Auditing Privacy Protection of Machine Unlearning

Yuechun Gu, Jiajie He, Keke Chen

Keywords: Machine unlearning, Auditing privacy, Privacy estimation, Membership inference attack

Abstract:

Machine unlearning aims to remove the effect of specific data from trained models to ensure individuals’ privacy. However, it’s arguable how to evaluate whether the privacy protection goal is achieved by machine unlearning. Furthermore, recent studies show unlearning may also increase the retained samples’ privacy risks. This paper takes a holistic approach to auditing both unlearned and retained samples’ privacy risks before and after unlearning.
We derive the privacy criteria for unlearned and retained samples, respectively, based on the perspectives of differential privacy and membership inference attacks. To make the auditing practical, we also develop an efficient membership inference attack, A-LiRA, utilizing data augmentation to reduce the cost of shadow model training. Our experimental findings indicate that existing machine unlearning algorithms do not consistently protect the privacy of unlearned samples and may inadvertently compromise the privacy of retained samples. For reproducibility, we have pubished our code.\footnote{ \url{https://anonymous.4open.science/r/Auditing-machine-unlearning-CB10/README.md}}

Primary Area: alignment, fairness, safety, privacy, and societal considerations

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 9561