Defending Membership Inference Attacks via Privacy-aware Sparsity Tuning

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Hu, Qiang, Zhang, Hengxiang, Wei, Hongxin
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910642018975744
author Hu, Qiang
Zhang, Hengxiang
Wei, Hongxin
author_facet Hu, Qiang
Zhang, Hengxiang
Wei, Hongxin
contents Over-parameterized models are typically vulnerable to membership inference attacks, which aim to determine whether a specific sample is included in the training of a given model. Previous Weight regularizations (e.g., L1 regularization) typically impose uniform penalties on all parameters, leading to a suboptimal tradeoff between model utility and privacy. In this work, we first show that only a small fraction of parameters substantially impact the privacy risk. In light of this, we propose Privacy-aware Sparsity Tuning (PAST), a simple fix to the L1 Regularization, by employing adaptive penalties to different parameters. Our key idea behind PAST is to promote sparsity in parameters that significantly contribute to privacy leakage. In particular, we construct the adaptive weight for each parameter based on its privacy sensitivity, i.e., the gradient of the loss gap with respect to the parameter. Using PAST, the network shrinks the loss gap between members and non-members, leading to strong resistance to privacy attacks. Extensive experiments demonstrate the superiority of PAST, achieving a state-of-the-art balance in the privacy-utility trade-off.
format Preprint
id arxiv_https___arxiv_org_abs_2410_06814
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Defending Membership Inference Attacks via Privacy-aware Sparsity Tuning
Hu, Qiang
Zhang, Hengxiang
Wei, Hongxin
Machine Learning
Artificial Intelligence
Over-parameterized models are typically vulnerable to membership inference attacks, which aim to determine whether a specific sample is included in the training of a given model. Previous Weight regularizations (e.g., L1 regularization) typically impose uniform penalties on all parameters, leading to a suboptimal tradeoff between model utility and privacy. In this work, we first show that only a small fraction of parameters substantially impact the privacy risk. In light of this, we propose Privacy-aware Sparsity Tuning (PAST), a simple fix to the L1 Regularization, by employing adaptive penalties to different parameters. Our key idea behind PAST is to promote sparsity in parameters that significantly contribute to privacy leakage. In particular, we construct the adaptive weight for each parameter based on its privacy sensitivity, i.e., the gradient of the loss gap with respect to the parameter. Using PAST, the network shrinks the loss gap between members and non-members, leading to strong resistance to privacy attacks. Extensive experiments demonstrate the superiority of PAST, achieving a state-of-the-art balance in the privacy-utility trade-off.
title Defending Membership Inference Attacks via Privacy-aware Sparsity Tuning
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2410.06814