFERENCE: BinExport2.Comment._Type.ValueType  # 6
+            """Data cross references."""
+            LOCAL_REFERENCE: BinExport2.Comment._Type.ValueType  # 7
+            """Local/stack variables."""
+
+        class Type(_Type, metaclass=_TypeEnumTypeWrapper): ...
+        DEFAULT: BinExport2.Comment.Type.ValueType  # 0
+        """A regular instruction comment. Typically displayed next to the
+        instruction disassembly.
+        """
+        ANTERIOR: BinExport2.Comment.Type.ValueType  # 1
+        """A comment line that is typically displayed before (above) the
+        instruction it refers to.
+        """
+        POSTERIOR: BinExport2.Comment.Type.ValueType  # 2
+        """Like ANTERIOR, but a typically displayed after (below)."""
+        FUNCTION: BinExport2.Comment.Type.ValueType  # 3
+        """Similar to an ANTERIOR comment, but applies to the beginning of an
+        identified function. Programs displaying the proto may choose to render
+        these differently (e.g. above an inferred function signature).
+        """
+        ENUM: BinExport2.Comment.Type.ValueType  # 4
+        """Named constants, bitfields and similar."""
+        LOCATION: BinExport2.Comment.Type.ValueType  # 5
+        """Named locations, usually the target of a jump."""
+        GLOBAL_REFERENCE: BinExport2.Comment.Type.ValueType  # 6
+        """Data cross references."""
+        LOCAL_REFERENCE: BinExport2.Comment.Type.ValueType  # 7
+        """Local/stack variables."""
+
+        INSTRUCTION_INDEX_FIELD_NUMBER: builtins.int
+        INSTRUCTION_OPERAND_INDEX_FIELD_NUMBER: builtins.int
+        OPERAND_EXPRESSION_INDEX_FIELD_NUMBER: builtins.int
+        STRING_TABLE_INDEX_FIELD_NUMBER: builtins.int
+        REPEATABLE_FIELD_NUMBER: builtins.int
+        TYPE_FIELD_NUMBER: builtins.int
+        instruction_index: builtins.int
+        """Index into the global instruction table. This is here to enable
+        comment processing without having to iterate over all instructions.
+        There is an N:M mapping of instructions to comments.
+        """
+        instruction_operand_index: builtins.int
+        """Index into the operand array local to an instruction."""
+        operand_expression_index: builtins.int
+        """Index into the expression array local to an operand, like in Reference.
+        This is not currently used, but allows to implement expression
+        substitutions.
+        """
+        string_table_index: builtins.int
+        """Index into the global string table."""
+        repeatable: builtins.bool
+        """Comment is propagated to all locations that reference the original
+        location.
+        """
+        type: global___BinExport2.Comment.Type.ValueType
+        def __init__(
+            self,
+            *,
+            instruction_index: builtins.int | None = ...,
+            instruction_operand_index: builtins.int | None = ...,
+            operand_expression_index: builtins.int | None = ...,
+            string_table_index: builtins.int | None = ...,
+            repeatable: builtins.bool | None = ...,
+            type: global___BinExport2.Comment.Type.ValueType | None = ...,
+        ) -> None: ...
+        def HasField(self, field_name: typing_extensions.Literal["instruction_index", b"instruction_index", "instruction_operand_index", b"instruction_operand_index", "operand_expression_index", b"operand_expression_index", "repeatable", b"repeatable", "string_table_index", b"string_table_index", "type", b"type"]) -> builtins.bool: ...
+        def ClearField(self, field_name: typing_extensions.Literal["instruction_index", b"instruction_index", "instruction_operand_index", b"instruction_operand_index", "operand_expression_index", b"operand_expression_index", "repeatable", b"repeatable", "string_table_index", b"string_table_index", "type", b"type"]) -> None: ...
+
+    @typing_extensions.final
+    class Section(google.protobuf.message.Message):
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        ADDRESS_FIELD_NUMBER: builtins.int
+        SIZE_FIELD_NUMBER: builtins.int
+        FLAG_R_FIELD_NUMBER: builtins.int
+        FLAG_W_FIELD_NUMBER: builtins.int
+        FLAG_X_FIELD_NUMBER: builtins.int
+        address: builtins.int
+        """Section start address."""
+        size: builtins.int
+        """Section size."""
+        flag_r: builtins.bool
+        """Read flag of the section, True when section is readable."""
+        flag_w: builtins.bool
+        """Write flag of the section, True when section is writable."""
+        flag_x: builtins.bool
+        """Execute flag of the section, True when section is executable."""
+        def __init__(
+            self,
+            *,
+            address: builtins.int | None = ...,
+            size: builtins.int | None = ...,
+            flag_r: builtins.bool | None = ...,
+            flag_w: builtins.bool | None = ...,
+            flag_x: builtins.bool | None = ...,
+        ) -> None: ...
+        def HasField(self, field_name: typing_extensions.Literal["address", b"address", "flag_r", b"flag_r", "flag_w", b"flag_w", "flag_x", b"flag_x", "size", b"size"]) -> builtins.bool: ...
+        def ClearField(self, field_name: typing_extensions.Literal["address", b"address", "flag_r", b"flag_r", "flag_w", b"flag_w", "flag_x", b"flag_x", "size", b"size"]) -> None: ...
+
+    @typing_extensions.final
+    class Library(google.protobuf.message.Message):
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        IS_STATIC_FIELD_NUMBER: builtins.int
+        LOAD_ADDRESS_FIELD_NUMBER: builtins.int
+        NAME_FIELD_NUMBER: builtins.int
+        is_static: builtins.bool
+        """If this library is statically linked."""
+        load_address: builtins.int
+        """Address where this library was loaded, 0 if unknown."""
+        name: builtins.str
+        """Name of the library (format is platform-dependent)."""
+        def __init__(
+            self,
+            *,
+            is_static: builtins.bool | None = ...,
+            load_address: builtins.int | None = ...,
+            name: builtins.str | None = ...,
+        ) -> None: ...
+        def HasField(self, field_name: typing_extensions.Literal["is_static", b"is_static", "load_address", b"load_address", "name", b"name"]) -> builtins.bool: ...
+        def ClearField(self, field_name: typing_extensions.Literal["is_static", b"is_static", "load_address", b"load_address", "name", b"name"]) -> None: ...
+
+    @typing_extensions.final
+    class Module(google.protobuf.message.Message):
+        DESCRIPTOR: google.protobuf.descriptor.Descriptor
+
+        NAME_FIELD_NUMBER: builtins.int
+        name: builtins.str
+        """Name, such as Java class name. Platform-dependent."""
+        def __init__(
+            self,
+            *,
+            name: builtins.str | None = ...,
+        ) -> None: ...
+        def HasField(self, field_name: typing_extensions.Literal["name", b"name"]) -> builtins.bool: ...
+        def ClearField(self, field_name: typing_extensions.Literal["name", b"name"]) -> None: ...
+
+    META_INFORMATION_FIELD_NUMBER: builtins.int
+    EXPRESSION_FIELD_NUMBER: builtins.int
+    OPERAND_FIELD_NUMBER: builtins.int
+    MNEMONIC_FIELD_NUMBER: builtins.int
+    INSTRUCTION_FIELD_NUMBER: builtins.int
+    BASIC_BLOCK_FIELD_NUMBER: builtins.int
+    FLOW_GRAPH_FIELD_NUMBER: builtins.int
+    CALL_GRAPH_FIELD_NUMBER: builtins.int
+    STRING_TABLE_FIELD_NUMBER: builtins.int
+    ADDRESS_COMMENT_FIELD_NUMBER: builtins.int
+    COMMENT_FIELD_NUMBER: builtins.int
+    STRING_REFERENCE_FIELD_NUMBER: builtins.int
+    EXPRESSION_SUBSTITUTION_FIELD_NUMBER: builtins.int
+    SECTION_FIELD_NUMBER: builtins.int
+    LIBRARY_FIELD_NUMBER: builtins.int
+    DATA_REFERENCE_FIELD_NUMBER: builtins.int
+    MODULE_FIELD_NUMBER: builtins.int
+    @property
+    def meta_information(self) -> global___BinExport2.Meta: ...
+    @property
+    def expression(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.Expression]: ...
+    @property
+    def operand(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.Operand]: ...
+    @property
+    def mnemonic(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.Mnemonic]: ...
+    @property
+    def instruction(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.Instruction]: ...
+    @property
+    def basic_block(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.BasicBlock]: ...
+    @property
+    def flow_graph(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.FlowGraph]: ...
+    @property
+    def call_graph(self) -> global___BinExport2.CallGraph: ...
+    @property
+    def string_table(self) -> google.protobuf.internal.containers.RepeatedScalarFieldContainer[builtins.str]: ...
+    @property
+    def address_comment(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.Reference]:
+        """No longer written. This is here so that BinDiff can work with older
+        BinExport files.
+        """
+    @property
+    def comment(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.Comment]:
+        """Rich comment index used for BinDiff's comment porting."""
+    @property
+    def string_reference(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.Reference]: ...
+    @property
+    def expression_substitution(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.Reference]: ...
+    @property
+    def section(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.Section]: ...
+    @property
+    def library(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.Library]: ...
+    @property
+    def data_reference(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.DataReference]: ...
+    @property
+    def module(self) -> google.protobuf.internal.containers.RepeatedCompositeFieldContainer[global___BinExport2.Module]: ...
+    def __init__(
+        self,
+        *,
+        meta_information: global___BinExport2.Meta | None = ...,
+        expression: collections.abc.Iterable[global___BinExport2.Expression] | None = ...,
+        operand: collections.abc.Iterable[global___BinExport2.Operand] | None = ...,
+        mnemonic: collections.abc.Iterable[global___BinExport2.Mnemonic] | None = ...,
+        instruction: collections.abc.Iterable[global___BinExport2.Instruction] | None = ...,
+        basic_block: collections.abc.Iterable[global___BinExport2.BasicBlock] | None = ...,
+        flow_graph: collections.abc.Iterable[global___BinExport2.FlowGraph] | None = ...,
+        call_graph: global___BinExport2.CallGraph | None = ...,
+        string_table: collections.abc.Iterable[builtins.str] | None = ...,
+        address_comment: collections.abc.Iterable[global___BinExport2.Reference] | None = ...,
+        comment: collections.abc.Iterable[global___BinExport2.Comment] | None = ...,
+        string_reference: collections.abc.Iterable[global___BinExport2.Reference] | None = ...,
+        expression_substitution: collections.abc.Iterable[global___BinExport2.Reference] | None = ...,
+        section: collections.abc.Iterable[global___BinExport2.Section] | None = ...,
+        library: collections.abc.Iterable[global___BinExport2.Library] | None = ...,
+        data_reference: collections.abc.Iterable[global___BinExport2.DataReference] | None = ...,
+        module: collections.abc.Iterable[global___BinExport2.Module] | None = ...,
+    ) -> None: ...
+    def HasField(self, field_name: typing_extensions.Literal["call_graph", b"call_graph", "meta_information", b"meta_information"]) -> builtins.bool: ...
+    def ClearField(self, field_name: typing_extensions.Literal["address_comment", b"address_comment", "basic_block", b"basic_block", "call_graph", b"call_graph", "comment", b"comment", "data_reference", b"data_reference", "expression", b"expression", "expression_substitution", b"expression_substitution", "flow_graph", b"flow_graph", "instruction", b"instruction", "library", b"library", "meta_information", b"meta_information", "mnemonic", b"mnemonic", "module", b"module", "operand", b"operand", "section", b"section", "string_reference", b"string_reference", "string_table", b"string_table"]) -> None: ...
+
+global___BinExport2 = BinExport2

