Year: 2025
Type: conference-paper
Abstract: Membership inference (MI) attacks try to determine if a data sample was used to train a machine learning model. For foundation models trained on unknown Web data, MI attacks are often used to detect copyrighted training materials, measure test set contamination, or audit machine unlearning. Unfortunately, we find that evaluations of MI attacks for foundation models are flawed, because they sample members and non-members from different distributions. For 8 pub-lished MI evaluation datasets, we show that blind attacks-that distinguish the member and non-member distributions without looking at any trained model-outperform state-of-the-art MI attacks. Existing evaluations thus tell us nothing about membership leakage of a foundation model's training data.
Authors: Debeshee Das, Jie Zhang, Florian Tramèr
Institution: ETH Zurich
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4411996540