fkie-cad/sandroid_dexray-interceptA Android malware analysis tool that creates comprehensive runtime profiles by hooking into application behavior across cryptography, file systems, databases, network communications, and process operations.
Android Binary API Tracer
Dexray Intercept is part of the dynamic Sandbox Sandroid. Its purpose is to create runtime profiles to track the behavior of an Android application. This is done utilizing frida.
Just install it with pip:
python3 -m pip install dexray-intercept
This will install Dexray Intercept as command line tool ammm or dexray-intercept.
Further it will provide a package dexray_intercept. More on how to use the package below.
Ensure that your Android device is rooted. The frida-server will be installed to the latest version automatically. Then you can use Dexray Intercept by just invoking the following command:
dexray-intercept <target app>
# or using its old name:
ammm <target app>
All hooks are disabled by default for optimal performance. Enable hooks based on your analysis needs:
# Enable specific hooks
dexray-intercept --enable-aes <app_name> # Enable AES crypto hooks
dexray-intercept --enable-web <app_name> # Enable web/HTTP hooks
dexray-intercept --enable-aes --enable-web <app_name> # Enable multiple hooks
# Enable hook groups
dexray-intercept --hooks-crypto <app_name> # Enable all crypto hooks
dexray-intercept --hooks-network <app_name> # Enable all network hooks
dexray-intercept --hooks-filesystem <app_name> # Enable all file system hooks
# Enable all hooks (performance impact)
dexray-intercept --hooks-all <app_name> # Enable all available hooks
# Use package identifier instead of app name
dexray-intercept -s com.example.package --hooks-crypto
--hooks-crypto (AES, encodings, keystore, certificates)--hooks-network (HTTP, sockets, SSL/TLS)--hooks-filesystem (file operations, databases, shared preferences)--hooks-ipc (intents, broadcasts, binder, shared preferences)--hooks-process (DEX unpacking, native libraries, runtime)--hooks-services (camera, location, telephony, bluetooth)Here an example on monitoring the chrome app on our AVD:
dexray-intercept Chrome
Dexray Intercept
⠀⠀⠀⠀⢀⣀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣀⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢀⣀⣀⣀⣀⣀⡀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠙⢷⣤⣤⣴⣶⣶⣦⣤⣤⡾⠋⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢀⣴⠾⠛⢉⣉⣉⣉⡉⠛⠷⣦⣄⠀⠀⠀⠀
⠀⠀⠀⠀⠀⣴⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣦⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢀⣴⠋⣠⣴⣿⣿⣿⣿⣿⡿⣿⣶⣌⠹⣷⡀⠀⠀
⠀⠀⠀⠀⣼⣿⣿⣉⣹⣿⣿⣿⣿⣏⣉⣿⣿⣧⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣼⠁⣴⣿⣿⣿⣿⣿⣿⣿⣿⣆⠉⠻⣧⠘⣷⠀⠀
⠀⠀⠀⢸⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢰⡇⢰⣿⣿⣿⣿⣿⣿⣿⣿⣿⡿⠀⠀⠈⠀⢹⡇⠀
⣠⣄⠀⢠⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠀⣠⣄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢸⡇⢸⣿⠛⣿⣿⣿⣿⣿⣿⡿⠃⠀⠀⠀⠀⢸⡇⠀
⣿⣿⡇⢸⣿⣿⣿Sandroid⣿⣿⣿⡇⢸⣿⣿⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⣷⠀⢿⡆⠈⠛⠻⠟⠛⠉⠀⠀⠀⠀⠀⠀⣾⠃⠀
⣿⣿⡇⢸⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡇⢸⣿⣿⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠸⣧⡀⠻⡄⠀⠀⠀⠀⠀⠀⠀⠀⠀⢀⣼⠃⠀⠀
⣿⣿⡇⢸⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡇⢸⣿⣿⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢼⠿⣦⣄⠀⠀⠀⠀⠀⠀⠀⣀⣴⠟⠁⠀⠀⠀
⣿⣿⡇⢸⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡇⢸⣿⣿⠀⠀⠀⠀⠀⠀⠀⠀⣠⣾⣿⣦⠀⠀⠈⠉⠛⠓⠲⠶⠖⠚⠋⠉⠀⠀⠀⠀⠀⠀
⠻⠟⠁⢸⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡇⠈⠻⠟⠀⠀⠀⠀⠀⠀⣠⣾⣿⣿⠟⠁⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠉⠉⣿⣿⣿⡏⠉⠉⢹⣿⣿⣿⠉⠉⠀⠀⠀⠀⠀⠀⠀⠀⣠⣾⣿⣿⠟⠁⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⣿⣿⣿⡇⠀⠀⢸⣿⣿⣿⠀⠀⠀⠀⠀⠀⠀⠀⠀⣾⣿⣿⠟⠁⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⣿⣿⣿⡇⠀⠀⢸⣿⣿⣿⠀⠀⠀⠀⠀⠀⠀⢀⣄⠈⠛⠁⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠈⠉⠉⠀⠀⠀⠀⠉⠉⠁⠀⠀⠀⠀⠀⠀⠀⠀⠁⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
[*] starting app profiling
[*] press Ctrl+C to stop the profiling ...
