Merge pull request #2742 from mandiant/idalib-tests

This commit is contained in:
Moritz
2026-01-13 21:48:30 +01:00
committed by GitHub
10 changed files with 290 additions and 40 deletions

View File

@@ -206,3 +206,54 @@ jobs:
GHIDRA_INSTALL_DIR: ${{ github.workspace }}/.github/ghidra/ghidra_${{ matrix.ghidra-version }}_PUBLIC
run: pytest -v tests/test_ghidra_features.py
idalib-tests:
name: IDA ${{ matrix.ida.version }} tests for ${{ matrix.python-version }}
runs-on: ubuntu-22.04
needs: [tests]
env:
IDA_LICENSE_ID: ${{ secrets.IDA_LICENSE_ID }}
strategy:
fail-fast: false
matrix:
python-version: ["3.10", "3.13"]
ida:
- version: 9.0
slug: "release/9.0/ida-essential/ida-essential_90_x64linux.run"
- version: 9.1
slug: "release/9.1/ida-essential/ida-essential_91_x64linux.run"
- version: 9.2
slug: "release/9.2/ida-essential/ida-essential_92_x64linux.run"
steps:
- name: Checkout capa with submodules
# do only run if IDA_LICENSE_ID is available, have to do this in every step, see https://github.com/orgs/community/discussions/26726#discussioncomment-3253118
if: ${{ env.IDA_LICENSE_ID != 0 }}
uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 # v4.1.1
with:
submodules: recursive
- name: Set up Python ${{ matrix.python-version }}
if: ${{ env.IDA_LICENSE_ID != 0 }}
uses: actions/setup-python@0a5c61591373683505ea898e09a3ea4f39ef2b9c # v5.0.0
with:
python-version: ${{ matrix.python-version }}
- name: Setup uv
if: ${{ env.IDA_LICENSE_ID != 0 }}
uses: astral-sh/setup-uv@61cb8a9741eeb8a550a1b8544337180c0fc8476b # v7.2.0
- name: Install dependencies
if: ${{ env.IDA_LICENSE_ID != 0 }}
run: sudo apt-get install -y libyaml-dev
- name: Install capa
if: ${{ env.IDA_LICENSE_ID != 0 }}
run: |
pip install -r requirements.txt
pip install -e .[dev,scripts]
pip install idapro
- name: Install IDA ${{ matrix.ida.version }}
if: ${{ env.IDA_LICENSE_ID != 0 }}
run: |
uv run hcli --disable-updates ida install --download-id ${{ matrix.ida.slug }} --license-id ${{ secrets.IDA_LICENSE_ID }} --set-default --yes
env:
HCLI_API_KEY: ${{ secrets.HCLI_API_KEY }}
IDA_LICENSE_ID: ${{ secrets.IDA_LICENSE_ID }}
- name: Run tests
if: ${{ env.IDA_LICENSE_ID != 0 }}
run: pytest -v tests/test_idalib_features.py # explicitly refer to the idalib tests for performance. other tests run above.

View File

@@ -138,6 +138,7 @@ repos:
- "--ignore=tests/test_ghidra_features.py"
- "--ignore=tests/test_ida_features.py"
- "--ignore=tests/test_viv_features.py"
- "--ignore=tests/test_idalib_features.py"
- "--ignore=tests/test_main.py"
- "--ignore=tests/test_scripts.py"
always_run: true

View File

@@ -58,6 +58,7 @@ Additionally a Binary Ninja bug has been fixed. Released binaries now include AR
### New Features
- ci: add support for arm64 binary releases
- tests: run tests against IDA via idalib @williballenthin #2742
### Breaking Changes

View File

@@ -18,6 +18,7 @@ import idaapi
import idautils
import capa.features.extractors.ida.helpers
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
@@ -50,10 +51,39 @@ def extract_recursive_call(fh: FunctionHandle):
yield Characteristic("recursive call"), fh.address
def extract_function_name(fh: FunctionHandle) -> Iterator[tuple[Feature, Address]]:
ea = fh.inner.start_ea
name = idaapi.get_name(ea)
if name.startswith("sub_"):
# skip default names, like "sub_401000"
return
yield FunctionName(name), fh.address
if name.startswith("_"):
# some linkers may prefix linked routines with a `_` to avoid name collisions.
# extract features for both the mangled and un-mangled representations.
# e.g. `_fwrite` -> `fwrite`
# see: https://stackoverflow.com/a/2628384/87207
yield FunctionName(name[1:]), fh.address
def extract_function_alternative_names(fh: FunctionHandle):
"""Get all alternative names for an address."""
for aname in capa.features.extractors.ida.helpers.get_function_alternative_names(fh.inner.start_ea):
yield FunctionName(aname), 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_recursive_call)
FUNCTION_HANDLERS = (
extract_function_calls_to,
extract_function_loop,
extract_recursive_call,
extract_function_name,
extract_function_alternative_names,
)

