Position: Membership Inference Attacks Cannot Prove that a Model Was Trained On Your Data

Publication
Proceedings of the 2025 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML 2025)

Related