This paper has been withdrawn by Huaijin Wang
No PDF available, click to view other formats
Abstract:The pervasive deployment of deep learning models across critical domains has concurrently intensified privacy concerns due to their inherent propensity for data memorization. While Membership Inference Attacks (MIAs) serve as the gold standard for auditing these privacy vulnerabilities, conventional MIA paradigms are increasingly constrained by the prohibitive computational costs of shadow model training and a precipitous performance degradation under low False Positive Rate constraints. To overcome these challenges, we introduce a novel perspective by leveraging the principles of model reprogramming as an active signal amplifier for privacy leakage. Building upon this insight, we present \texttt{ReproMIA}, a unified and efficient proactive framework for membership inference. We rigorously substantiate, both theoretically and empirically, how our methodology proactively induces and magnifies latent privacy footprints embedded within the model's representations. We provide specialized instantiations of \texttt{ReproMIA} across diverse architectural paradigms, including LLMs, Diffusion Models, and Classification Models. Comprehensive experimental evaluations across more than ten benchmarks and a variety of model architectures demonstrate that \texttt{ReproMIA} consistently and substantially outperforms existing state-of-the-art baselines, achieving a transformative leap in performance specifically within low-FPR regimes, such as an average of 5.25\% AUC and 10.68\% TPR@1\%FPR increase over the runner-up for LLMs, as well as 3.70\% and 12.40\% respectively for Diffusion Models.
| Comments: | This version was posted without enough prior discussion with my collaborator. Thus, it is being withdrawn pending further internal review. The authors do not wish this version to be considered part of the active scientific record in its current form |
| Subjects: | Machine Learning (cs.LG); Cryptography and Security (cs.CR) |
| Cite as: | arXiv:2603.28942 [cs.LG] |
| (or arXiv:2603.28942v3 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2603.28942 arXiv-issued DOI via DataCite |
From: Huaijin Wang [view email]
[v1]
Mon, 30 Mar 2026 19:35:10 UTC (1,572 KB)
[v2]
Sat, 4 Apr 2026 10:08:01 UTC (1,572 KB)
[v3]
Wed, 15 Apr 2026 17:23:35 UTC (1 KB) (withdrawn)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。