View File

@@ -20,6 +20,7 @@ import idaapi
import ida_nalt
import idautils
import ida_bytes
import ida_funcs
import ida_segment
from capa.features.address import AbsoluteVirtualAddress
@@ -436,3 +437,16 @@ def is_basic_block_return(bb: idaapi.BasicBlock) -> bool:
def has_sib(oper: idaapi.op_t) -> bool:
# via: https://reverseengineering.stackexchange.com/a/14300
return oper.specflag1 == 1
def find_alternative_names(cmt: str):
for line in cmt.split("\n"):
if line.startswith("Alternative name is '") and line.endswith("'"):
name = line[len("Alternative name is '") : -1] # Extract name between quotes
yield name
def get_function_alternative_names(fva: int):
"""Get all alternative names for an address."""
yield from find_alternative_names(ida_bytes.get_cmt(fva, False) or "")
yield from find_alternative_names(ida_funcs.get_func_cmt(idaapi.get_func(fva), False) or "")

View File

@@ -22,6 +22,7 @@ import idautils
import capa.features.extractors.helpers
import capa.features.extractors.ida.helpers
from capa.features.file import FunctionName
from capa.features.insn import API, MAX_STRUCTURE_SIZE, Number, Offset, Mnemonic, OperandNumber, OperandOffset
from capa.features.common import MAX_BYTES_FEATURE_SIZE, THUNK_CHAIN_DEPTH_DELTA, Bytes, String, Feature, Characteristic
from capa.features.address import Address, AbsoluteVirtualAddress
@@ -129,8 +130,8 @@ def extract_insn_api_features(fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle)
# not a function (start)
return
if target_func.flags & idaapi.FUNC_LIB:
name = idaapi.get_name(target_func.start_ea)
name = idaapi.get_name(target_func.start_ea)
if target_func.flags & idaapi.FUNC_LIB or not name.startswith("sub_"):
yield API(name), ih.address
if name.startswith("_"):
# some linkers may prefix linked routines with a `_` to avoid name collisions.
@@ -139,6 +140,10 @@ def extract_insn_api_features(fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle)
# see: https://stackoverflow.com/a/2628384/87207
yield API(name[1:]), ih.address
for altname in capa.features.extractors.ida.helpers.get_function_alternative_names(target_func.start_ea):
yield FunctionName(altname), ih.address
yield API(altname), ih.address
def extract_insn_number_features(
fh: FunctionHandle, bbh: BBHandle, ih: InsnHandle

View File

@@ -12,7 +12,6 @@
# See the License for the specific language governing permissions and
# limitations under the License.
import io
import os
import logging
import datetime
@@ -23,24 +22,13 @@ from pathlib import Path
from rich.console import Console
from typing_extensions import assert_never
import capa.perf
import capa.rules
import capa.engine
import capa.helpers
import capa.version
import capa.render.json
import capa.rules.cache
import capa.render.default
import capa.render.verbose
import capa.features.common
import capa.features.freeze as frz
import capa.render.vverbose
import capa.features.extractors
import capa.render.result_document
import capa.render.result_document as rdoc
import capa.features.extractors.common
import capa.features.extractors.base_extractor
import capa.features.extractors.cape.extractor
from capa.rules import RuleSet
from capa.engine import MatchResults
from capa.exceptions import UnsupportedOSError, UnsupportedArchError, UnsupportedFormatError
@@ -346,12 +334,24 @@ def get_extractor(
import capa.features.extractors.ida.extractor
logger.debug("idalib: opening database...")
# 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()):
with console.status("analyzing program...", spinner="dots"):
if idapro.open_database(str(input_path), run_auto_analysis=True):
raise RuntimeError("failed to analyze input file")
idapro.enable_console_messages(False)
with console.status("analyzing program...", spinner="dots"):
# we set the primary and secondary Lumina servers to 0.0.0.0 to disable Lumina,
# which sometimes provides bad names, including overwriting names from debug info.
#
# use -R to load resources, which can help us embedded PE files.
#
# return values from open_database:
# 0 - Success
# 2 - User cancelled or 32-64 bit conversion failed
# 4 - Database initialization failed
# -1 - Generic errors (database already open, auto-analysis failed, etc.)
# -2 - User cancelled operation
ret = idapro.open_database(
str(input_path), run_auto_analysis=True, args="-Olumina:host=0.0.0.0 -Osecondary_lumina:host=0.0.0.0 -R"
)
if ret != 0:
raise RuntimeError("failed to analyze input file")
logger.debug("idalib: waiting for analysis...")
ida_auto.auto_wait()

