Prerequisites [UV & Just]

UV is the package manager used in the repo, while Just allows to define custom commands to ease the use of uv commands.
You can check the installation guides for UV and Just at:

• UV: https://docs.astral.sh/uv/getting-started/installation/
• Just: https://github.com/casey/just?tab=readme-ov-file#installation

It follows a quick guide to install them on Linux.

Install UV and Just on Linux

UV

To install UV, run:

curl -LsSf https://astral.sh/uv/install.sh | sh

Just

To install Just, run:

curl --proto '=https' --tlsv1.2 -sSf https://just.systems/install.sh | bash -s -- --to DEST

Replacing DEST with the path to the destination folder (e.g. ~/.local/bin). Ensure that the chosen location is in the PATH.

Getting Started

Go to the root of the project to execute just commands.
To see all available commands, run:

just help

To install all dependencies, run:

just install

After the dependencies have been installed, you can either:

• Activate the virtual environment that has been created in the .venv folder and run python scripts as usual.
• Use directly the uv run command to skip the environment activation step.

The source code and the scripts are organized in the audit directory as follows:

audit
├── README.md
├── audit                   # Source code
│   ├── attacks
│   │   ├── __init__.py
│   │   ├── attack_utils.py
│   │   └── lira.py
│   ├── constants.py
│   ├── dataset.py
│   ├── main.py
│   ├── models.py
│   ├── one_round_audits
│   │   ├── __init__.py
│   │   ├── approximate.py
│   │   ├── diffusion.py
│   │   ├── global_loss.py
│   │   ├── lira.py
│   │   ├── local_rank.py
│   │   └── utils
│   │       ├── __init__.py
│   │       ├── eps_estimation.py
│   │       ├── guided_diffusion.py
│   │       ├── neighbors.py
│   │       └── stable_diffusion.py
│   ├── train
│   │   ├── __init__.py
│   │   ├── privacy_engine.py
│   │   └── train.py
│   └── utils.py
├── pyproject.toml
└── scripts                 # Scripts to train models and perform audits via slurm
    ├── attack.sh
    ├── audit_dpsgd.sh
    ├── audit_relax_loss.sh
    ├── audit_undefended.sh
    ├── train_dpsgd.sh
    ├── train_relax_loss.sh
    └── train_undefended.sh