Homomorphic Encryption and Secure Aggregation Based Vertical-Horizontal Federated Learning for Flight Operation Data Sharing
Xinyan Li, Huimin Zhao, Xintao Chen, Wu Deng · 2024
Aiming at the privacy risks and insufficient data sharing in traditional flight operation data sharing based on federated learning, a homomorphic encryption and secure aggregation based vertical-horizontal federated learning for flight operation data sharing is proposed. Vertical federated learning based on homomorphic encryption is used to process vertical partitioned data owned by different types of data subjects such as airports, airlines, and air traffic control. Furthermore, horizontal federated learning based on secret sharing is used to aggregate model parameters of the same type entities. The scope of data sharing is expanded. The experimental results show that while ensuring data security, the vertical-horizontal federated learning achieves similar results as centralized machine learning methods without privacy protection. A security and effective basis for civil aviation departments to make strategic business decisions is provides.