Research on smart city data encryption and communication efficiency improvement under federated learning framework

Abstract

To improve the data communication processing capability of smart city informatization construction and defend against malicious joint attacks of internal communication participants, we studied the use of function encryption, blockchain, differential privacy and other technologies to defend against weight disclosure, participation in collusion attacks, single point failures and other issues in the federal learning process, and introduced edge computing and asynchronous communication in the classic federal learning framework, Improve the communication efficiency of the model while ensuring the accuracy. The research results show that the accuracy of FE-BDP algorithm can maintain above 95% for a long time, and the change is not significant with the increase of the number of users, indicating that the algorithm has strong stability. The loss value of FE-LDP model is significantly smaller than that of other models and can be stabilized below 0.05. The data aggregation time of blockchain encryption technology is less than 1.0s. The edge-asynchronous communication framework can be applied to multiple urban scenarios and achieve effective data communication, with the highest accuracy rate of 93.87%and the lowest communication cost of 1315.29s. The research results show that the security encryption fusion technology can effectively protect user data privacy, and the edge-asynchronous communication framework has obvious effect on improving communication efficiency, which has important application value for promoting the construction of smart city informatization.