Thank you for your interest in contributing! This framework is the companion code to a peer-reviewed publication — contributions must maintain academic rigour.
- New attack implementations (e.g. C&W attack, ZOO attack)
- New defense modules (e.g. randomised smoothing, TRADES)
- Construction-specific datasets (real anonymised data, more scenarios)
- Empirical validation (running attacks on real construction AI models)
- Bug fixes and documentation
git clone https://github.com/Ayush-2703/acis-framework.git
cd acis-framework
pip install -e ".[dev]"
pre-commit install- All attacks must subclass
BaseAttackand returnAttackResult - All new threats must reference a published paper in the docstring
- Tests required for all new modules (
pytest tests/ -v) - Docstrings follow numpy style
- Fork the repository
- Create a feature branch:
git checkout -b feat/your-attack-name - Add tests for your changes
- Run
pytest tests/ -v— all tests must pass - Submit a pull request with a clear description
If your contribution introduces a new attack or defense, please cite the
original paper in the docstring and add it to the reference list in README.md.