capa/features/extractors/binexport2/extractor.py

@@ -0,0 +1,130 @@
+# Copyright (C) 2023 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 Set, List, Tuple, Iterator
+
+import capa.features.extractors.elf
+import capa.features.extractors.common
+import capa.features.extractors.binexport2.file
+import capa.features.extractors.binexport2.insn
+import capa.features.extractors.binexport2.helpers
+import capa.features.extractors.binexport2.function
+import capa.features.extractors.binexport2.basicblock
+from capa.features.common import OS, Arch, Format, Feature
+from capa.features.address import Address, AbsoluteVirtualAddress
+from capa.features.extractors.binexport2 import (
+    AddressSpace,
+    AnalysisContext,
+    BinExport2Index,
+    FunctionContext,
+    BasicBlockContext,
+    BinExport2Analysis,
+    InstructionContext,
+)
+from capa.features.extractors.base_extractor import (
+    BBHandle,
+    InsnHandle,
+    SampleHashes,
+    FunctionHandle,
+    StaticFeatureExtractor,
+)
+from capa.features.extractors.binexport2.binexport2_pb2 import BinExport2
+
+logger = logging.getLogger(__name__)
+
+
+class BinExport2FeatureExtractor(StaticFeatureExtractor):
+    def __init__(self, be2: BinExport2, buf: bytes):
+        super().__init__(hashes=SampleHashes.from_bytes(buf))
+        self.be2: BinExport2 = be2
+        self.buf: bytes = buf
+        self.idx: BinExport2Index = BinExport2Index(self.be2)
+        self.analysis: BinExport2Analysis = BinExport2Analysis(self.be2, self.idx, self.buf)
+        address_space: AddressSpace = AddressSpace.from_buf(buf, self.analysis.base_address)
+        self.ctx: AnalysisContext = AnalysisContext(self.buf, self.be2, self.idx, self.analysis, address_space)
+
+        self.global_features: List[Tuple[Feature, Address]] = []
+        self.global_features.extend(list(capa.features.extractors.common.extract_format(self.buf)))
+        self.global_features.extend(list(capa.features.extractors.common.extract_os(self.buf)))
+        self.global_features.extend(list(capa.features.extractors.common.extract_arch(self.buf)))
+
+        self.format: Set[str] = set()
+        self.os: Set[str] = set()
+        self.arch: Set[str] = set()
+
+        for feature, _ in self.global_features:
+            assert isinstance(feature.value, str)
+
+            if isinstance(feature, Format):
+                self.format.add(feature.value)
+            elif isinstance(feature, OS):
+                self.os.add(feature.value)
+            elif isinstance(feature, Arch):
+                self.arch.add(feature.value)
+            else:
+                raise ValueError("unexpected global feature: %s", feature)
+
+    def get_base_address(self) -> AbsoluteVirtualAddress:
+        return AbsoluteVirtualAddress(self.analysis.base_address)
+
+    def extract_global_features(self) -> Iterator[Tuple[Feature, Address]]:
+        yield from self.global_features
+
+    def extract_file_features(self) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.binexport2.file.extract_features(self.be2, self.buf)
+
+    def get_functions(self) -> Iterator[FunctionHandle]:
+        for flow_graph_index, flow_graph in enumerate(self.be2.flow_graph):
+            entry_basic_block_index: int = flow_graph.entry_basic_block_index
+            flow_graph_address: int = self.idx.get_basic_block_address(entry_basic_block_index)
+
+            vertex_idx: int = self.idx.vertex_index_by_address[flow_graph_address]
+            be2_vertex: BinExport2.CallGraph.Vertex = self.be2.call_graph.vertex[vertex_idx]
+
+            # skip thunks
+            if capa.features.extractors.binexport2.helpers.is_vertex_type(
+                be2_vertex, BinExport2.CallGraph.Vertex.Type.THUNK
+            ):
+                continue
+
+            yield FunctionHandle(
+                AbsoluteVirtualAddress(flow_graph_address),
+                inner=FunctionContext(self.ctx, flow_graph_index, self.format, self.os, self.arch),
+            )
+
+    def extract_function_features(self, fh: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.binexport2.function.extract_features(fh)
+
+    def get_basic_blocks(self, fh: FunctionHandle) -> Iterator[BBHandle]:
+        fhi: FunctionContext = fh.inner
+        flow_graph_index: int = fhi.flow_graph_index
+        flow_graph: BinExport2.FlowGraph = self.be2.flow_graph[flow_graph_index]
+
+        for basic_block_index in flow_graph.basic_block_index:
+            basic_block_address: int = self.idx.get_basic_block_address(basic_block_index)
+            yield BBHandle(
+                address=AbsoluteVirtualAddress(basic_block_address),
+                inner=BasicBlockContext(basic_block_index),
+            )
+
+    def extract_basic_block_features(self, fh: FunctionHandle, bbh: BBHandle) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.binexport2.basicblock.extract_features(fh, bbh)
+
+    def get_instructions(self, fh: FunctionHandle, bbh: BBHandle) -> Iterator[InsnHandle]:
+        bbi: BasicBlockContext = bbh.inner
+        basic_block: BinExport2.BasicBlock = self.be2.basic_block[bbi.basic_block_index]
+        for instruction_index, _, instruction_address in self.idx.basic_block_instructions(basic_block):
+            yield InsnHandle(
+                address=AbsoluteVirtualAddress(instruction_address),
+                inner=InstructionContext(instruction_index),
+            )
+
+    def extract_insn_features(
+        self, fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+    ) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.binexport2.insn.extract_features(fh, bbh, ih)

capa/features/extractors/binexport2/file.py

@@ -0,0 +1,80 @@
+# Copyright (C) 2023 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
+from typing import Tuple, Iterator
+
+import pefile
+from elftools.elf.elffile import ELFFile
+
+import capa.features.common
+import capa.features.extractors.common
+import capa.features.extractors.pefile
+import capa.features.extractors.elffile
+from capa.features.common import Feature
+from capa.features.address import Address
+from capa.features.extractors.binexport2.binexport2_pb2 import BinExport2
+
+logger = logging.getLogger(__name__)
+
+
+def extract_file_export_names(_be2: BinExport2, buf: bytes) -> Iterator[Tuple[Feature, Address]]:
+    if buf.startswith(capa.features.extractors.common.MATCH_PE):
+        pe: pefile.PE = pefile.PE(data=buf)
+        yield from capa.features.extractors.pefile.extract_file_export_names(pe)
+    elif buf.startswith(capa.features.extractors.common.MATCH_ELF):
+        elf: ELFFile = ELFFile(io.BytesIO(buf))
+        yield from capa.features.extractors.elffile.extract_file_export_names(elf)
+    else:
+        logger.warning("unsupported format")
+
+
+def extract_file_import_names(_be2: BinExport2, buf: bytes) -> Iterator[Tuple[Feature, Address]]:
+    if buf.startswith(capa.features.extractors.common.MATCH_PE):
+        pe: pefile.PE = pefile.PE(data=buf)
+        yield from capa.features.extractors.pefile.extract_file_import_names(pe)
+    elif buf.startswith(capa.features.extractors.common.MATCH_ELF):
+        elf: ELFFile = ELFFile(io.BytesIO(buf))
+        yield from capa.features.extractors.elffile.extract_file_import_names(elf)
+    else:
+        logger.warning("unsupported format")
+
+
+def extract_file_section_names(_be2: BinExport2, buf: bytes) -> Iterator[Tuple[Feature, Address]]:
+    if buf.startswith(capa.features.extractors.common.MATCH_PE):
+        pe: pefile.PE = pefile.PE(data=buf)
+        yield from capa.features.extractors.pefile.extract_file_section_names(pe)
+    elif buf.startswith(capa.features.extractors.common.MATCH_ELF):
+        elf: ELFFile = ELFFile(io.BytesIO(buf))
+        yield from capa.features.extractors.elffile.extract_file_section_names(elf)
+    else:
+        logger.warning("unsupported format")
+
+
+def extract_file_strings(_be2: BinExport2, buf: bytes) -> Iterator[Tuple[Feature, Address]]:
+    yield from capa.features.extractors.common.extract_file_strings(buf)
+
+
+def extract_file_format(_be2: BinExport2, buf: bytes) -> Iterator[Tuple[Feature, Address]]:
+    yield from capa.features.extractors.common.extract_format(buf)
+
+
+def extract_features(be2: BinExport2, buf: bytes) -> Iterator[Tuple[Feature, Address]]:
+    """extract file features"""
+    for file_handler in FILE_HANDLERS:
+        for feature, addr in file_handler(be2, buf):
+            yield feature, addr
+
+
+FILE_HANDLERS = (
+    extract_file_export_names,
+    extract_file_import_names,
+    extract_file_strings,
+    extract_file_section_names,
+    extract_file_format,
+)

capa/features/extractors/binexport2/function.py

@@ -0,0 +1,72 @@
+# Copyright (C) 2023 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
+
+from capa.features.file import FunctionName
+from capa.features.common import Feature, Characteristic
+from capa.features.address import Address, AbsoluteVirtualAddress
+from capa.features.extractors import loops
+from capa.features.extractors.binexport2 import BinExport2Index, FunctionContext
+from capa.features.extractors.base_extractor import FunctionHandle
+from capa.features.extractors.binexport2.binexport2_pb2 import BinExport2
+
+
+def extract_function_calls_to(fh: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+
+    be2: BinExport2 = fhi.ctx.be2
+    idx: BinExport2Index = fhi.ctx.idx
+
+    flow_graph_index: int = fhi.flow_graph_index
+    flow_graph_address: int = idx.flow_graph_address_by_index[flow_graph_index]
+    vertex_index: int = idx.vertex_index_by_address[flow_graph_address]
+
+    for caller_index in idx.callers_by_vertex_index[vertex_index]:
+        caller: BinExport2.CallGraph.Vertex = be2.call_graph.vertex[caller_index]
+        caller_address: int = caller.address
+        yield Characteristic("calls to"), AbsoluteVirtualAddress(caller_address)
+
+
+def extract_function_loop(fh: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+
+    be2: BinExport2 = fhi.ctx.be2
+
+    flow_graph_index: int = fhi.flow_graph_index
+    flow_graph: BinExport2.FlowGraph = be2.flow_graph[flow_graph_index]
+
+    edges: List[Tuple[int, int]] = []
+    for edge in flow_graph.edge:
+        edges.append((edge.source_basic_block_index, edge.target_basic_block_index))
+
+    if loops.has_loop(edges):
+        yield Characteristic("loop"), fh.address
+
+
+def extract_function_name(fh: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+
+    be2: BinExport2 = fhi.ctx.be2
+    idx: BinExport2Index = fhi.ctx.idx
+    flow_graph_index: int = fhi.flow_graph_index
+
+    flow_graph_address: int = idx.flow_graph_address_by_index[flow_graph_index]
+    vertex_index: int = idx.vertex_index_by_address[flow_graph_address]
+    vertex: BinExport2.CallGraph.Vertex = be2.call_graph.vertex[vertex_index]
+
+    if vertex.HasField("mangled_name"):
+        yield FunctionName(vertex.mangled_name), fh.address
+
+
+def extract_features(fh: FunctionHandle) -> Iterator[Tuple[Feature, Address]]:
+    for func_handler in FUNCTION_HANDLERS:
+        for feature, addr in func_handler(fh):
+            yield feature, addr
+
+
+FUNCTION_HANDLERS = (extract_function_calls_to, extract_function_loop, extract_function_name)