View File

@@ -109,6 +109,13 @@ dependencies = [
]
dynamic = ["version"]
[tool.pytest.ini_options]
filterwarnings = [
"ignore:builtin type SwigPyPacked has no __module__ attribute:DeprecationWarning",
"ignore:builtin type SwigPyObject has no __module__ attribute:DeprecationWarning",
"ignore:builtin type swigvarlink has no __module__ attribute:DeprecationWarning",
]
[tool.setuptools.dynamic]
version = {attr = "capa.version.__version__"}

View File

@@ -12,7 +12,7 @@
# See the License for the specific language governing permissions and
# limitations under the License.
import logging
import contextlib
import collections
from pathlib import Path
@@ -20,7 +20,6 @@ from functools import lru_cache
import pytest
import capa.main
import capa.features.file
import capa.features.insn
import capa.features.common
@@ -53,6 +52,7 @@ from capa.features.extractors.base_extractor import (
)
from capa.features.extractors.dnfile.extractor import DnfileFeatureExtractor
logger = logging.getLogger(__name__)
CD = Path(__file__).resolve().parent
DOTNET_DIR = CD / "data" / "dotnet"
DNFILE_TESTFILES = DOTNET_DIR / "dnfile-testfiles"
@@ -200,6 +200,73 @@ def get_binja_extractor(path: Path):
return extractor
# we can't easily cache this because the extractor relies on global state (the opened database)
# which also has to be closed elsewhere. so, the idalib tests will just take a little bit to run.
def get_idalib_extractor(path: Path):
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 idapro
import ida_auto
import capa.features.extractors.ida.extractor
logger.debug("idalib: opening database...")
idapro.enable_console_messages(False)
# we set the primary and secondary Lumina servers to 0.0.0.0 to disable Lumina,
# which sometimes provides bad names, including overwriting names from debug info.
#
# use -R to load resources, which can help us embedded PE files.
#
# return values from open_database:
# 0 - Success
# 2 - User cancelled or 32-64 bit conversion failed
# 4 - Database initialization failed
# -1 - Generic errors (database already open, auto-analysis failed, etc.)
# -2 - User cancelled operation
ret = idapro.open_database(
str(path), run_auto_analysis=True, args="-Olumina:host=0.0.0.0 -Osecondary_lumina:host=0.0.0.0 -R"
)
if ret != 0:
raise RuntimeError("failed to analyze input file")
logger.debug("idalib: waiting for analysis...")
ida_auto.auto_wait()
logger.debug("idalib: opened database.")
extractor = capa.features.extractors.ida.extractor.IdaFeatureExtractor()
fixup_idalib(path, extractor)
return extractor
def fixup_idalib(path: Path, extractor):
"""
IDA fixups to overcome differences between backends
"""
import idaapi
import ida_funcs
def remove_library_id_flag(fva):
f = idaapi.get_func(fva)
f.flags &= ~ida_funcs.FUNC_LIB
ida_funcs.update_func(f)
if "kernel32-64" in path.name:
# remove (correct) library function id, so we can test x64 thunk
remove_library_id_flag(0x1800202B0)
if "al-khaser_x64" in path.name:
# remove (correct) library function id, so we can test x64 nested thunk
remove_library_id_flag(0x14004B4F0)
@lru_cache(maxsize=1)
def get_cape_extractor(path):
from capa.helpers import load_json_from_path
@@ -914,20 +981,8 @@ FEATURE_PRESENCE_TESTS = sorted(
("mimikatz", "function=0x4556E5", capa.features.insn.API("advapi32.LsaQueryInformationPolicy"), False),
("mimikatz", "function=0x4556E5", capa.features.insn.API("LsaQueryInformationPolicy"), True),
# insn/api: x64
(
"kernel32-64",
"function=0x180001010",
capa.features.insn.API("RtlVirtualUnwind"),
True,
),
("kernel32-64", "function=0x180001010", capa.features.insn.API("RtlVirtualUnwind"), True),
# insn/api: x64 thunk
(
"kernel32-64",
"function=0x1800202B0",
capa.features.insn.API("RtlCaptureContext"),
True,
),
("kernel32-64", "function=0x1800202B0", capa.features.insn.API("RtlCaptureContext"), True),
# insn/api: x64 nested thunk
("al-khaser x64", "function=0x14004B4F0", capa.features.insn.API("__vcrt_GetModuleHandle"), True),
@@ -1015,20 +1070,20 @@ FEATURE_PRESENCE_TESTS = sorted(
("pma16-01", "file", OS(OS_WINDOWS), True),
("pma16-01", "file", OS(OS_LINUX), False),
("mimikatz", "file", OS(OS_WINDOWS), True),
("pma16-01", "function=0x404356", OS(OS_WINDOWS), True),
("pma16-01", "function=0x404356,bb=0x4043B9", OS(OS_WINDOWS), True),
("pma16-01", "function=0x401100", OS(OS_WINDOWS), True),
("pma16-01", "function=0x401100,bb=0x401130", OS(OS_WINDOWS), True),
("mimikatz", "function=0x40105D", OS(OS_WINDOWS), True),
("pma16-01", "file", Arch(ARCH_I386), True),
("pma16-01", "file", Arch(ARCH_AMD64), False),
("mimikatz", "file", Arch(ARCH_I386), True),
("pma16-01", "function=0x404356", Arch(ARCH_I386), True),
("pma16-01", "function=0x404356,bb=0x4043B9", Arch(ARCH_I386), True),
("pma16-01", "function=0x401100", Arch(ARCH_I386), True),
("pma16-01", "function=0x401100,bb=0x401130", Arch(ARCH_I386), True),
("mimikatz", "function=0x40105D", Arch(ARCH_I386), True),
("pma16-01", "file", Format(FORMAT_PE), True),
("pma16-01", "file", Format(FORMAT_ELF), False),
("mimikatz", "file", Format(FORMAT_PE), True),
# format is also a global feature
("pma16-01", "function=0x404356", Format(FORMAT_PE), True),
("pma16-01", "function=0x401100", Format(FORMAT_PE), True),
("mimikatz", "function=0x456BB9", Format(FORMAT_PE), True),
# elf support
("7351f.elf", "file", OS(OS_LINUX), True),

