FedTop:a constraint-loosed federated learning aggregation method against poisoning attack  

作　　者：Che WANG Zhenhao WU Jianbo GAO Jiashuo ZHANG Junjie XIA Feng GAO Zhi GUAN Zhong CHEN 

机构地区：[1]School of Computer Science,Peking University,Beijing 100871,China [2]Peking University Chongqing Research Institute of Big Data,Chongqing 401329,China [3]China Unicom,Beijing 100033,China [4]National Engineering Research Center for Software Engineering,Peking University,Beijing 100871,China

出　　处：《Frontiers of Computer Science》2024年第5期233-235,共3页计算机科学前沿（英文版）

基　　金：This work was supported by the MoST Science and Technology Innovation Project of Xiong'an(2022XAGG0115);the National Natural Science Foundation of China(Grant Nos.62202011,62172010).

摘　　要：Federated learning(FL)is a decentralized machine learning paradigm,which has significant advantages in protecting data privacy[1].However,FL is vulnerable to poisoning attacks that malicious participants perform attacks by injecting dirty data or abnormal model parameters during the local model training and aim to manipulate the performance of the global model[2].

关 键 词：CONSTRAINT LOOSE AGGREGATION 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TP309[自动化与计算机技术—控制科学与工程]