capa/features/extractors/binexport2/helpers.py

@@ -0,0 +1,650 @@
+# Copyright (C) 2024 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
+from typing import Set, Dict, List, Tuple, Union, Iterator, Optional
+from collections import defaultdict
+from dataclasses import dataclass
+
+import capa.features.extractors.helpers
+import capa.features.extractors.binexport2.helpers
+from capa.features.common import ARCH_I386, ARCH_AMD64, ARCH_AARCH64
+from capa.features.extractors.binexport2.binexport2_pb2 import BinExport2
+
+HAS_ARCH32 = {ARCH_I386}
+HAS_ARCH64 = {ARCH_AARCH64, ARCH_AMD64}
+
+HAS_ARCH_INTEL = {ARCH_I386, ARCH_AMD64}
+HAS_ARCH_ARM = {ARCH_AARCH64}
+
+
+def mask_immediate(arch: Set[str], immediate: int) -> int:
+    if arch & HAS_ARCH64:
+        immediate &= 0xFFFFFFFFFFFFFFFF
+    elif arch & HAS_ARCH32:
+        immediate &= 0xFFFFFFFF
+    return immediate
+
+
+def twos_complement(arch: Set[str], immediate: int, default: Optional[int] = None) -> int:
+    if default is not None:
+        return capa.features.extractors.helpers.twos_complement(immediate, default)
+    elif arch & HAS_ARCH64:
+        return capa.features.extractors.helpers.twos_complement(immediate, 64)
+    elif arch & HAS_ARCH32:
+        return capa.features.extractors.helpers.twos_complement(immediate, 32)
+    return immediate
+
+
+def is_address_mapped(be2: BinExport2, address: int) -> bool:
+    """return True if the given address is mapped"""
+    sections_with_perms: Iterator[BinExport2.Section] = filter(lambda s: s.flag_r or s.flag_w or s.flag_x, be2.section)
+    return any(section.address <= address < section.address + section.size for section in sections_with_perms)
+
+
+def is_vertex_type(vertex: BinExport2.CallGraph.Vertex, type_: BinExport2.CallGraph.Vertex.Type.ValueType) -> bool:
+    return vertex.HasField("type") and vertex.type == type_
+
+
+# internal to `build_expression_tree`
+# this is unstable: it is subject to change, so don't rely on it!
+def _prune_expression_tree_empty_shifts(
+    be2: BinExport2,
+    operand: BinExport2.Operand,
+    expression_tree: List[List[int]],
+    tree_index: int,
+):
+    expression_index = operand.expression_index[tree_index]
+    expression = be2.expression[expression_index]
+    children_tree_indexes: List[int] = expression_tree[tree_index]
+
+    if expression.type == BinExport2.Expression.OPERATOR:
+        if len(children_tree_indexes) == 0 and expression.symbol in ("lsl", "lsr"):
+            # Ghidra may emit superfluous lsl nodes with no children.
+            # https://github.com/mandiant/capa/pull/2340/files#r1750003919
+            # Which is maybe: https://github.com/NationalSecurityAgency/ghidra/issues/6821#issuecomment-2295394697
+            #
+            # Which seems to be as if the shift wasn't there (shift of #0)
+            # so we want to remove references to this node from any parent nodes.
+            for tree_node in expression_tree:
+                if tree_index in tree_node:
+                    tree_node.remove(tree_index)
+
+            return
+
+    for child_tree_index in children_tree_indexes:
+        _prune_expression_tree_empty_shifts(be2, operand, expression_tree, child_tree_index)
+
+
+# internal to `build_expression_tree`
+# this is unstable: it is subject to change, so don't rely on it!
+def _prune_expression_tree_empty_commas(
+    be2: BinExport2,
+    operand: BinExport2.Operand,
+    expression_tree: List[List[int]],
+    tree_index: int,
+):
+    expression_index = operand.expression_index[tree_index]
+    expression = be2.expression[expression_index]
+    children_tree_indexes: List[int] = expression_tree[tree_index]
+
+    if expression.type == BinExport2.Expression.OPERATOR:
+        if len(children_tree_indexes) == 1 and expression.symbol == ",":
+            # Due to the above pruning of empty LSL or LSR expressions,
+            # the parents might need to be fixed up.
+            #
+            # Specifically, if the pruned node was part of a comma list with two children,
+            # now there's only a single child, which renders as an extra comma,
+            # so we replace references to the comma node with the immediate child.
+            #
+            # A more correct way of doing this might be to walk up the parents and do fixups,
+            # but I'm not quite sure how to do this yet. Just do two passes right now.
+            child = children_tree_indexes[0]
+
+            for tree_node in expression_tree:
+                tree_node.index
+                if tree_index in tree_node:
+                    tree_node[tree_node.index(tree_index)] = child
+
+            return
+
+    for child_tree_index in children_tree_indexes:
+        _prune_expression_tree_empty_commas(be2, operand, expression_tree, child_tree_index)
+
+
+# internal to `build_expression_tree`
+# this is unstable: it is subject to change, so don't rely on it!
+def _prune_expression_tree(
+    be2: BinExport2,
+    operand: BinExport2.Operand,
+    expression_tree: List[List[int]],
+):
+    _prune_expression_tree_empty_shifts(be2, operand, expression_tree, 0)
+    _prune_expression_tree_empty_commas(be2, operand, expression_tree, 0)
+
+
+# this is unstable: it is subject to change, so don't rely on it!
+def _build_expression_tree(
+    be2: BinExport2,
+    operand: BinExport2.Operand,
+) -> List[List[int]]:
+    # The reconstructed expression tree layout, linking parent nodes to their children.
+    #
+    # There is one list of integers for each expression in the operand.
+    # These integers are indexes of other expressions in the same operand,
+    # which are the children of that expression.
+    #
+    # So:
+    #
+    #   [ [1, 3], [2], [], [4], [5], []]
+    #
+    # means the first expression has two children, at index 1 and 3,
+    # and the tree looks like:
+    #
+    #        0
+    #       / \
+    #      1   3
+    #      |   |
+    #      2   4
+    #          |
+    #          5
+    #
+    # Remember, these are the indices into the entries in operand.expression_index.
+    if len(operand.expression_index) == 0:
+        # Ghidra bug where empty operands (no expressions) may
+        # exist (see https://github.com/NationalSecurityAgency/ghidra/issues/6817)
+        return []
+
+    tree: List[List[int]] = []
+    for i, expression_index in enumerate(operand.expression_index):
+        children = []
+
+        # scan all subsequent expressions, looking for those that have parent_index == current.expression_index
+        for j, candidate_index in enumerate(operand.expression_index[i + 1 :]):
+            candidate = be2.expression[candidate_index]
+
+            if candidate.parent_index == expression_index:
+                children.append(i + j + 1)
+
+        tree.append(children)
+
+    _prune_expression_tree(be2, operand, tree)
+    _prune_expression_tree(be2, operand, tree)
+
+    return tree
+
+
+def _fill_operand_expression_list(
+    be2: BinExport2,
+    operand: BinExport2.Operand,
+    expression_tree: List[List[int]],
+    tree_index: int,
+    expression_list: List[BinExport2.Expression],
+):
+    """
+    Walk the given expression tree and collect the expression nodes in-order.
+    """
+    expression_index = operand.expression_index[tree_index]
+    expression = be2.expression[expression_index]
+    children_tree_indexes: List[int] = expression_tree[tree_index]
+
+    if expression.type == BinExport2.Expression.REGISTER:
+        assert len(children_tree_indexes) == 0
+        expression_list.append(expression)
+        return
+
+    elif expression.type == BinExport2.Expression.SYMBOL:
+        assert len(children_tree_indexes) <= 1
+        expression_list.append(expression)
+
+        if len(children_tree_indexes) == 0:
+            return
+        elif len(children_tree_indexes) == 1:
+            # like: v
+            # from: mov v0.D[0x1], x9
+            #           |
+            #           0
+            #           .
+            #           |
+            #           D
+            child_index = children_tree_indexes[0]
+            _fill_operand_expression_list(be2, operand, expression_tree, child_index, expression_list)
+            return
+        else:
+            raise NotImplementedError(len(children_tree_indexes))
+
+    elif expression.type == BinExport2.Expression.IMMEDIATE_INT:
+        assert len(children_tree_indexes) == 0
+        expression_list.append(expression)
+        return
+
+    elif expression.type == BinExport2.Expression.SIZE_PREFIX:
+        # like: b4
+        #
+        # We might want to use this occasionally, such as to disambiguate the
+        # size of MOVs into/out of memory. But I'm not sure when/where we need that yet.
+        #
+        # IDA spams this size prefix hint *everywhere*, so we can't rely on the exporter
+        # to provide it only when necessary.
+        assert len(children_tree_indexes) == 1
+        child_index = children_tree_indexes[0]
+        _fill_operand_expression_list(be2, operand, expression_tree, child_index, expression_list)
+        return
+
+    elif expression.type == BinExport2.Expression.OPERATOR:
+        if len(children_tree_indexes) == 1:
+            # prefix operator, like "ds:"
+            expression_list.append(expression)
+            child_index = children_tree_indexes[0]
+            _fill_operand_expression_list(be2, operand, expression_tree, child_index, expression_list)
+            return
+
+        elif len(children_tree_indexes) == 2:
+            # infix operator: like "+" in "ebp+10"
+            child_a = children_tree_indexes[0]
+            child_b = children_tree_indexes[1]
+            _fill_operand_expression_list(be2, operand, expression_tree, child_a, expression_list)
+            expression_list.append(expression)
+            _fill_operand_expression_list(be2, operand, expression_tree, child_b, expression_list)
+            return
+
+        elif len(children_tree_indexes) == 3:
+            # infix operator: like "+" in "ebp+ecx+10"
+            child_a = children_tree_indexes[0]
+            child_b = children_tree_indexes[1]
+            child_c = children_tree_indexes[2]
+            _fill_operand_expression_list(be2, operand, expression_tree, child_a, expression_list)
+            expression_list.append(expression)
+            _fill_operand_expression_list(be2, operand, expression_tree, child_b, expression_list)
+            expression_list.append(expression)
+            _fill_operand_expression_list(be2, operand, expression_tree, child_c, expression_list)
+            return
+
+        else:
+            raise NotImplementedError(len(children_tree_indexes))
+
+    elif expression.type == BinExport2.Expression.DEREFERENCE:
+        assert len(children_tree_indexes) == 1
+        expression_list.append(expression)
+
+        child_index = children_tree_indexes[0]
+        _fill_operand_expression_list(be2, operand, expression_tree, child_index, expression_list)
+        return
+
+    elif expression.type == BinExport2.Expression.IMMEDIATE_FLOAT:
+        raise NotImplementedError(expression.type)
+
+    else:
+        raise NotImplementedError(expression.type)
+
+
+def get_operand_expressions(be2: BinExport2, op: BinExport2.Operand) -> List[BinExport2.Expression]:
+    tree = _build_expression_tree(be2, op)
+
+    expressions: List[BinExport2.Expression] = []
+    _fill_operand_expression_list(be2, op, tree, 0, expressions)
+
+    return expressions
+
+
+def get_operand_register_expression(be2: BinExport2, operand: BinExport2.Operand) -> Optional[BinExport2.Expression]:
+    if len(operand.expression_index) == 1:
+        expression: BinExport2.Expression = be2.expression[operand.expression_index[0]]
+        if expression.type == BinExport2.Expression.REGISTER:
+            return expression
+    return None
+
+
+def get_operand_immediate_expression(be2: BinExport2, operand: BinExport2.Operand) -> Optional[BinExport2.Expression]:
+    if len(operand.expression_index) == 1:
+        # - type: IMMEDIATE_INT
+        #   immediate: 20588728364
+        #   parent_index: 0
+        expression: BinExport2.Expression = be2.expression[operand.expression_index[0]]
+        if expression.type == BinExport2.Expression.IMMEDIATE_INT:
+            return expression
+
+    elif len(operand.expression_index) == 2:
+        # from IDA, which provides a size hint for every operand,
+        # we get the following pattern for immediate constants:
+        #
+        # - type: SIZE_PREFIX
+        #   symbol: "b8"
+        # - type: IMMEDIATE_INT
+        #   immediate: 20588728364
+        #   parent_index: 0
+        expression0: BinExport2.Expression = be2.expression[operand.expression_index[0]]
+        expression1: BinExport2.Expression = be2.expression[operand.expression_index[1]]
+
+        if expression0.type == BinExport2.Expression.SIZE_PREFIX:
+            if expression1.type == BinExport2.Expression.IMMEDIATE_INT:
