Evaluating the Dynamics of Membership Privacy in Deep Learning

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Chen, Yuetian, Wang, Zhiqi, Baracaldo, Nathalie, Kadhe, Swanand Ravindra, Yu, Lei
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908477610262528
author Chen, Yuetian
Wang, Zhiqi
Baracaldo, Nathalie
Kadhe, Swanand Ravindra
Yu, Lei
author_facet Chen, Yuetian
Wang, Zhiqi
Baracaldo, Nathalie
Kadhe, Swanand Ravindra
Yu, Lei
contents Membership inference attacks (MIAs) pose a critical threat to the privacy of training data in deep learning. Despite significant progress in attack methodologies, our understanding of when and how models encode membership information during training remains limited. This paper presents a dynamic analytical framework for dissecting and quantifying privacy leakage dynamics at the individual sample level. By tracking per-sample vulnerabilities on an FPR-TPR plane throughout training, our framework systematically measures how factors such as dataset complexity, model architecture, and optimizer choice influence the rate and severity at which samples become vulnerable. Crucially, we discover a robust correlation between a sample's intrinsic learning difficulty, and find that the privacy risk of samples highly vulnerable in the final trained model is largely determined early during training. Our results thus provide a deeper understanding of how privacy risks dynamically emerge during training, laying the groundwork for proactive, privacy-aware model training strategies.
format Preprint
id arxiv_https___arxiv_org_abs_2507_23291
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Evaluating the Dynamics of Membership Privacy in Deep Learning
Chen, Yuetian
Wang, Zhiqi
Baracaldo, Nathalie
Kadhe, Swanand Ravindra
Yu, Lei
Machine Learning
Artificial Intelligence
Membership inference attacks (MIAs) pose a critical threat to the privacy of training data in deep learning. Despite significant progress in attack methodologies, our understanding of when and how models encode membership information during training remains limited. This paper presents a dynamic analytical framework for dissecting and quantifying privacy leakage dynamics at the individual sample level. By tracking per-sample vulnerabilities on an FPR-TPR plane throughout training, our framework systematically measures how factors such as dataset complexity, model architecture, and optimizer choice influence the rate and severity at which samples become vulnerable. Crucially, we discover a robust correlation between a sample's intrinsic learning difficulty, and find that the privacy risk of samples highly vulnerable in the final trained model is largely determined early during training. Our results thus provide a deeper understanding of how privacy risks dynamically emerge during training, laying the groundwork for proactive, privacy-aware model training strategies.
title Evaluating the Dynamics of Membership Privacy in Deep Learning
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2507.23291