PRoVeFL: Private Robust and Verifiable Aggregation in Federated Learning

PRoVeFL is a modular federated learning framework designed to combine privacy, Byzantine robustness, and verifiable aggregation. It uses multiple servers and multi-key fully homomorphic encryption: clients encrypt local updates and distribute encrypted shares, while selected computations are moved into the plaintext domain under stated privacy constraints. The framework supports robust aggregation methods including Krum, Trimmed Mean, FLTrust, norm clipping, and MESAS. According to the abstract, PRoVeFL improves runtime by up to 100x over Prio and 10x over ELSA under comparable distributed-trust security guarantees. The detailed evaluation and implementation costs require inspection of the paper.