logo
科言猫学术研究的AI总结
language
手机端手机端
  • Phone Feature微信小程序
简体中文简体中文
notice消息通知
博客博客
delete
arrow
返回

ProVFL: Property Inference Attacks Against Vertical Federated Learning

delete2025-01-01
delete0
PRE
AI
L
Li Bai
X
Xinwei Zhang
S
Sen Zhang
Q
Qingqing Ye
H
Haibo Hu
DOI：10.1109/TIFS.2025.3581743delete
delete原文链接
delete原文求助
delete分享
delete收藏
摘要

摘要

En 中文
近期研究表明，隐私泄露可能发生在垂直联邦学习（VFL）中，其中各方持有相同样本的分割特征。尽管包括标签和特征推断在内的各种攻击专注于VFL中的记录级隐私风险，但很少有研究深入探讨分布级隐私威胁。本文探讨了VFL中的属性推断攻击（PIAs），其中恶意方试图推断受害者训练集中目标属性的全局分布信息。我们的关键观察是，VFL中间结果的Lp范数分布可能反映训练集中目标属性的比例。受此启发，我们提出了ProVFL，一个包含分布比较和相关性增强模块的新型PIA框架。为了实现属性推断，我们通过创建具有不同比例的中间结果群体来设计分布比较模块，旨在学习Lp范数分布与其比例之间的关系。然后，我们理论分析了影响攻击有效性的因素，并基于标签替换和模型精炼开发了相关性增强模块，以放大属性信息泄露。广泛的实验结果表明，我们的攻击可以实现低至1%的估计误差的推断。这构成了VFL设置中从私有训练数据泄露属性信息的即时威胁。
Keyword:
Property inference attacks
vertical federated learning
privacy attacks
defense mechanisms

期刊

IEEE Transactions on Information Forensics and Security 封面图
IEEE Transactions on Information Forensics and Security
IF:
8
论文数:
5.3K
被引数:
2.3W

机构

T
The Hong Kong Polytechnic University
学者数:
5.1K
论文数: 3.0K
被引数: 17
引用论文

引用论文

LDP-Purifier: Defending against Poisoning Attacks in Local Differential PrivacyLDP-Purifier：防御本地差分隐私中的污染攻击
errLecture Notes in Computer Science
IF0
err2024-10-27
err0
PREAI
errLeixia Wang; Qingqing Ye; Haibo Hu; Xiaofeng Meng; Kai Huang
err分享
err收藏
Comprehensive Analysis of Privacy Leakage in Vertical Federated Learning During Prediction
errProceedings on Privacy Enhancing Technologies
IF0
err2022-03-03
err0
errOAAI
errXue Jiang; Xuebing Zhou; Jens Grossklags
err分享
err收藏
Hacking smart machines with smarter ones: How to extract meaningful data from machine learning classifiers
errInternational Journal of Security and Networks
IF0
err2015-01-01
err0
errOAAI
errGiuseppe Ateniese; Luigi V. Mancini; Angelo Spognardi; Antonio Villani; Domenico Vitali; Giovanni Felici
err分享
err收藏
Group Property Inference Attacks Against Graph Neural Networks
errProceedings of the 2022 ACM SIGSAC Conference on Computer and Communications Security
IF0
err2022-11-07
err0
errOAAI
errXiuling Wang; Wendy Hui Wang
err分享
err收藏
Dual coordinate descent methods for logistic regression and maximum entropy modelslogistic回归和最大熵模型的双坐标下降方法
errMACHINE LEARNING
IF2.9
err2010-11-11
err265
errOAAI
errYu, Hsiang-Fu; Huang, Fang-Lan; Lin, Chih-Jen
err分享
err收藏
HashVFL: Defending Against Data Reconstruction Attacks in Vertical Federated Learning
errIEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY
IF8
err2024-01-01
err4
errOAAI
errQiu, Pengyu; Zhang, Xuhong; Ji, Shouling; Fu, Chong; Yang, Xing; Wang, Ting
err分享
err收藏
Vertical Federated Learning: Concepts, Advances, and Challenges垂直联合学习: 概念、进展与挑战
errIEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
IF10.4
err2024-07-01
err40
errOAAI
errLiu, Yang; Kang, Yan; Zou, Tianyuan; Pu, Yanhong; He, Yuanqin; Ye, Xiaozhou; Ouyang, Ye; Zhang, Ya-Qin; Yang, Qiang
err分享
err收藏
Deep Learning with Differential Privacy具有差分隐私的深度学习
errProceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security
IF0
err2016-10-24
err0
errOAAI
errMartin Abadi; Andy Chu; Ian Goodfellow; H. Brendan McMahan; Ilya Mironov; Kunal Talwar; Li Zhang
err分享
err收藏
Property Inference Attacks Against GANs
errProceedings 2022 Network and Distributed System Security Symposium
IF0
err2022-01-01
err0
errOAAI
errJunhao Zhou; Yufei Chen; Chao Shen; Yang Zhang
err分享
err收藏
Membership Inference Attacks Against Machine Learning Models针对机器学习模型的成员推理攻击
err2017 IEEE Symposium on Security and Privacy (SP)
IF0
err2017-05-01
err0
errOAAI
errReza Shokri; Marco Stronati; Congzheng Song; Vitaly Shmatikov
err分享
err收藏
学者 查看更多内容