Gen-LRA: Towards a Principled Membership Inference Attack for Generative Models

Joshua Ward, Chi-Hua Wang, Guang Cheng

Keywords: Privacy, Membership Inference Attacks, Generative Models

Abstract:

Evaluating the potential privacy leakage of synthetic data is an important but unresolved problem. Most existing adversarial auditing frameworks for synthetic data rely on heuristics and unreasonable assumptions to attack the failure modes of generative models, exhibiting limited capability to describe and detect the privacy exposure of training data. In this paper, we study designing Membership Inference Attacks (MIAs) that specifically exploit the observation that generative models tend to memorize certain data points in their training sets, leading to significant local overfitting. Here, we propose Generative Likelihood Ratio Attack (Gen-LRA), a novel, computationally efficient shadow-box MIA that, with no assumption of model knowledge or access, attacks the generated synthetic dataset by conducting a hypothesis test that it is locally overfit to potential training data. Assessed over a comprehensive benchmark spanning diverse datasets, model architectures, and attack parameters, we find that Gen-LRA consistently dominates other MIAs for generative models across multiple performance metrics. These results underscore Gen-LRA's effectiveness as an interpretable and robust privacy auditing tool, highlighting the significant privacy risks posed by generative model overfitting in real-world applications

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: 12851