车联网中联邦学习模型低时延传输迁移方法研究OA

Federated learning,due to its distributed and privacy-preserving characteristics,has attracted widespread atten-tion in the field of data security in vehicular networks.The asynchronous federated learning mechanism can better adapt to the dynamic changes of vehicle computing power and network conditions,and at the same time improve the efficiency of global model updates and realize effective protection of local privacy data.However,the malicious vehicles in feder-ated learning training may perform poisoning attacks by uploading malicious models to the global model,which in turn af-fects the local training of normal vehicles.During model dissemination,although increasing the number of candidate mod-els can improve the probability of avoiding malicious models,it will significantly increase communication latency and af-fect system performance.To balance security and latency,a federated learning model transmission migration method was proposed.The interaction process between moving vehicles and roadside units(RSUs)on urban roads were modeled,as well as the security of model dissemination.Through reinforcement learning,the vehicle-to-RSU transmission migration strategy was optimized,ensuring the security of model dissemination while effectively reducing communication latency.Simulation results show that,compared with baseline methods,the proposed method reduces the average transmission la-tency by about 7%,which verifies its advantages in terms of security and communication latency.