stratosphereips/awesome-ml-privacy-attacks

Organizes 100+ peer-reviewed papers across four attack vectors—membership inference, model reconstruction, property inference, and model extraction—with linked author implementations where available. Complements curated survey papers and references established testing tools like TensorFlow Privacy and IBM's Adversarial Robustness Toolbox to help researchers evaluate ML privacy vulnerabilities in practice. Maintains categorized sections addressing both white-box and black-box attack scenarios, enabling systematic exploration of privacy threats across different model architectures and training paradigms.

633 stars. No commits in the last 6 months.