ml-analyzer

The goal of this project is to implement a ML model analysis framework. To extract the model in the APK and perform some simple attack experiments.

How to start

Before starting

We are using pipenv to manage python dependencies, so you may need to install pipenv first:

pip install pipenv

Once you have done that, you can install all the dependencies by the following command:

pipenv sync

This will also create a standalone python virtual environment with all packages installed.

Run

Before running, we need to spawn a pipenv shell environment first:

pipenv shell

Now you can get help message of ml-analyzer tool by:

python main.py --help

usage: ml-analyzer [-h] [--adb-serial ADB_SERIAL] {detect,extract,run} ...

A ML model analysis framework.

positional arguments:
  {detect,extract,run}  sub-command help

optional arguments:
  -h, --help            show this help message and exit
  --adb-serial ADB_SERIAL
                        A serial number which can be used to identify a
                        connected android device. can be found in `adb device
                        -l`.

detect ML framework

python main.py detect --apk ./tests/apks/tflite_example_image_classification.apk

extract ML framework

python main.py extract --apk ./tests/apks/tflite_example_image_classification.apk

Test

We write test code and use pytest to execute these tests. You do not need to install pytest manually, which is already included in the virtual environment.

Just enter the pipenv shell and run:

pytest --log-cli-level DEBUG

Scripts

There are some tools that can be used in the process in the scripts directory