+                return expression1
+
+    return None
+
+
+def get_instruction_mnemonic(be2: BinExport2, instruction: BinExport2.Instruction) -> str:
+    return be2.mnemonic[instruction.mnemonic_index].name.lower()
+
+
+def get_instruction_operands(be2: BinExport2, instruction: BinExport2.Instruction) -> List[BinExport2.Operand]:
+    return [be2.operand[operand_index] for operand_index in instruction.operand_index]
+
+
+def split_with_delimiters(s: str, delimiters: Tuple[str, ...]) -> Iterator[str]:
+    """
+    Splits a string by any of the provided delimiter characters,
+    including the delimiters in the results.
+
+    Args:
+        string: The string to split.
+        delimiters: A string containing the characters to use as delimiters.
+    """
+    start = 0
+    for i, char in enumerate(s):
+        if char in delimiters:
+            yield s[start:i]
+            yield char
+            start = i + 1
+
+    if start < len(s):
+        yield s[start:]
+
+
+BinExport2OperandPattern = Union[str, Tuple[str, ...]]
+
+
+@dataclass
+class BinExport2InstructionPattern:
+    """
+    This describes a way to match disassembled instructions, with mnemonics and operands.
+
+    You can specify constraints on the instruction, via:
+      - the mnemonics, like "mov",
+      - number of operands, and
+      - format of each operand, "[reg, reg, #int]".
+
+    During matching, you can also capture a single element, to see its concrete value.
+    For example, given the pattern:
+
+        mov reg0, #int0  ; capture int0
+
+    and the instruction:
+
+        mov eax, 1
+
+    Then the capture will contain the immediate integer 1.
+
+    This matcher uses the BinExport2 data layout under the hood.
+    """
+
+    mnemonics: Tuple[str, ...]
+    operands: Tuple[Union[str, BinExport2OperandPattern], ...]
+    capture: Optional[str]
+
+    @classmethod
+    def from_str(cls, query: str):
+        """
+        Parse a pattern string into a Pattern instance.
+        The supported syntax is like this:
+
+            br      reg
+            br      reg                          ; capture reg
+            br      reg(stack)                   ; capture reg
+            br      reg(not-stack)               ; capture reg
+            mov     reg0, reg1                   ; capture reg0
+            adrp    reg, #int                    ; capture #int
+            add     reg, reg, #int               ; capture #int
+            ldr     reg0, [reg1]                 ; capture reg1
+            ldr|str reg, [reg, #int]             ; capture #int
+            ldr|str reg, [reg(stack), #int]      ; capture #int
+            ldr|str reg, [reg(not-stack), #int]  ; capture #int
+            ldr|str reg, [reg, #int]!            ; capture #int
+            ldr|str reg, [reg], #int             ; capture #int
+            ldp|stp reg, reg, [reg, #int]        ; capture #int
+            ldp|stp reg, reg, [reg, #int]!       ; capture #int
+            ldp|stp reg, reg, [reg], #int        ; capture #int
+        """
+        #
+        # The implementation of the parser here is obviously ugly.
+        # Its handwritten and probably fragile. But since we don't
+        # expect this to be widely used, its probably ok.
+        # Don't hesitate to rewrite this if it becomes more important.
+        #
+        # Note that this doesn't have to be very performant.
+        # We expect these patterns to be parsed once upfront and then reused
+        # (globally at the module level?) rather than within any loop.
+        #
+
+        pattern, _, comment = query.strip().partition(";")
+
+        # we don't support fs: yet
+        assert ":" not in pattern
+
+        # from "capture #int" to "#int"
+        if comment:
+            comment = comment.strip()
+            assert comment.startswith("capture ")
+            capture = comment[len("capture ") :]
+        else:
+            capture = None
+
+        # from "ldr|str ..." to ["ldr", "str"]
+        pattern = pattern.strip()
+        mnemonic, _, rest = pattern.partition(" ")
+        mnemonics = mnemonic.split("|")
+
+        operands: List[Union[str, Tuple[str, ...]]] = []
+        while rest:
+            rest = rest.strip()
+            if not rest.startswith("["):
+                # If its not a dereference, which looks like `[op, op, op, ...]`,
+                # then its a simple operand, which we can split by the next comma.
+                operand, _, rest = rest.partition(", ")
+                rest = rest.strip()
+                operands.append(operand)
+
+            else:
+                # This looks like a dereference, something like `[op, op, op, ...]`.
+                # Since these can't be nested, look for the next ] and then parse backwards.
+                deref_end = rest.index("]")
+                try:
+                    deref_end = rest.index(", ", deref_end)
+                    deref_end += len(", ")
+                except ValueError:
+                    deref = rest
+                    rest = ""
+                else:
+                    deref = rest[:deref_end]
+                    rest = rest[deref_end:]
+                    rest = rest.strip()
+                    deref = deref.rstrip(" ")
+                    deref = deref.rstrip(",")
+
+                # like: [reg, #int]!
+                has_postindex_writeback = deref.endswith("!")
+
+                deref = deref.rstrip("!")
+                deref = deref.rstrip("]")
+                deref = deref.lstrip("[")
+
+                parts = tuple(split_with_delimiters(deref, (",", "+", "*")))
+                parts = tuple(s.strip() for s in parts)
+
+                # emit operands in this order to match
+                # how BinExport2 expressions are flatted
+                # by get_operand_expressions
+                if has_postindex_writeback:
+                    operands.append(("!", "[") + parts)
+                else:
+                    operands.append(("[",) + parts)
+
+        for operand in operands:  # type: ignore
+            # Try to ensure we've parsed the operands correctly.
+            # This is just sanity checking.
+            for o in (operand,) if isinstance(operand, str) else operand:
+                # operands can look like:
+                #  - reg
+                #  - reg0
+                #  - reg(stack)
+                #  - reg0(stack)
+                #  - reg(not-stack)
+                #  - reg0(not-stack)
+                #  - #int
+                #  - #int0
+                # and a limited set of supported operators.
+                # use an inline regex so that its easy to read. not perf critical.
+                assert re.match(r"^(reg|#int)[0-9]?(\(stack\)|\(not-stack\))?$", o) or o in ("[", ",", "!", "+", "*")
+
+        return cls(tuple(mnemonics), tuple(operands), capture)
+
+    @dataclass
+    class MatchResult:
+        operand_index: int
+        expression_index: int
+        expression: BinExport2.Expression
+
+    def match(
+        self, mnemonic: str, operand_expressions: List[List[BinExport2.Expression]]
+    ) -> Optional["BinExport2InstructionPattern.MatchResult"]:
+        """
+        Match the given BinExport2 data against this pattern.
+
+        The BinExport2 expression tree must have been flattened, such as with
+        capa.features.extractors.binexport2.helpers.get_operand_expressions.
+
+        If there's a match, the captured Expression instance is returned.
+        Otherwise, you get None back.
+        """
+        if mnemonic not in self.mnemonics:
+            return None
+
+        if len(self.operands) != len(operand_expressions):
+            return None
+
+        captured = None
+
+        for operand_index, found_expressions in enumerate(operand_expressions):
+            wanted_expressions = self.operands[operand_index]
+
+            # from `"reg"` to `("reg", )`
+            if isinstance(wanted_expressions, str):
+                wanted_expressions = (wanted_expressions,)
+            assert isinstance(wanted_expressions, tuple)
+
+            if len(wanted_expressions) != len(found_expressions):
+                return None
+
+            for expression_index, (wanted_expression, found_expression) in enumerate(
+                zip(wanted_expressions, found_expressions)
+            ):
+                if wanted_expression.startswith("reg"):
+                    if found_expression.type != BinExport2.Expression.REGISTER:
+                        return None
+
+                    if wanted_expression.endswith(")"):
+                        if wanted_expression.endswith("(not-stack)"):
+                            # intel 64: rsp, esp, sp,
+                            # intel 32: ebp, ebp, bp
+                            # arm: sp
+                            register_name = found_expression.symbol.lower()
+                            if register_name in ("rsp", "esp", "sp", "rbp", "ebp", "bp"):
+                                return None
+
+                        elif wanted_expression.endswith("(stack)"):
+                            register_name = found_expression.symbol.lower()
+                            if register_name not in ("rsp", "esp", "sp", "rbp", "ebp", "bp"):
+                                return None
+
+                        else:
+                            raise ValueError("unexpected expression suffix", wanted_expression)
+
+                    if self.capture == wanted_expression:
+                        captured = BinExport2InstructionPattern.MatchResult(
+                            operand_index, expression_index, found_expression
+                        )
+
+                elif wanted_expression.startswith("#int"):
+                    if found_expression.type != BinExport2.Expression.IMMEDIATE_INT:
+                        return None
+
+                    if self.capture == wanted_expression:
+                        captured = BinExport2InstructionPattern.MatchResult(
+                            operand_index, expression_index, found_expression
+                        )
+
+                elif wanted_expression == "[":
+                    if found_expression.type != BinExport2.Expression.DEREFERENCE:
+                        return None
+
+                elif wanted_expression in (",", "!", "+", "*"):
+                    if found_expression.type != BinExport2.Expression.OPERATOR:
+                        return None
+
+                    if found_expression.symbol != wanted_expression:
+                        return None
+
+                else:
+                    raise ValueError(found_expression)
+
+        if captured:
+            return captured
+        else:
+            # There were no captures, so
+            # return arbitrary non-None expression
+            return BinExport2InstructionPattern.MatchResult(operand_index, expression_index, found_expression)
+
+
+class BinExport2InstructionPatternMatcher:
+    """Index and match a collection of instruction patterns."""
+
+    def __init__(self, queries: List[BinExport2InstructionPattern]):
+        self.queries = queries
+        # shard the patterns by (mnemonic, #operands)
+        self._index: Dict[Tuple[str, int], List[BinExport2InstructionPattern]] = defaultdict(list)
+
+        for query in queries:
+            for mnemonic in query.mnemonics:
+                self._index[(mnemonic.lower(), len(query.operands))].append(query)
+
+    @classmethod
+    def from_str(cls, patterns: str):
+        return cls(
+            [
+                BinExport2InstructionPattern.from_str(line)
+                for line in filter(
+                    lambda line: not line.startswith("#"), (line.strip() for line in patterns.split("\n"))
+                )
+            ]
+        )
+
+    def match(
+        self, mnemonic: str, operand_expressions: List[List[BinExport2.Expression]]
+    ) -> Optional[BinExport2InstructionPattern.MatchResult]:
+        queries = self._index.get((mnemonic.lower(), len(operand_expressions)), [])
+        for query in queries:
+            captured = query.match(mnemonic.lower(), operand_expressions)
+            if captured:
+                return captured
+
+        return None
+
+    def match_with_be2(
+        self, be2: BinExport2, instruction_index: int
+    ) -> Optional[BinExport2InstructionPattern.MatchResult]:
+        instruction: BinExport2.Instruction = be2.instruction[instruction_index]
+        mnemonic: str = get_instruction_mnemonic(be2, instruction)
+
+        if (mnemonic.lower(), len(instruction.operand_index)) not in self._index:
+            # verify that we might have a hit before we realize the operand expression list
+            return None
+
+        operands = []
+        for operand_index in instruction.operand_index:
+            operands.append(get_operand_expressions(be2, be2.operand[operand_index]))
+
+        return self.match(mnemonic, operands)

