计算机应用研究2025，Vol.42Issue(6)：1601-1610,10.DOI:10.19734/j.issn.1001-3695.2024.10.0451
联邦学习中隐私保护聚合机制综述
Survey of privacy-preserving aggregation mechanisms in federated learning
摘要
Abstract
As a new distributed machine learning(DML)framework,FL can effectively protect the local data privacy of par-ticipants by aggregating the local model parameters uploaded by participants to train the global model.However,these local model parameters still have the risk of revealing the privacy of participants.As a critical step in FL,the privacy-preserving ag-gregation(PPAgg)mechanism has become a key technology for addressing privacy issues.This paper first introduced the con-cept of FL and its associated privacy and security threats.It then highlighted the core ideas and key procedures of PPAgg mechanisms by integrating existing privacy-preserving techniques in FL.This paper analyzed typical PPAgg mechanisms in de-tail,focusing on their primary advantages and limitations,as well as the specific application scenarios where they were sui-table.Finally,this paper summarized and analyzed current PPAgg mechanisms,explored emerging challenges and development directions for FL,and proposed potential solutions to address these issues.关键词
联邦学习/隐私保护/聚合机制/区块链/安全多方计算Key words
federated learning(FL)/privacy-preserving/aggregation mechanism/blockchain/secure multi-party computa-tion分类
计算机与自动化引用本文复制引用
仇健,马海英,王占君,沈金宇..联邦学习中隐私保护聚合机制综述[J].计算机应用研究,2025,42(6):1601-1610,10.基金项目
南通市自然科学基金面上项目(JC2023069) （JC2023069）
南通大学信息科学技术学院研究生科研与实践创新计划资助项目(NTUSISTPR24_07) （NTUSISTPR24_07）
