• 中国精品科技期刊
  • CCF推荐A类中文期刊
  • 计算领域高质量科技期刊T1类
高级检索

基于矩阵映射的拜占庭鲁棒联邦学习算法

刘飚, 张方佼, 王文鑫, 谢康, 张健毅

刘飚, 张方佼, 王文鑫, 谢康, 张健毅. 基于矩阵映射的拜占庭鲁棒联邦学习算法[J]. 计算机研究与发展, 2021, 58(11): 2416-2429. DOI: 10.7544/issn1000-1239.2021.20210633
引用本文: 刘飚, 张方佼, 王文鑫, 谢康, 张健毅. 基于矩阵映射的拜占庭鲁棒联邦学习算法[J]. 计算机研究与发展, 2021, 58(11): 2416-2429. DOI: 10.7544/issn1000-1239.2021.20210633
Liu Biao, Zhang Fangjiao, Wang Wenxin, Xie Kang, Zhang Jianyi. A Byzantine-Robust Federated Learning Algorithm Based on Matrix Mapping[J]. Journal of Computer Research and Development, 2021, 58(11): 2416-2429. DOI: 10.7544/issn1000-1239.2021.20210633
Citation: Liu Biao, Zhang Fangjiao, Wang Wenxin, Xie Kang, Zhang Jianyi. A Byzantine-Robust Federated Learning Algorithm Based on Matrix Mapping[J]. Journal of Computer Research and Development, 2021, 58(11): 2416-2429. DOI: 10.7544/issn1000-1239.2021.20210633
刘飚, 张方佼, 王文鑫, 谢康, 张健毅. 基于矩阵映射的拜占庭鲁棒联邦学习算法[J]. 计算机研究与发展, 2021, 58(11): 2416-2429. CSTR: 32373.14.issn1000-1239.2021.20210633
引用本文: 刘飚, 张方佼, 王文鑫, 谢康, 张健毅. 基于矩阵映射的拜占庭鲁棒联邦学习算法[J]. 计算机研究与发展, 2021, 58(11): 2416-2429. CSTR: 32373.14.issn1000-1239.2021.20210633
Liu Biao, Zhang Fangjiao, Wang Wenxin, Xie Kang, Zhang Jianyi. A Byzantine-Robust Federated Learning Algorithm Based on Matrix Mapping[J]. Journal of Computer Research and Development, 2021, 58(11): 2416-2429. CSTR: 32373.14.issn1000-1239.2021.20210633
Citation: Liu Biao, Zhang Fangjiao, Wang Wenxin, Xie Kang, Zhang Jianyi. A Byzantine-Robust Federated Learning Algorithm Based on Matrix Mapping[J]. Journal of Computer Research and Development, 2021, 58(11): 2416-2429. CSTR: 32373.14.issn1000-1239.2021.20210633

基于矩阵映射的拜占庭鲁棒联邦学习算法

基金项目: 国家重点研发计划项目(2018YFB1004100);信息网络安全公安部重点实验室(公安部第三研究所)开放基金资助课题(C18612);中国科学院网络测评技术重点实验室(中国科学院信息工程研究所)项目(KFKT2019-004)
详细信息
  • 中图分类号: TP391; TP309.2