capa/features/extractors/binexport2/insn.py

@@ -0,0 +1,254 @@
+# Copyright (C) 2023 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 capa.features.extractors.helpers
+import capa.features.extractors.strings
+import capa.features.extractors.binexport2.helpers
+import capa.features.extractors.binexport2.arch.arm.insn
+import capa.features.extractors.binexport2.arch.intel.insn
+from capa.features.insn import API, Mnemonic
+from capa.features.common import Bytes, String, Feature, Characteristic
+from capa.features.address import Address, AbsoluteVirtualAddress
+from capa.features.extractors.binexport2 import (
+    AddressSpace,
+    AnalysisContext,
+    BinExport2Index,
+    FunctionContext,
+    ReadMemoryError,
+    BinExport2Analysis,
+    InstructionContext,
+)
+from capa.features.extractors.base_extractor import BBHandle, InsnHandle, FunctionHandle
+from capa.features.extractors.binexport2.helpers import HAS_ARCH_ARM, HAS_ARCH_INTEL
+from capa.features.extractors.binexport2.binexport2_pb2 import BinExport2
+
+logger = logging.getLogger(__name__)
+
+
+def extract_insn_api_features(fh: FunctionHandle, _bbh: BBHandle, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+
+    be2: BinExport2 = fhi.ctx.be2
+    be2_index: BinExport2Index = fhi.ctx.idx
+    be2_analysis: BinExport2Analysis = fhi.ctx.analysis
+    insn: BinExport2.Instruction = be2.instruction[ii.instruction_index]
+
+    for addr in insn.call_target:
+        addr = be2_analysis.thunks.get(addr, addr)
+
+        if addr not in be2_index.vertex_index_by_address:
+            # disassembler did not define function at address
+            logger.debug("0x%x is not a vertex", addr)
+            continue
+
+        vertex_idx: int = be2_index.vertex_index_by_address[addr]
+        vertex: BinExport2.CallGraph.Vertex = be2.call_graph.vertex[vertex_idx]
+
+        if not capa.features.extractors.binexport2.helpers.is_vertex_type(
+            vertex, BinExport2.CallGraph.Vertex.Type.IMPORTED
+        ):
+            continue
+
+        if not vertex.HasField("mangled_name"):
+            logger.debug("vertex %d does not have mangled_name", vertex_idx)
+            continue
+
+        api_name: str = vertex.mangled_name
+        for name in capa.features.extractors.helpers.generate_symbols("", api_name):
+            yield API(name), ih.address
+
+
+def extract_insn_number_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+
+    if fhi.arch & HAS_ARCH_INTEL:
+        yield from capa.features.extractors.binexport2.arch.intel.insn.extract_insn_number_features(fh, bbh, ih)
+    elif fhi.arch & HAS_ARCH_ARM:
+        yield from capa.features.extractors.binexport2.arch.arm.insn.extract_insn_number_features(fh, bbh, ih)
+
+
+def extract_insn_bytes_features(fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+
+    ctx: AnalysisContext = fhi.ctx
+    be2: BinExport2 = ctx.be2
+    idx: BinExport2Index = ctx.idx
+    address_space: AddressSpace = ctx.address_space
+
+    instruction_index: int = ii.instruction_index
+
+    if instruction_index in idx.string_reference_index_by_source_instruction_index:
+        # disassembler already identified string reference from instruction
+        return
+
+    reference_addresses: List[int] = []
+
+    if instruction_index in idx.data_reference_index_by_source_instruction_index:
+        for data_reference_index in idx.data_reference_index_by_source_instruction_index[instruction_index]:
+            data_reference: BinExport2.DataReference = be2.data_reference[data_reference_index]
+            data_reference_address: int = data_reference.address
+
+            if data_reference_address in idx.insn_address_by_index:
+                # appears to be code
+                continue
+
+            reference_addresses.append(data_reference_address)
+
+    for reference_address in reference_addresses:
+        try:
+            # if at end of segment then there might be an overrun here.
+            buf: bytes = address_space.read_memory(reference_address, 0x100)
+        except ReadMemoryError:
+            logger.debug("failed to read memory: 0x%x", reference_address)
+            continue
+
+        if capa.features.extractors.helpers.all_zeros(buf):
+            continue
+
+        is_string: bool = False
+
+        # note: we *always* break after the first iteration
+        for s in capa.features.extractors.strings.extract_ascii_strings(buf):
+            if s.offset != 0:
+                break
+
+            yield String(s.s), ih.address
+            is_string = True
+            break
+
+        # note: we *always* break after the first iteration
+        for s in capa.features.extractors.strings.extract_unicode_strings(buf):
+            if s.offset != 0:
+                break
+
+            yield String(s.s), ih.address
+            is_string = True
+            break
+
+        if not is_string:
+            yield Bytes(buf), ih.address
+
+
+def extract_insn_string_features(
+    fh: FunctionHandle, _bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+
+    be2: BinExport2 = fhi.ctx.be2
+    idx: BinExport2Index = fhi.ctx.idx
+
+    instruction_index: int = ii.instruction_index
+
+    if instruction_index in idx.string_reference_index_by_source_instruction_index:
+        for string_reference_index in idx.string_reference_index_by_source_instruction_index[instruction_index]:
+            string_reference: BinExport2.Reference = be2.string_reference[string_reference_index]
+            string_index: int = string_reference.string_table_index
+            string: str = be2.string_table[string_index]
+            yield String(string), ih.address
+
+
+def extract_insn_offset_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+
+    if fhi.arch & HAS_ARCH_INTEL:
+        yield from capa.features.extractors.binexport2.arch.intel.insn.extract_insn_offset_features(fh, bbh, ih)
+    elif fhi.arch & HAS_ARCH_ARM:
+        yield from capa.features.extractors.binexport2.arch.arm.insn.extract_insn_offset_features(fh, bbh, ih)
+
+
+def extract_insn_nzxor_characteristic_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+
+    if fhi.arch & HAS_ARCH_INTEL:
+        yield from capa.features.extractors.binexport2.arch.intel.insn.extract_insn_nzxor_characteristic_features(
+            fh, bbh, ih
+        )
+    elif fhi.arch & HAS_ARCH_ARM:
+        yield from capa.features.extractors.binexport2.arch.arm.insn.extract_insn_nzxor_characteristic_features(
+            fh, bbh, ih
+        )
+
+
+def extract_insn_mnemonic_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+
+    be2: BinExport2 = fhi.ctx.be2
+
+    instruction: BinExport2.Instruction = be2.instruction[ii.instruction_index]
+    mnemonic: BinExport2.Mnemonic = be2.mnemonic[instruction.mnemonic_index]
+    mnemonic_name: str = mnemonic.name.lower()
+    yield Mnemonic(mnemonic_name), ih.address
+
+
+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.
+    """
+    fhi: FunctionContext = fh.inner
+    ii: InstructionContext = ih.inner
+
+    be2: BinExport2 = fhi.ctx.be2
+
+    instruction: BinExport2.Instruction = be2.instruction[ii.instruction_index]
+    for call_target_address in instruction.call_target:
+        addr: AbsoluteVirtualAddress = AbsoluteVirtualAddress(call_target_address)
+        yield Characteristic("calls from"), addr
+
+        if fh.address == addr:
+            yield Characteristic("recursive call"), addr
+
+
+def extract_function_indirect_call_characteristic_features(
+    fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle
+) -> Iterator[Tuple[Feature, Address]]:
+    fhi: FunctionContext = fh.inner
+
+    if fhi.arch & HAS_ARCH_INTEL:
+        yield from capa.features.extractors.binexport2.arch.intel.insn.extract_function_indirect_call_characteristic_features(
+            fh, bbh, ih
+        )
+    elif fhi.arch & HAS_ARCH_ARM:
+        yield from capa.features.extractors.binexport2.arch.arm.insn.extract_function_indirect_call_characteristic_features(
+            fh, bbh, ih
+        )
+
+
+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, bbh, insn):
+            yield feature, ea
+
+
+INSTRUCTION_HANDLERS = (
+    extract_insn_api_features,
+    extract_insn_number_features,
+    extract_insn_bytes_features,
+    extract_insn_string_features,
+    extract_insn_offset_features,
+    extract_insn_nzxor_characteristic_features,
+    extract_insn_mnemonic_features,
+    extract_function_calls_from,
+    extract_function_indirect_call_characteristic_features,
+)

capa/features/extractors/binja/basicblock.py

@@ -7,17 +7,15 @@
 # See the License for the specific language governing permissions and limitations under the License.
 
 import string
-import struct
 from typing import Tuple, Iterator
 
-from binaryninja import Function, Settings
+from binaryninja import Function
 from binaryninja import BasicBlock as BinjaBasicBlock
 from binaryninja import (
     BinaryView,
     SymbolType,
     RegisterValueType,
     VariableSourceType,
-    MediumLevelILSetVar,
     MediumLevelILOperation,
     MediumLevelILBasicBlock,
     MediumLevelILInstruction,
@@ -29,11 +27,6 @@ from capa.features.basicblock import BasicBlock
 from capa.features.extractors.helpers import MIN_STACKSTRING_LEN
 from capa.features.extractors.base_extractor import BBHandle, FunctionHandle
 
-use_const_outline: bool = False
-settings: Settings = Settings()
-if settings.contains("analysis.outlining.builtins") and settings.get_bool("analysis.outlining.builtins"):
-    use_const_outline = True
-
 
 def get_printable_len_ascii(s: bytes) -> int:
     """Return string length if all operand bytes are ascii or utf16-le printable"""
@@ -65,7 +58,7 @@ def get_stack_string_len(f: Function, il: MediumLevelILInstruction) -> int:
 
     addr = target.value.value
     sym = bv.get_symbol_at(addr)
-    if not sym or sym.type != SymbolType.LibraryFunctionSymbol:
+    if not sym or sym.type not in [SymbolType.LibraryFunctionSymbol, SymbolType.SymbolicFunctionSymbol]:
         return 0
 
     if sym.name not in ["__builtin_strncpy", "__builtin_strcpy", "__builtin_wcscpy"]:
@@ -91,52 +84,6 @@ def get_stack_string_len(f: Function, il: MediumLevelILInstruction) -> int:
     return max(get_printable_len_ascii(bytes(s)), get_printable_len_wide(bytes(s)))
 
 
-def get_printable_len(il: MediumLevelILSetVar) -> int:
-    """Return string length if all operand bytes are ascii or utf16-le printable"""
-    width = il.dest.type.width
-    value = il.src.value.value
-
-    if width == 1:
-        chars = struct.pack("<B", value & 0xFF)
-    elif width == 2:
-        chars = struct.pack("<H", value & 0xFFFF)
-    elif width == 4:
-        chars = struct.pack("<I", value & 0xFFFFFFFF)
-    elif width == 8:
-        chars = struct.pack("<Q", value & 0xFFFFFFFFFFFFFFFF)
-    else:
-        return 0
-
-    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_: bytes):
-        if all(c == 0x00 for c in chars_[1::2]):
-            return is_printable_ascii(chars_[::2])
-
-    if is_printable_ascii(chars):
-        return width
-
-    if is_printable_utf16le(chars):
-        return width // 2
-
-    return 0
-
-
-def is_mov_imm_to_stack(il: MediumLevelILInstruction) -> bool:
-    """verify instruction moves immediate onto stack"""
-    if il.operation != MediumLevelILOperation.MLIL_SET_VAR:
-        return False
-
-    if il.src.operation != MediumLevelILOperation.MLIL_CONST:
-        return False
-
-    if il.dest.source_type != VariableSourceType.StackVariableSourceType:
-        return False
-
-    return True
-
-
 def bb_contains_stackstring(f: Function, bb: MediumLevelILBasicBlock) -> bool:
     """check basic block for stackstring indicators
 
@@ -144,14 +91,10 @@ def bb_contains_stackstring(f: Function, bb: MediumLevelILBasicBlock) -> bool:
     """
     count = 0
     for il in bb:
-        if use_const_outline:
-            count += get_stack_string_len(f, il)
-        else:
-            if is_mov_imm_to_stack(il):
-                count += get_printable_len(il)
+        count += get_stack_string_len(f, il)
+        if count > MIN_STACKSTRING_LEN:
+            return True
 
-    if count > MIN_STACKSTRING_LEN:
-        return True
     return False
 
 

capa/features/extractors/binja/extractor.py

@@ -28,7 +28,7 @@ from capa.features.extractors.base_extractor import (
 
 class BinjaFeatureExtractor(StaticFeatureExtractor):
     def __init__(self, bv: binja.BinaryView):
-        super().__init__(hashes=SampleHashes.from_bytes(bv.file.raw.read(0, len(bv.file.raw))))
+        super().__init__(hashes=SampleHashes.from_bytes(bv.file.raw.read(0, bv.file.raw.length)))
         self.bv = bv
         self.global_features: List[Tuple[Feature, Address]] = []
         self.global_features.extend(capa.features.extractors.binja.file.extract_file_format(self.bv))

capa/features/extractors/binja/file.py

@@ -5,8 +5,6 @@
 # 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
 from typing import Tuple, Iterator
 
 from binaryninja import Segment, BinaryView, SymbolType, SymbolBinding
@@ -20,67 +18,40 @@ from capa.features.address import NO_ADDRESS, Address, FileOffsetAddress, Absolu
 from capa.features.extractors.binja.helpers import read_c_string, unmangle_c_name
 
 
-def check_segment_for_pe(bv: BinaryView, seg: Segment) -> Iterator[Tuple[int, int]]:
-    """check segment for embedded PE
-
-    adapted for binja from:
-    https://github.com/vivisect/vivisect/blob/7be4037b1cecc4551b397f840405a1fc606f9b53/PE/carve.py#L19
-    """
-    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)
-    ]
-
-    todo = []
-    # If this is the first segment of the binary, skip the first bytes. Otherwise, there will always be a matched
-    # PE at the start of the binaryview.
-    start = seg.start
-    if bv.view_type == "PE" and start == bv.start:
+def check_segment_for_pe(bv: BinaryView, seg: Segment) -> Iterator[Tuple[Feature, Address]]:
+    """check segment for embedded PE"""
+    start = 0
+    if bv.view_type == "PE" and seg.start == bv.start:
+        # If this is the first segment of the binary, skip the first bytes.
+        # Otherwise, there will always be a matched PE at the start of the binaryview.
         start += 1
 
-    for mzx, pex, i in mz_xor:
-        for off, _ in bv.find_all_data(start, seg.end, mzx):
-            todo.append((off, mzx, pex, i))
+    buf = bv.read(seg.start, seg.length)
 
