面向车联网数据隐私保护的高效分布式模型共享策略  

Efficient distributed model sharing strategy for data privacy protection in Internet of vehicles

作　　者：莫梓嘉 高志鹏[1] 杨杨[1] 林怡静 孙山 赵晨 MO Zijia;GAO Zhipeng;YANG Yang;LIN Yijing;SUN Shan;ZHAO Chen(State Key Laboratory of Networking and Switching Technology,Beijing University of Posts and Telecommunications,Beijing 100876,China)

机构地区：[1]北京邮电大学网络与交换技术国家重点实验室,北京100876

出　　处：《通信学报》2022年第4期83-94,共12页Journal on Communications

基　　金：国家自然科学基金资助项目(No.62072049)。

摘　　要：针对车联网隐私数据共享面临的效率问题,提出了基于区块链的高效分布式模型共享策略。针对车联网场景下多实体、多角色的数据共享需求,通过在车辆、路边单元和基站之间构建主从链架构,实现了分布式模型安全共享;提出了基于激励机制的异步联邦学习算法,以激励车辆及路边单元参与优化过程;构造了混合PBFT的改进DPo S共识算法来降低通信成本、提高共识效率。实验分析表明,所提机制能够提高数据共享效率,并具有一定的可扩展性。Aiming at the efficiency problem of privacy data sharing in the Internet of vehicles(IoV),an efficient distributed model sharing strategy based on blockchain was proposed.In response to the data sharing requirements among multiple entities and roles in the IoV,a master-slave chain architecture was built between vehicles,roadside units,and base stations to achieve secure sharing of distributed models.An asynchronous federated learning algorithm based on motivate mechanism was proposed to encourage vehicles and roadside units to participate in the optimization process.An improved DPo S consensus algorithm with hybrid PBFT was constructed to reduce communication costs and improve consensus efficiency.Experimental analysis shows that the proposed mechanism can improve the efficiency of data sharing and has certain scalability.

关 键 词：区块链 车联网 联邦学习 边缘计算 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]