View File

@@ -0,0 +1,86 @@
# Copyright 2020 Google LLC
#
# 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
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# 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 pathlib import Path
import pytest
import fixtures
import capa.features.extractors.ida.idalib
from capa.features.file import FunctionName
from capa.features.insn import API
from capa.features.common import Characteristic
logger = logging.getLogger(__name__)
idalib_present = capa.features.extractors.ida.idalib.has_idalib()
if idalib_present:
try:
import idapro # noqa: F401 [imported but unused]
import ida_kernwin
kernel_version: str = ida_kernwin.get_kernel_version()
except ImportError:
idalib_present = False
kernel_version = "0.0"
@pytest.mark.skipif(idalib_present is False, reason="Skip idalib tests if the idalib Python API is not installed")
@fixtures.parametrize(
"sample,scope,feature,expected",
fixtures.FEATURE_PRESENCE_TESTS + fixtures.FEATURE_SYMTAB_FUNC_TESTS,
indirect=["sample", "scope"],
)
def test_idalib_features(sample: Path, scope, feature, expected):
if kernel_version in {"9.0", "9.1"} and sample.name.startswith("2bf18d"):
if isinstance(feature, (API, FunctionName)) and feature.value == "__libc_connect":
# see discussion here: https://github.com/mandiant/capa/pull/2742#issuecomment-3674146335
#
# > i confirmed that there were changes in 9.2 related to the ELF loader handling names,
# > so I think its reasonable to conclude that 9.1 and older had a bug that
# > prevented this name from surfacing.
pytest.xfail(f"IDA {kernel_version} does not extract all ELF symbols")
if kernel_version in {"9.0"} and sample.name.startswith("Practical Malware Analysis Lab 12-04.exe_"):
if isinstance(feature, Characteristic) and feature.value == "embedded pe":
# see discussion here: https://github.com/mandiant/capa/pull/2742#issuecomment-3667086165
#
# idalib for IDA 9.0 doesn't support argv arguments, so we can't ask that resources are loaded
pytest.xfail("idalib 9.0 does not support loading resource segments")
try:
fixtures.do_test_feature_presence(fixtures.get_idalib_extractor, sample, scope, feature, expected)
finally:
logger.debug("closing database...")
import idapro
idapro.close_database(save=False)
logger.debug("closed database.")
@pytest.mark.skipif(idalib_present is False, reason="Skip idalib tests if the idalib Python API is not installed")
@fixtures.parametrize(
"sample,scope,feature,expected",
fixtures.FEATURE_COUNT_TESTS,
indirect=["sample", "scope"],
)
def test_idalib_feature_counts(sample, scope, feature, expected):
try:
fixtures.do_test_feature_count(fixtures.get_idalib_extractor, sample, scope, feature, expected)
finally:
logger.debug("closing database...")
import idapro
idapro.close_database(save=False)
logger.debug("closed database.")