-    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 seg.end < (e_lfanew + 4):
-            continue
-
-        newoff = struct.unpack("<I", capa.features.extractors.helpers.xor_static(bv.read(e_lfanew, 4), i))[0]
-
-        peoff = off + newoff
-        if seg.end < (peoff + 2):
-            continue
-
-        if bv.read(peoff, 2) == pex:
-            yield off, i
+    for offset, _ in capa.features.extractors.helpers.carve_pe(buf, start):
+        yield Characteristic("embedded pe"), FileOffsetAddress(seg.start + offset)
 
 
 def extract_file_embedded_pe(bv: BinaryView) -> Iterator[Tuple[Feature, Address]]:
     """extract embedded PE features"""
     for seg in bv.segments:
-        for ea, _ in check_segment_for_pe(bv, seg):
-            yield Characteristic("embedded pe"), FileOffsetAddress(ea)
+        yield from check_segment_for_pe(bv, seg)
 
 
 def extract_file_export_names(bv: BinaryView) -> Iterator[Tuple[Feature, Address]]:
     """extract function exports"""
-    for sym in bv.get_symbols_of_type(SymbolType.FunctionSymbol):
+    for sym in bv.get_symbols_of_type(SymbolType.FunctionSymbol) + bv.get_symbols_of_type(SymbolType.DataSymbol):
         if sym.binding in [SymbolBinding.GlobalBinding, SymbolBinding.WeakBinding]:
             name = sym.short_name
-            yield Export(name), AbsoluteVirtualAddress(sym.address)
-            unmangled_name = unmangle_c_name(name)
-            if name != unmangled_name:
-                yield Export(unmangled_name), AbsoluteVirtualAddress(sym.address)
+            if name.startswith("__forwarder_name(") and name.endswith(")"):
+                yield Export(name[17:-1]), AbsoluteVirtualAddress(sym.address)
+                yield Characteristic("forwarded export"), AbsoluteVirtualAddress(sym.address)
+            else:
+                yield Export(name), AbsoluteVirtualAddress(sym.address)
+
+                unmangled_name = unmangle_c_name(name)
+                if name != unmangled_name:
+                    yield Export(unmangled_name), AbsoluteVirtualAddress(sym.address)
 
     for sym in bv.get_symbols_of_type(SymbolType.DataSymbol):
         if sym.binding not in [SymbolBinding.GlobalBinding]:

capa/features/extractors/binja/find_binja_api.py

@@ -5,31 +5,175 @@
 # 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 sys
+import logging
 import subprocess
+import importlib.util
+from typing import Optional
 from pathlib import Path
 
+logger = logging.getLogger(__name__)
+
+
 # When the script gets executed as a standalone executable (via PyInstaller), `import binaryninja` does not work because
 # we have excluded the binaryninja module in `pyinstaller.spec`. The trick here is to call the system Python and try
 # to find out the path of the binaryninja module that has been installed.
-# Note, including the binaryninja module in the `pyintaller.spec` would not work, since the binaryninja module tries to
+# Note, including the binaryninja module in the `pyinstaller.spec` would not work, since the binaryninja module tries to
 # find the binaryninja core e.g., `libbinaryninjacore.dylib`, using a relative path. And this does not work when the
 # binaryninja module is extracted by the PyInstaller.
-code = r"""
+CODE = r"""
 from pathlib import Path
 from importlib import util
 spec = util.find_spec('binaryninja')
 if spec is not None:
     if len(spec.submodule_search_locations) > 0:
-            path = Path(spec.submodule_search_locations[0])
-            # encode the path with utf8 then convert to hex, make sure it can be read and restored properly
-            print(str(path.parent).encode('utf8').hex())
+        path = Path(spec.submodule_search_locations[0])
+        # encode the path with utf8 then convert to hex, make sure it can be read and restored properly
+        print(str(path.parent).encode('utf8').hex())
 """
 
 
-def find_binja_path() -> Path:
-    raw_output = subprocess.check_output(["python", "-c", code]).decode("ascii").strip()
-    return Path(bytes.fromhex(raw_output).decode("utf8"))
+def find_binaryninja_path_via_subprocess() -> Optional[Path]:
+    raw_output = subprocess.check_output(["python", "-c", CODE]).decode("ascii").strip()
+    output = bytes.fromhex(raw_output).decode("utf8")
+    if not output.strip():
+        return None
+    return Path(output)
+
+
+def get_desktop_entry(name: str) -> Optional[Path]:
+    """
+    Find the path for the given XDG Desktop Entry name.
+
+    Like:
+
+        >> get_desktop_entry("com.vector35.binaryninja.desktop")
+        Path("~/.local/share/applications/com.vector35.binaryninja.desktop")
+    """
+    assert sys.platform in ("linux", "linux2")
+    assert name.endswith(".desktop")
+
+    data_dirs = os.environ.get("XDG_DATA_DIRS", "/usr/share") + f":{Path.home()}/.local/share"
+    for data_dir in data_dirs.split(":"):
+        applications = Path(data_dir) / "applications"
+        for application in applications.glob("*.desktop"):
+            if application.name == name:
+                return application
+
+    return None
+
+
+def get_binaryninja_path(desktop_entry: Path) -> Optional[Path]:
+    # from: Exec=/home/wballenthin/software/binaryninja/binaryninja %u
+    # to:        /home/wballenthin/software/binaryninja/
+    for line in desktop_entry.read_text(encoding="utf-8").splitlines():
+        if not line.startswith("Exec="):
+            continue
+
+        if not line.endswith("binaryninja %u"):
+            continue
+
+        binaryninja_path = Path(line[len("Exec=") : -len("binaryninja %u")])
+        if not binaryninja_path.exists():
+            return None
+
+        return binaryninja_path
+
+    return None
+
+
+def validate_binaryninja_path(binaryninja_path: Path) -> bool:
+    if not binaryninja_path:
+        return False
+
+    module_path = binaryninja_path / "python"
+    if not module_path.is_dir():
+        return False
+
+    if not (module_path / "binaryninja" / "__init__.py").is_file():
+        return False
+
+    return True
+
+
+def find_binaryninja() -> Optional[Path]:
+    binaryninja_path = find_binaryninja_path_via_subprocess()
+    if not binaryninja_path or not validate_binaryninja_path(binaryninja_path):
+        if sys.platform == "linux" or sys.platform == "linux2":
+            # ok
+            logger.debug("detected OS: linux")
+        elif sys.platform == "darwin":
+            logger.warning("unsupported platform to find Binary Ninja: %s", sys.platform)
+            return False
+        elif sys.platform == "win32":
+            logger.warning("unsupported platform to find Binary Ninja: %s", sys.platform)
+            return False
+        else:
+            logger.warning("unsupported platform to find Binary Ninja: %s", sys.platform)
+            return False
+
+        desktop_entry = get_desktop_entry("com.vector35.binaryninja.desktop")
+        if not desktop_entry:
+            logger.debug("failed to find Binary Ninja application")
+            return None
+        logger.debug("found Binary Ninja application: %s", desktop_entry)
+
+        binaryninja_path = get_binaryninja_path(desktop_entry)
+        if not binaryninja_path:
+            logger.debug("failed to determine Binary Ninja installation path")
+            return None
+
+        if not validate_binaryninja_path(binaryninja_path):
+            logger.debug("failed to validate Binary Ninja installation")
+            return None
+
+    logger.debug("found Binary Ninja installation: %s", binaryninja_path)
+
+    return binaryninja_path / "python"
+
+
+def is_binaryninja_installed() -> bool:
+    """Is the binaryninja module ready to import?"""
+    try:
+        return importlib.util.find_spec("binaryninja") is not None
+    except ModuleNotFoundError:
+        return False
+
+
+def has_binaryninja() -> bool:
+    if is_binaryninja_installed():
+        logger.debug("found installed Binary Ninja API")
+        return True
+
+    logger.debug("Binary Ninja API not installed, searching...")
+
+    binaryninja_path = find_binaryninja()
+    if not binaryninja_path:
+        logger.debug("failed to find Binary Ninja installation")
+
+    logger.debug("found Binary Ninja API: %s", binaryninja_path)
+    return binaryninja_path is not None
+
+
+def load_binaryninja() -> bool:
+    try:
+        import binaryninja
+
+        return True
+    except ImportError:
+        binaryninja_path = find_binaryninja()
+        if not binaryninja_path:
+            return False
+
+        sys.path.append(binaryninja_path.absolute().as_posix())
+        try:
+            import binaryninja  # noqa: F401 unused import
+
+            return True
+        except ImportError:
+            return False
 
 
 if __name__ == "__main__":
-    print(find_binja_path())
+    print(find_binaryninja_path_via_subprocess())

capa/features/extractors/cape/call.py

@@ -9,6 +9,7 @@
 import logging
 from typing import Tuple, Iterator
 
+import capa.features.extractors.helpers
 from capa.helpers import assert_never
 from capa.features.insn import API, Number
 from capa.features.common import String, Feature
@@ -50,7 +51,8 @@ def extract_call_features(ph: ProcessHandle, th: ThreadHandle, ch: CallHandle) -
         else:
             assert_never(value)
 
-    yield API(call.api), ch.address
+    for name in capa.features.extractors.helpers.generate_symbols("", call.api):
+        yield API(name), ch.address
 
 
 def extract_features(ph: ProcessHandle, th: ThreadHandle, ch: CallHandle) -> Iterator[Tuple[Feature, Address]]:

capa/features/extractors/cape/global_.py

@@ -48,7 +48,7 @@ def extract_format(report: CapeReport) -> Iterator[Tuple[Feature, Address]]:
     else:
         logger.warning("unknown file format, file command output: %s", report.target.file.type)
         raise ValueError(
-            "unrecognized file format from the CAPE report; output of file command: {report.target.file.type}"
+            f"unrecognized file format from the CAPE report; output of file command: {report.target.file.type}"
         )
 
 
@@ -73,7 +73,7 @@ def extract_os(report: CapeReport) -> Iterator[Tuple[Feature, Address]]:
         else:
             # if the operating system information is missing from the cape report, it's likely a bug
             logger.warning("unrecognized OS: %s", file_output)
-            raise ValueError("unrecognized OS from the CAPE report; output of file command: {file_output}")
+            raise ValueError(f"unrecognized OS from the CAPE report; output of file command: {file_output}")
     else:
         # the sample is shellcode
         logger.debug("unsupported file format, file command output: %s", file_output)

capa/features/extractors/cape/models.py

@@ -46,7 +46,7 @@ class FlexibleModel(BaseModel):
 
 
 # use this type to indicate that we won't model this data.
-# because its not relevant to our use in capa.
+# because it's not relevant to our use in capa.
 #
 # while its nice to have full coverage of the data shape,
 # it can easily change and break our parsing.
@@ -230,7 +230,7 @@ class File(FlexibleModel):
     sha1: str
     sha256: str
     sha512: str
-    sha3_384: str
+    sha3_384: Optional[str] = None
     ssdeep: str
     # unsure why this would ever be "False"
     tlsh: Optional[Union[str, bool]] = None
@@ -356,8 +356,8 @@ class Behavior(ExactModel):
     anomaly: List[str]
     encryptedbuffers: List[EncryptedBuffer]
     # these are small objects that describe atomic events,
-    # like file move, registery access.
-    # we'll detect the same with our API call analyis.
+    # like file move, registry access.
+    # we'll detect the same with our API call analysis.
     enhanced: Skip = None
 
 
@@ -398,7 +398,7 @@ class CapeReport(FlexibleModel):
     behavior: Behavior
 
     # post-processed results: payloads and extracted configs
-    CAPE: Optional[Cape] = None
+    CAPE: Optional[Union[Cape, List]] = None
     dropped: Optional[List[File]] = None
     procdump: Optional[List[ProcessFile]] = None
     procmemory: ListTODO

capa/features/extractors/common.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2021 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
@@ -75,7 +75,7 @@ def extract_format(buf: bytes) -> Iterator[Tuple[Feature, Address]]:
         #  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"))
+        logger.debug("unknown file format: %s", buf[:4].hex())
         return
 
 

capa/features/extractors/dnfile/extractor.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2022 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/dnfile/file.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2022 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/dnfile/function.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2022 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/dnfile/helpers.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2022 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
@@ -83,7 +83,7 @@ def read_dotnet_user_string(pe: dnfile.dnPE, token: StringToken) -> Optional[str
         return None
 
     try:
-        user_string: Optional[dnfile.stream.UserString] = pe.net.user_strings.get_us(token.rid)
+        user_string: Optional[dnfile.stream.UserString] = pe.net.user_strings.get(token.rid)
     except UnicodeDecodeError as e:
         logger.debug("failed to decode #US stream index 0x%08x (%s)", token.rid, e)
         return None
@@ -119,14 +119,14 @@ def get_dotnet_managed_imports(pe: dnfile.dnPE) -> Iterator[DnType]:
         access: Optional[str]
 
