FLAT: An Efficient Federated Learning Scheme With Adaptive Topology and Secure Aggregation

Federated learning (FL, a system where multiple computers train an AI model together without sending raw data to one central location) has privacy risks because servers can sometimes reconstruct original training data from model gradients (the values used to improve a model). This paper proposes FLAT, a new FL scheme that uses decentralization and secure multi-party computation (MPC, a technique where multiple parties jointly compute a result without revealing their individual data) to reduce privacy risks while making the system faster, achieving 2.7x to 4.3x speed improvements in local networks.