[*] Filesystem profiling informations:
[*] [Libc::read] Read FD (anon_inode:[eventfd],0x7ac6b67540,8)
[*] Filesystem profiling informations:
[*] [Libc::read] Read FD (anon_inode:[eventfd],0x7fcb41c990,8
Install Dexray Intercept as a package and use the new modular architecture:
from dexray_intercept import AppProfiler, setup_frida_device
from dexray_intercept.services.hook_manager import HookManager
# Connect to device and get process
device = setup_frida_device()
process = device.attach("com.example.app")
# Configure hooks (all disabled by default for performance)
hook_config = {
'aes_hooks': True,
'web_hooks': True,
'file_system_hooks': True,
'keystore_hooks': True
}
# Create profiler with new architecture
profiler = AppProfiler(
process,
verbose_mode=True,
output_format="JSON",
hook_config=hook_config,
enable_stacktrace=True
)
# Start profiling
script = profiler.start_profiling()
# ... let app run and collect data ...
# Get results
profile_data = profiler.get_profile_data()
json_output = profiler.get_profiling_log_as_json()
# Runtime hook management
profiler.enable_hook('socket_hooks', True) # Enable more hooks at runtime
enabled_hooks = profiler.get_enabled_hooks() # Check what's enabled
# Stop profiling
profiler.stop_profiling()
Enable specific hook groups based on your analysis needs:
# Crypto hooks
hook_config = {
'aes_hooks': True,
'encodings_hooks': True,
'keystore_hooks': True
}
# Network hooks
hook_config = {
'web_hooks': True,
'socket_hooks': True
}
# File system hooks
hook_config = {
'file_system_hooks': True,
'database_hooks': True
}
# Enable all hooks (performance impact)
profiler.enable_all_hooks()
# Enable hook groups
profiler.enable_hook_group('crypto') # Enable all crypto-related hooks
The old API is still available for backward compatibility:
from dexray_intercept import AppProfilerLegacy
# OR use environment variable: DEXRAY_FORCE_OLD_ARCH=true
profiler = AppProfilerLegacy(process_session, verbose=True, output_format="CMD",
base_path=None, deactivate_unlink=False)
profiler.instrument() # Old method name
# ...
profiler.finish_app_profiling() # Old method name
In order to run it as a package in Sandroid ensure that you also installed the JobManager from AndroidFridaManager. This allows running multpitle frida sessions in different threads.
All you have to do is running the following code:
from AndroidFridaManager import JobManager
from dexray_intercept import AppProfiler
job_manager = JobManager()
app_package = "net.classwindexampleyear.bookseapiececountry"
profiler = AppProfiler(job_manager.process_session, True, output_format="JSON", base_path=None, deactivate_unlink=False)
frida_script_path = profiler.get_frida_script()
job_manager.setup_frida_session(app_package, profiler.on_appProfiling_message)
job = job_manager.start_job(frida_script_path, custom_hooking_handler_name=profiler.on_appProfiling_message)
# close only the job and the frida session keeps active to run other frida scripts
# job_manager.stop_job_with_id(job.job_id)
job_manager.stop_app_with_closing_frida(app_package) # stops the frida session and the app and all frida jobs
profiler.write_profiling_log() # write the log data to profile.json
# instead of writing it to a file the JSON output will just be returned
# profiler.get_profiling_log_as_JSON()
Ensure that no other part of your code is trying to connect to the frida server (no other frida session).
In order to test this you can try the following sample: catelites_2018_01_19.apk. The name for the package is net.classwindexampleyear.bookseapiececountry. Ensure that your AVD is running on Android 9, so that the sample can execute everything of its malicious code. You can install this sample simple with adb install samples/unpacking/catelites_2018_01_19.apk.
In order to compile this project ensure that npm and frida-compile running on your system and installed into your path. Since frida version 17.0 frida-compile gets installed via pip install frida-tools.
Than just invoke the following command in to get the latest frida agent compiled:
$ cd <AppProfiling-Project>
> Dexray [email protected] build
> frida-compile agent/hooking_profile_loader.ts -o src/dexray_intercept/profiling.js
$ npm install frida-java-bridge@latest --save
$ npm install --save-dev @types/frida-gum@latest
> Dexray [email protected] prepare
> npm run build
up to date, audited 75 packages in 6s
19 packages are looking for funding
run `npm fund` for details
found 0 vulnerabilities
This ensures that the latest frida scripts/hooks are used in dexray-intercept.
In order to do adjustments in the python code it is recommend to install dexray-intercept with pip utilizing the editable mode:
python3 -m pip install -e .
This way local changed in the python code gets reflected without creating a new version of the package.
Comprehensive documentation is available covering installation, usage, API reference, and development:
By just invoking the following command in this directory the setup.py should be used to install dexray-intercept as a local python package to your system:
python3 -m pip install .
In order to compile the TypeScript frida hooks we need the frida-compile (link) project. Which will be bundled with frida-tools.
python3 -m pip install frida-tools
Besides this we need also support for frida-java-bridge and the internal frida types:
npm install frida-java-bridge@latest --save
npm install --save-dev @types/frida-gum@latest
When unpacking, applications may load DexCode—previously pointed to distinct memory blocks—into a DexFile, which represents the code being executed. For instance, some applications may restore instructions immediately before execution. In such cases, Sandroid is unable to revert the instructions back into the DexFile. Further research is necessary to resolve this issue
Dexray Intercept builds upon the excellent work of various open-source projects and researchers in the Android security and dynamic analysis community. We would like to acknowledge the following projects that have inspired or contributed to our implementation:
We extend our gratitude to these projects and their maintainers for advancing the state of Android security analysis and making their work available to the community.