         # assume .NET imports starting with get_/set_ are used to access a property
-        if member_ref.Name.startswith("get_"):
+        member_ref_name: str = str(member_ref.Name)
+        if member_ref_name.startswith("get_"):
             access = FeatureAccess.READ
-        elif member_ref.Name.startswith("set_"):
+        elif member_ref_name.startswith("set_"):
             access = FeatureAccess.WRITE
         else:
             access = None
 
-        member_ref_name: str = member_ref.Name
         if member_ref_name.startswith(("get_", "set_")):
             # remove get_/set_ from MemberRef name
             member_ref_name = member_ref_name[4:]
@@ -212,7 +212,7 @@ def get_dotnet_managed_methods(pe: dnfile.dnPE) -> Iterator[DnType]:
             token: int = calculate_dotnet_token_value(method.table.number, method.row_index)
             access: Optional[str] = accessor_map.get(token)
 
-            method_name: str = method.row.Name
+            method_name: str = str(method.row.Name)
             if method_name.startswith(("get_", "set_")):
                 # remove get_/set_
                 method_name = method_name[4:]
@@ -289,8 +289,8 @@ def get_dotnet_unmanaged_imports(pe: dnfile.dnPE) -> Iterator[DnUnmanagedMethod]
             logger.debug("ImplMap[0x%X] ImportScope row is None", rid)
             module = ""
         else:
-            module = impl_map.ImportScope.row.Name
-        method: str = impl_map.ImportName
+            module = str(impl_map.ImportScope.row.Name)
+        method: str = str(impl_map.ImportName)
 
         member_forward_table: int
         if impl_map.MemberForwarded.table is None:
@@ -320,8 +320,11 @@ def get_dotnet_table_row(pe: dnfile.dnPE, table_index: int, row_index: int) -> O
     if row_index - 1 <= 0:
         return None
 
+    table: Optional[dnfile.base.ClrMetaDataTable] = pe.net.mdtables.tables.get(table_index)
+    if table is None:
+        return None
+
     try:
-        table = pe.net.mdtables.tables.get(table_index, [])
         return table[row_index - 1]
     except IndexError:
         return None
@@ -334,7 +337,7 @@ def resolve_nested_typedef_name(
 
     if index in nested_class_table:
         typedef_name = []
-        name = typedef.TypeName
+        name = str(typedef.TypeName)
 
         # Append the current typedef name
         typedef_name.append(name)
@@ -343,24 +346,24 @@ def resolve_nested_typedef_name(
             # Iterate through the typedef table to resolve the nested name
             table_row = get_dotnet_table_row(pe, dnfile.mdtable.TypeDef.number, nested_class_table[index])
             if table_row is None:
-                return typedef.TypeNamespace, tuple(typedef_name[::-1])
+                return str(typedef.TypeNamespace), tuple(typedef_name[::-1])
 
-            name = table_row.TypeName
+            name = str(table_row.TypeName)
             typedef_name.append(name)
             index = nested_class_table[index]
 
         # Document the root enclosing details
         table_row = get_dotnet_table_row(pe, dnfile.mdtable.TypeDef.number, nested_class_table[index])
         if table_row is None:
-            return typedef.TypeNamespace, tuple(typedef_name[::-1])
+            return str(typedef.TypeNamespace), tuple(typedef_name[::-1])
 
-        enclosing_name = table_row.TypeName
+        enclosing_name = str(table_row.TypeName)
         typedef_name.append(enclosing_name)
 
-        return table_row.TypeNamespace, tuple(typedef_name[::-1])
+        return str(table_row.TypeNamespace), tuple(typedef_name[::-1])
 
     else:
-        return typedef.TypeNamespace, (typedef.TypeName,)
+        return str(typedef.TypeNamespace), (str(typedef.TypeName),)
 
 
 def resolve_nested_typeref_name(
@@ -370,29 +373,29 @@ def resolve_nested_typeref_name(
     # If the ResolutionScope decodes to a typeRef type then it is nested
     if isinstance(typeref.ResolutionScope.table, dnfile.mdtable.TypeRef):
         typeref_name = []
-        name = typeref.TypeName
+        name = str(typeref.TypeName)
         # Not appending the current typeref name to avoid potential duplicate
 
         # Validate index
         table_row = get_dotnet_table_row(pe, dnfile.mdtable.TypeRef.number, index)
         if table_row is None:
-            return typeref.TypeNamespace, (typeref.TypeName,)
+            return str(typeref.TypeNamespace), (str(typeref.TypeName),)
 
         while isinstance(table_row.ResolutionScope.table, dnfile.mdtable.TypeRef):
             # Iterate through the typeref table to resolve the nested name
             typeref_name.append(name)
-            name = table_row.TypeName
+            name = str(table_row.TypeName)
             table_row = get_dotnet_table_row(pe, dnfile.mdtable.TypeRef.number, table_row.ResolutionScope.row_index)
             if table_row is None:
-                return typeref.TypeNamespace, tuple(typeref_name[::-1])
+                return str(typeref.TypeNamespace), tuple(typeref_name[::-1])
 
         # Document the root enclosing details
-        typeref_name.append(table_row.TypeName)
+        typeref_name.append(str(table_row.TypeName))
 
-        return table_row.TypeNamespace, tuple(typeref_name[::-1])
+        return str(table_row.TypeNamespace), tuple(typeref_name[::-1])
 
     else:
-        return typeref.TypeNamespace, (typeref.TypeName,)
+        return str(typeref.TypeNamespace), (str(typeref.TypeName),)
 
 
 def get_dotnet_nested_class_table_index(pe: dnfile.dnPE) -> Dict[int, int]:

capa/features/extractors/dnfile/insn.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2022 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/dnfile/types.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2022 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/dotnetfile.py

@@ -1,4 +1,4 @@
-# Copyright (C) 2023 Mandiant, Inc. All Rights Reserved.
+# Copyright (C) 2022 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
@@ -49,8 +49,8 @@ logger = logging.getLogger(__name__)
 
 
 def extract_file_format(**kwargs) -> Iterator[Tuple[Format, Address]]:
-    yield Format(FORMAT_PE), NO_ADDRESS
     yield Format(FORMAT_DOTNET), NO_ADDRESS
+    yield Format(FORMAT_PE), NO_ADDRESS
 
 
 def extract_file_import_names(pe: dnfile.dnPE, **kwargs) -> Iterator[Tuple[Import, Address]]:
@@ -78,12 +78,12 @@ def extract_file_namespace_features(pe: dnfile.dnPE, **kwargs) -> Iterator[Tuple
     for _, typedef in iter_dotnet_table(pe, dnfile.mdtable.TypeDef.number):
         # emit internal .NET namespaces
         assert isinstance(typedef, dnfile.mdtable.TypeDefRow)
-        namespaces.add(typedef.TypeNamespace)
+        namespaces.add(str(typedef.TypeNamespace))
 
     for _, typeref in iter_dotnet_table(pe, dnfile.mdtable.TypeRef.number):
         # emit external .NET namespaces
         assert isinstance(typeref, dnfile.mdtable.TypeRefRow)
-        namespaces.add(typeref.TypeNamespace)
+        namespaces.add(str(typeref.TypeNamespace))
 
     # namespaces may be empty, discard
     namespaces.discard("")

capa/features/extractors/drakvuf/call.py

@@ -0,0 +1,58 @@
+# Copyright (C) 2024 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 Tuple, Iterator
+
+import capa.features.extractors.helpers
+from capa.features.insn import API, Number
+from capa.features.common import String, Feature
+from capa.features.address import Address
+from capa.features.extractors.base_extractor import CallHandle, ThreadHandle, ProcessHandle
+from capa.features.extractors.drakvuf.models import Call
+
+logger = logging.getLogger(__name__)
+
+
+def extract_call_features(ph: ProcessHandle, th: ThreadHandle, ch: CallHandle) -> Iterator[Tuple[Feature, Address]]:
+    """
+    This method extracts the given call's features (such as API name and arguments),
+    and returns them as API, Number, and String features.
+
+    args:
+      ph: process handle (for defining the extraction scope)
+      th: thread handle (for defining the extraction scope)
+      ch: call handle (for defining the extraction scope)
+
+    yields:
+      Feature, address; where Feature is either: API, Number, or String.
+    """
+    call: Call = ch.inner
+
+    # list similar to disassembly: arguments right-to-left, call
+    for arg_value in reversed(call.arguments.values()):
+        try:
+            yield Number(int(arg_value, 0)), ch.address
+        except ValueError:
+            # DRAKVUF automatically resolves the contents of memory addresses, (e.g. Arg1="0xc6f217efe0:\"ntdll.dll\"").
+            # For those cases we yield the entire string as it, since yielding the address only would
+            # likely not provide any matches, and yielding just the memory contentswould probably be misleading,
+            # but yielding the entire string would be helpful for an analyst looking at the verbose output
+            yield String(arg_value), ch.address
+
+    for name in capa.features.extractors.helpers.generate_symbols("", call.name):
+        yield API(name), ch.address
+
+
+def extract_features(ph: ProcessHandle, th: ThreadHandle, ch: CallHandle) -> Iterator[Tuple[Feature, Address]]:
+    for handler in CALL_HANDLERS:
+        for feature, addr in handler(ph, th, ch):
+            yield feature, addr
+
+
+CALL_HANDLERS = (extract_call_features,)

capa/features/extractors/drakvuf/extractor.py

@@ -0,0 +1,96 @@
+# Copyright (C) 2024 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 Dict, List, Tuple, Union, Iterator
+
+import capa.features.extractors.drakvuf.call
+import capa.features.extractors.drakvuf.file
+import capa.features.extractors.drakvuf.thread
+import capa.features.extractors.drakvuf.global_
+import capa.features.extractors.drakvuf.process
+from capa.features.common import Feature, Characteristic
+from capa.features.address import NO_ADDRESS, Address, ThreadAddress, ProcessAddress, AbsoluteVirtualAddress, _NoAddress
+from capa.features.extractors.base_extractor import (
+    CallHandle,
+    SampleHashes,
+    ThreadHandle,
+    ProcessHandle,
+    DynamicFeatureExtractor,
+)
+from capa.features.extractors.drakvuf.models import Call, DrakvufReport
+from capa.features.extractors.drakvuf.helpers import index_calls
+
+logger = logging.getLogger(__name__)
+
+
+class DrakvufExtractor(DynamicFeatureExtractor):
+    def __init__(self, report: DrakvufReport):
+        super().__init__(
+            # DRAKVUF currently does not yield hash information about the sample in its output
+            hashes=SampleHashes(md5="", sha1="", sha256="")
+        )
+
+        self.report: DrakvufReport = report
+
+        # sort the api calls to prevent going through the entire list each time
+        self.sorted_calls: Dict[ProcessAddress, Dict[ThreadAddress, List[Call]]] = index_calls(report)
+
+        # pre-compute these because we'll yield them at *every* scope.
+        self.global_features = list(capa.features.extractors.drakvuf.global_.extract_features(self.report))
+
+    def get_base_address(self) -> Union[AbsoluteVirtualAddress, _NoAddress, None]:
+        # DRAKVUF currently does not yield information about the PE's address
+        return NO_ADDRESS
+
+    def extract_global_features(self) -> Iterator[Tuple[Feature, Address]]:
+        yield from self.global_features
+
+    def extract_file_features(self) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.drakvuf.file.extract_features(self.report)
+
+    def get_processes(self) -> Iterator[ProcessHandle]:
+        yield from capa.features.extractors.drakvuf.file.get_processes(self.sorted_calls)
+
+    def extract_process_features(self, ph: ProcessHandle) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.drakvuf.process.extract_features(ph)
+
+    def get_process_name(self, ph: ProcessHandle) -> str:
+        return ph.inner["process_name"]
+
+    def get_threads(self, ph: ProcessHandle) -> Iterator[ThreadHandle]:
+        yield from capa.features.extractors.drakvuf.process.get_threads(self.sorted_calls, ph)
+
+    def extract_thread_features(self, ph: ProcessHandle, th: ThreadHandle) -> Iterator[Tuple[Feature, Address]]:
+        if False:
+            # force this routine to be a generator,
+            # but we don't actually have any elements to generate.
+            yield Characteristic("never"), NO_ADDRESS
+        return
+
+    def get_calls(self, ph: ProcessHandle, th: ThreadHandle) -> Iterator[CallHandle]:
+        yield from capa.features.extractors.drakvuf.thread.get_calls(self.sorted_calls, ph, th)
+
+    def get_call_name(self, ph: ProcessHandle, th: ThreadHandle, ch: CallHandle) -> str:
+        call: Call = ch.inner
+        call_name = "{}({}){}".format(
+            call.name,
+            ", ".join(f"{arg_name}={arg_value}" for arg_name, arg_value in call.arguments.items()),
+            (f" -> {getattr(call, 'return_value', '')}"),  # SysCalls don't have a return value, while WinApi calls do
+        )
+        return call_name
+
+    def extract_call_features(
+        self, ph: ProcessHandle, th: ThreadHandle, ch: CallHandle
+    ) -> Iterator[Tuple[Feature, Address]]:
+        yield from capa.features.extractors.drakvuf.call.extract_features(ph, th, ch)
+
+    @classmethod
+    def from_report(cls, report: Iterator[Dict]) -> "DrakvufExtractor":
+        dr = DrakvufReport.from_raw_report(report)
+        return DrakvufExtractor(report=dr)