A Byzantine-Robust Federated Learning Algorithm Based on Matrix Mapping

Funds: This work was supported by the National Key Research and Development Program of China (2018YFB1004100), the Opening Project of Key Lab of Information Network Security of Ministry of Public Security (The Third Research Institute of Ministry of Public Security) (C18612), and the Project of CAS Key Laboratory of Network Assessment Technology (Institute of Information Engineering, Chinese Academy of Sciences) (KFKT2019-004).
  • 摘要: 联邦学习(federated learning)由于参数服务器端只收集客户端模型而不接触客户端本地数据,从而更好地保护数据隐私.然而其基础聚合算法FedAvg容易受到拜占庭客户端攻击.针对此问题,很多研究提出了不同聚合算法,但这些聚合算法存在防守能力不足、模型假设不贴合实际等问题.因此,提出一种新型的拜占庭鲁棒聚合算法.与现有聚合算法不同,该算法侧重于检测Softmax层的概率分布.具体地,参数服务器在收集客户端模型之后,通过构造的矩阵去映射模型的更新部分来获取此模型的Softmax层概率分布,排除分布异常的客户端模型.实验结果表明:在不降低FedAvg精度的前提下,在阻碍收敛攻击中,将拜占庭容忍率从40%提高到45%,在后门攻击中实现对边缘后门攻击的防守.此外,根据目前最先进的自适应攻击框架,设计出专门针对该聚合算法的自适应攻击,并进行了实验评估,实验结果显示,该聚合算法可以防御至少30%的拜占庭客户端.
    Abstract: Federated learning can better protect data privacy because the parameter server only collects the client model and does not touch the local data of the client. However, its basic aggregation algorithm FedAvg is vulnerable to Byzantine client attacks. In response to this problem, many studies have proposed different aggregation algorithms, but these aggregation algorithms have insufficient defensive capabilities, and the model assumptions do not fit the reality. Therefore, we propose a new type of Byzantine robust aggregation algorithm. Different from the existing aggregation algorithms, our algorithm focuses on detecting the probability distribution of the Softmax layer. Specifically, after collecting the client model, the parameter server obtains the Softmax layer probability distribution of the model through the generated matrix to map the updated part of the model, and eliminates the client model with abnormal distribution. The experimental results show that without reducing the accuracy of FedAvg, the Byzantine tolerance rate is increased from 40% to 45% in convergence prevention attacks, and the defense against edge-case backdoor attacks is realized in backdoor attacks. In addition, according to the current state-of-the-art adaptive attack framework, an adaptive attack is designed specifically for our algorithm, and experimental evaluations have been carried out. The experimental results show that our aggregation algorithm can defend at least 30% of Byzantine clients.
  • 期刊类型引用(10)

    1. 崔玉礼,黄丽君. 基于图卷积神经网络的WSN零动态攻击检测方法. 太原学院学报(自然科学版). 2025(01): 78-84 . 百度学术
    2. 何戡,陈金喆,宗学军,齐济,孙永超. 基于油气集输半实物仿真平台的工控网络安全测试研究. 化工自动化及仪表. 2024(02): 274-283 . 百度学术
    3. 李卫峰,冯光辉. 基于动态特征选择的恶意网络行为检测仿真. 计算机仿真. 2024(02): 410-414 . 百度学术
    4. 马佳利,郭渊博,方晨,陈庆礼,张琦. 基于数字孪生的工业互联网安全检测与响应研究. 通信学报. 2024(06): 87-100 . 百度学术
    5. 李一鑫. 面向工业网络场景的基于1DLA-CNN和DCNN-IDS算法的网络安全检测模型研究. 自动化与仪器仪表. 2024(07): 138-142 . 百度学术
    6. 过珺. 基于优先级诊断树的工控网络入侵数据关联挖掘方法. 齐齐哈尔大学学报(自然科学版). 2024(04): 11-16 . 百度学术
    7. 王泽鹏 ,马超 ,张壮壮 ,吴黎兵 ,石小川 . 动态决策驱动的工控网络数据要素威胁检测方法. 计算机研究与发展. 2024(10): 2404-2416 . 本站查看
    8. 刘奇旭,陈艳辉,尼杰硕,罗成,柳彩云,曹雅琴,谭儒,冯云,张越. 基于机器学习的工业互联网入侵检测综述. 计算机研究与发展. 2022(05): 994-1014 . 本站查看
    9. 赵明明,司红星,刘潮. 基于数据挖掘与关联分析的工控设备异常运行状态自动化检测方法分析. 信息安全与通信保密. 2022(04): 2-10 . 百度学术
    10. 刘广睿,张伟哲,李欣洁. 基于边缘样本的智能网络入侵检测系统数据污染防御方法. 计算机研究与发展. 2022(10): 2348-2361 . 本站查看

    其他类型引用(4)

计量
  • 文章访问数:  620
  • HTML全文浏览量:  3
  • PDF下载量:  545
  • 被引次数: 14
出版历程
  • 发布日期:  2021-10-31

目录

    /

    返回文章
    返回