计算机科学
正确性
可验证秘密共享
方案(数学)
博弈论
无线
计算机安全
人工智能
过程(计算)
人机交互
电信
算法
操作系统
数学分析
经济
微观经济学
集合(抽象数据类型)
程序设计语言
数学
作者
Haitao Deng,Jing Hu,Rohit Sharma,Mingsen Mo,Yongjun Ren
标识
DOI:10.1016/j.comcom.2023.04.026
摘要
The continued spread of COVID-19 seriously endangers the physical and mental health of people in all countries. It is an important method to establish inter agency COVID-19 detection and prevention system based on game theory through wireless communication and artificial intelligence. Federated learning (FL) as a privacy preserving machine learning framework has received extensive attention. From the perspective of game theory, FL can be regarded as a process in which multiple participants play games against each other to maximize their own interests. This requires that the user's data is not leaked during the training process. However, existing studies have proved that the privacy protection capability of FL is insufficient. In addition, the existing way of realizing privacy protection through multiple rounds of communication between participants increases the burden of wireless communication. To this end, this paper considers the security model of FL based on game theory, and proposes our scheme, NVAS, a non-interactive verifiable privacy-preserving FL aggregation scheme in wireless communication environments. The NVAS can protect user privacy during FL training without unnecessary interaction between participants, which can better motivate more participants to join and provide high-quality training data. Furthermore, we designed a concise and efficient verification algorithm to ensure the correctness of model aggregation. Finally, the security and feasibility of the scheme are analyzed.
科研通智能强力驱动
Strongly Powered by AbleSci AI