MIBench: A Comprehensive Framework for Benchmarking Model Inversion Attack and Defense

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Qiu, Yixiang, Yu, Hongyao, Fang, Hao, Zhuang, Tianqu, Yu, Wenbo, Chen, Bin, Wang, Xuan, Xia, Shu-Tao, Xu, Ke
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913726456659968
author Qiu, Yixiang
Yu, Hongyao
Fang, Hao
Zhuang, Tianqu
Yu, Wenbo
Chen, Bin
Wang, Xuan
Xia, Shu-Tao
Xu, Ke
author_facet Qiu, Yixiang
Yu, Hongyao
Fang, Hao
Zhuang, Tianqu
Yu, Wenbo
Chen, Bin
Wang, Xuan
Xia, Shu-Tao
Xu, Ke
contents Model Inversion (MI) attacks aim at leveraging the output information of target models to reconstruct privacy-sensitive training data, raising critical concerns regarding the privacy vulnerabilities of Deep Neural Networks (DNNs). Unfortunately, in tandem with the rapid evolution of MI attacks, the absence of a comprehensive benchmark with standardized metrics and reproducible implementations has emerged as a formidable challenge. This deficiency has hindered objective comparison of methodological advancements and reliable assessment of defense efficacy. To address this critical gap, we build the first practical benchmark named MIBench for systematic evaluation of model inversion attacks and defenses. This benchmark bases on an extensible and reproducible modular-based toolbox which currently integrates a total of 19 state-of-the-art attack and defense methods and encompasses 9 standardized evaluation protocols. Capitalizing on this foundation, we conduct extensive evaluation from multiple perspectives to holistically compare and analyze various methods across different scenarios, such as the impact of target resolution, model predictive power, defense performance and adversarial robustness.
format Preprint
id arxiv_https___arxiv_org_abs_2410_05159
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle MIBench: A Comprehensive Framework for Benchmarking Model Inversion Attack and Defense
Qiu, Yixiang
Yu, Hongyao
Fang, Hao
Zhuang, Tianqu
Yu, Wenbo
Chen, Bin
Wang, Xuan
Xia, Shu-Tao
Xu, Ke
Computer Vision and Pattern Recognition
Cryptography and Security
Model Inversion (MI) attacks aim at leveraging the output information of target models to reconstruct privacy-sensitive training data, raising critical concerns regarding the privacy vulnerabilities of Deep Neural Networks (DNNs). Unfortunately, in tandem with the rapid evolution of MI attacks, the absence of a comprehensive benchmark with standardized metrics and reproducible implementations has emerged as a formidable challenge. This deficiency has hindered objective comparison of methodological advancements and reliable assessment of defense efficacy. To address this critical gap, we build the first practical benchmark named MIBench for systematic evaluation of model inversion attacks and defenses. This benchmark bases on an extensible and reproducible modular-based toolbox which currently integrates a total of 19 state-of-the-art attack and defense methods and encompasses 9 standardized evaluation protocols. Capitalizing on this foundation, we conduct extensive evaluation from multiple perspectives to holistically compare and analyze various methods across different scenarios, such as the impact of target resolution, model predictive power, defense performance and adversarial robustness.
title MIBench: A Comprehensive Framework for Benchmarking Model Inversion Attack and Defense
topic Computer Vision and Pattern Recognition
Cryptography and Security
url https://arxiv.org/abs/2410.05159