capa/features/extractors/drakvuf/file.py

@@ -0,0 +1,56 @@
+# Copyright (C) 2024 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 Dict, List, Tuple, Iterator
+
+from capa.features.file import Import
+from capa.features.common import Feature
+from capa.features.address import Address, ThreadAddress, ProcessAddress, AbsoluteVirtualAddress
+from capa.features.extractors.helpers import generate_symbols
+from capa.features.extractors.base_extractor import ProcessHandle
+from capa.features.extractors.drakvuf.models import Call, DrakvufReport
+
+logger = logging.getLogger(__name__)
+
+
+def get_processes(calls: Dict[ProcessAddress, Dict[ThreadAddress, List[Call]]]) -> Iterator[ProcessHandle]:
+    """
+    Get all the created processes for a sample.
+    """
+    for proc_addr, calls_per_thread in calls.items():
+        sample_call = next(iter(calls_per_thread.values()))[0]  # get process name
+        yield ProcessHandle(proc_addr, inner={"process_name": sample_call.process_name})
+
+
+def extract_import_names(report: DrakvufReport) -> Iterator[Tuple[Feature, Address]]:
+    """
+    Extract imported function names.
+    """
+    if report.loaded_dlls is None:
+        return
+    dlls = report.loaded_dlls
+
+    for dll in dlls:
+        dll_base_name = dll.name.split("\\")[-1]
+        for function_name, function_address in dll.imports.items():
+            for name in generate_symbols(dll_base_name, function_name, include_dll=True):
+                yield Import(name), AbsoluteVirtualAddress(function_address)
+
+
+def extract_features(report: DrakvufReport) -> Iterator[Tuple[Feature, Address]]:
+    for handler in FILE_HANDLERS:
+        for feature, addr in handler(report):
+            yield feature, addr
+
+
+FILE_HANDLERS = (
+    # TODO(yelhamer): extract more file features from other DRAKVUF plugins
+    # https://github.com/mandiant/capa/issues/2169
+    extract_import_names,
+)

capa/features/extractors/drakvuf/global_.py

@@ -0,0 +1,44 @@
+# Copyright (C) 2024 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 Tuple, Iterator
+
+from capa.features.common import OS, FORMAT_PE, ARCH_AMD64, OS_WINDOWS, Arch, Format, Feature
+from capa.features.address import NO_ADDRESS, Address
+from capa.features.extractors.drakvuf.models import DrakvufReport
+
+logger = logging.getLogger(__name__)
+
+
+def extract_format(report: DrakvufReport) -> Iterator[Tuple[Feature, Address]]:
+    # DRAKVUF sandbox currently supports only Windows as the guest: https://drakvuf-sandbox.readthedocs.io/en/latest/usage/getting_started.html
+    yield Format(FORMAT_PE), NO_ADDRESS
+
+
+def extract_os(report: DrakvufReport) -> Iterator[Tuple[Feature, Address]]:
+    # DRAKVUF sandbox currently supports only PE files: https://drakvuf-sandbox.readthedocs.io/en/latest/usage/getting_started.html
+    yield OS(OS_WINDOWS), NO_ADDRESS
+
+
+def extract_arch(report: DrakvufReport) -> Iterator[Tuple[Feature, Address]]:
+    # DRAKVUF sandbox currently supports only x64 Windows as the guest: https://drakvuf-sandbox.readthedocs.io/en/latest/usage/getting_started.html
+    yield Arch(ARCH_AMD64), NO_ADDRESS
+
+
+def extract_features(report: DrakvufReport) -> Iterator[Tuple[Feature, Address]]:
+    for global_handler in GLOBAL_HANDLER:
+        for feature, addr in global_handler(report):
+            yield feature, addr
+
+
+GLOBAL_HANDLER = (
+    extract_format,
+    extract_os,
+    extract_arch,
+)

capa/features/extractors/drakvuf/helpers.py

@@ -0,0 +1,39 @@
+# Copyright (C) 2024 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 itertools
+from typing import Dict, List
+
+from capa.features.address import ThreadAddress, ProcessAddress
+from capa.features.extractors.drakvuf.models import Call, DrakvufReport
+
+
+def index_calls(report: DrakvufReport) -> Dict[ProcessAddress, Dict[ThreadAddress, List[Call]]]:
+    # this method organizes calls into processes and threads, and then sorts them based on
+    # timestamp so that we can address individual calls per index (CallAddress requires call index)
+    result: Dict[ProcessAddress, Dict[ThreadAddress, List[Call]]] = {}
+    for call in itertools.chain(report.syscalls, report.apicalls):
+        if call.pid == 0:
+            # DRAKVUF captures api/native calls from all processes running on the system.
+            # we ignore the pid 0 since it's a system process and it's unlikely for it to
+            # be hijacked or so on, in addition to capa addresses not supporting null pids
+            continue
+        proc_addr = ProcessAddress(pid=call.pid, ppid=call.ppid)
+        thread_addr = ThreadAddress(process=proc_addr, tid=call.tid)
+        if proc_addr not in result:
+            result[proc_addr] = {}
+        if thread_addr not in result[proc_addr]:
+            result[proc_addr][thread_addr] = []
+
+        result[proc_addr][thread_addr].append(call)
+
+    for proc, threads in result.items():
+        for thread in threads:
+            result[proc][thread].sort(key=lambda call: call.timestamp)
+
+    return result

capa/features/extractors/drakvuf/models.py

@@ -0,0 +1,137 @@
+# Copyright (C) 2024 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 Any, Dict, List, Iterator
+
+from pydantic import Field, BaseModel, ConfigDict, model_validator
+
+logger = logging.getLogger(__name__)
+
+
+REQUIRED_SYSCALL_FIELD_NAMES = {
+    "Plugin",
+    "TimeStamp",
+    "PID",
+    "PPID",
+    "TID",
+    "UserName",
+    "UserId",
+    "ProcessName",
+    "Method",
+    "EventUID",
+    "Module",
+    "vCPU",
+    "CR3",
+    "Syscall",
+    "NArgs",
+}
+
+
+class ConciseModel(BaseModel):
+    ConfigDict(extra="ignore")
+
+
+class DiscoveredDLL(ConciseModel):
+    plugin_name: str = Field(alias="Plugin")
+    event: str = Field(alias="Event")
+    name: str = Field(alias="DllName")
+    pid: int = Field(alias="PID")
+
+
+class LoadedDLL(ConciseModel):
+    plugin_name: str = Field(alias="Plugin")
+    event: str = Field(alias="Event")
+    name: str = Field(alias="DllName")
+    imports: Dict[str, int] = Field(alias="Rva")
+
+
+class Call(ConciseModel):
+    plugin_name: str = Field(alias="Plugin")
+    timestamp: str = Field(alias="TimeStamp")
+    process_name: str = Field(alias="ProcessName")
+    ppid: int = Field(alias="PPID")
+    pid: int = Field(alias="PID")
+    tid: int = Field(alias="TID")
+    name: str = Field(alias="Method")
+    arguments: Dict[str, str]
+
+
+class WinApiCall(Call):
+    # This class models Windows API calls captured by DRAKVUF (DLLs, etc.).
+    arguments: Dict[str, str] = Field(alias="Arguments")
+    event: str = Field(alias="Event")
+    return_value: str = Field(alias="ReturnValue")
+
+    @model_validator(mode="before")
+    @classmethod
+    def build_arguments(cls, values: Dict[str, Any]) -> Dict[str, Any]:
+        args = values["Arguments"]
+        values["Arguments"] = dict(arg.split("=", 1) for arg in args)
+        return values
+
+
+class SystemCall(Call):
+    # This class models native Windows API calls captured by DRAKVUF.
+    # Schema: {
+    #   "Plugin": "syscall",
+    #   "TimeStamp": "1716999134.582553",
+    #   "PID": 3888, "PPID": 2852, "TID": 368, "UserName": "SessionID", "UserId": 2,
+    #   "ProcessName": "\\Device\\HarddiskVolume2\\Windows\\explorer.exe",
+    #   "Method": "NtSetIoCompletionEx",
+    #   "EventUID": "0x27",
+    #   "Module": "nt",
+    #   "vCPU": 0,
+    #   "CR3": "0x119b1002",
+    #   "Syscall": 419,
+    #   "NArgs": 6,
+    #   "IoCompletionHandle": "0xffffffff80001ac0", "IoCompletionReserveHandle": "0xffffffff8000188c",
+    #   "KeyContext": "0x0", "ApcContext": "0x2", "IoStatus": "0x7ffb00000000", "IoStatusInformation": "0x0"
+    # }
+    # The keys up until "NArgs" are common to all the native calls that DRAKVUF reports, with
+    # the remaining keys representing the call's specific arguments.
+    syscall_number: int = Field(alias="Syscall")
+    module: str = Field(alias="Module")
+    nargs: int = Field(alias="NArgs")
+
+    @model_validator(mode="before")
+    @classmethod
+    def build_extra(cls, values: Dict[str, Any]) -> Dict[str, Any]:
+        # DRAKVUF stores argument names and values as entries in the syscall's entry.
+        # This model validator collects those arguments into a list in the model.
+        values["arguments"] = {
+            name: value for name, value in values.items() if name not in REQUIRED_SYSCALL_FIELD_NAMES
+        }
+        return values
+
+
+class DrakvufReport(ConciseModel):
+    syscalls: List[SystemCall] = []
+    apicalls: List[WinApiCall] = []
+    discovered_dlls: List[DiscoveredDLL] = []
+    loaded_dlls: List[LoadedDLL] = []
+
+    @classmethod
+    def from_raw_report(cls, entries: Iterator[Dict]) -> "DrakvufReport":
+        report = cls()
+
+        for entry in entries:
+            plugin = entry.get("Plugin")
+            # TODO(yelhamer): add support for more DRAKVUF plugins
+            # https://github.com/mandiant/capa/issues/2181
+            if plugin == "syscall":
+                report.syscalls.append(SystemCall(**entry))
+            elif plugin == "apimon":
+                event = entry.get("Event")
+                if event == "api_called":
+                    report.apicalls.append(WinApiCall(**entry))
+                elif event == "dll_loaded":
+                    report.loaded_dlls.append(LoadedDLL(**entry))
+                elif event == "dll_discovered":
+                    report.discovered_dlls.append(DiscoveredDLL(**entry))
+
+        return report

capa/features/extractors/drakvuf/process.py

@@ -0,0 +1,40 @@
+# Copyright (C) 2024 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 Dict, List, Tuple, Iterator
+
+from capa.features.common import String, Feature
+from capa.features.address import Address, ThreadAddress, ProcessAddress
+from capa.features.extractors.base_extractor import ThreadHandle, ProcessHandle
+from capa.features.extractors.drakvuf.models import Call
+
+logger = logging.getLogger(__name__)
+
+
+def get_threads(
+    calls: Dict[ProcessAddress, Dict[ThreadAddress, List[Call]]], ph: ProcessHandle
+) -> Iterator[ThreadHandle]:
+    """
+    Get the threads associated with a given process.
+    """
+    for thread_addr in calls[ph.address]:
+        yield ThreadHandle(address=thread_addr, inner={})
+
+
+def extract_process_name(ph: ProcessHandle) -> Iterator[Tuple[Feature, Address]]:
+    yield String(ph.inner["process_name"]), ph.address
+
+
+def extract_features(ph: ProcessHandle) -> Iterator[Tuple[Feature, Address]]:
+    for handler in PROCESS_HANDLERS:
+        for feature, addr in handler(ph):
+            yield feature, addr
+
+
+PROCESS_HANDLERS = (extract_process_name,)