National Cryptographic Algorithms based Personalized Federated Learning with Flight Data

This paper designed FedPDA (Federated Learning Personalized Data Authentication) a new architecture that combines personalized federated learning and domestic encryption algorithms. The FedPDA enhanced the adaptability and prediction accuracy of specific flight data, -divided the model parameters into global general parameters and personalized specific parameters. In addition, FedPDA uses domestic encryption standards SM2 and SM3 to ensure the security and integrity of flight data. The experimental results show that FedPDA can not only effectively protect user privacy and data security, but also significantly improve the stability and efficiency of the heterogeneous flight data.