联邦学习安全威胁综述  

A Survey on Threats to Federated Learning

作　　者：王坤庆 刘婧 李晨 赵语杭 吕浩然 李鹏[1] 刘炳莹 Wang Kunqing;Liu Jing;Li Chen;Zhao Yuhang;LÜHaoran;LiPeng;Liu Bingying(Chinese People's Armed Police Force,Beijing 100089;School of Life Sciences,QiluNormal University,Jinan 250200;Information Center of Ministry of Ecology and Environment,Beijing 100029;School of Cyberspace Science and Technology,Beijing Institute of Technology,Beijing 100081)

机构地区：[1]中国人民武装警察部队,北京100089 [2]齐鲁师范学院生命科学学院,济南250200 [3]生态环境部信息中心,心北京100029 [4]北京理工大学网络空间安全学院,北京100081

出　　处：《信息安全研究》2022年第3期223-234,共12页Journal of Information Security Research

基　　金：国家自然科学基金项目(61876019)。

摘　　要：当前,联邦学习已被认为是解决数据孤岛和隐私保护的有效解决方案,其自身安全性和隐私保护问题一直备受工业界和学术界关注.现有的联邦学习系统已被证明存在诸多漏洞,这些漏洞可被联邦学习系统内部或外部的攻击者所利用,破坏联邦学习数据的安全性.首先对特定场景下联邦学习的概念、分类和威胁模型进行介绍;其次介绍联邦学习的机密性、完整性、可用性(CIA)模型;然后对破坏联邦学习CIA模型的攻击方法进行分类研究;最后对CIA模型当前面临的问题挑战和未来研究方向进行分析和总结.At present, federated learning has been considered as an effective solution to solve data island and privacy protection. Its own security and privacy protection issues have attracted widespread attentions from industry and academia. The existing federated learning systems have been proven to have vulnerabilities. These vulnerabilities can be exploited by adversaries, whether within or without the system, to destroy data security. Firstly, this paper introduces the concept, classification and threat models of federated learning in specific scenarios. Secondly, it introduces the confidentiality, integrity, and availability(CIA) model of federated learning. Then, it carries out a classification study on the attack methods that destroy the federated learning CIA model. Finally, it explores the current challenges and future research directions of federated learning CIA model.

关 键 词：联邦学习 隐私泄露 机密性、完整性、可用性模型 成员攻击 生成对抗网络攻击 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TP183[自动化与计算机技术—控制科学与工程]