A blockchain-enabled privacy-preserving and auditable federated learning scheme for 6G edge intelligent networks

To address the challenges of model parameter leakage, central server single-point failure, and insufficient auditability in federated learning (FL) for 6 G edge scenarios, this paper proposes a blockchain-enabled privacy-preserving and auditable FL scheme (BPFL). BPFL replaces traditional centralized servers with blockchain to achieve distributed model storage and aggregation, eliminating single-point failure risks. It integrates optimized Paillier homomorphic encryption, single-mask protection, and Shamir threshold sharing to ensure encrypted uploads and secure mask cancelation of client models. A robust aggregation protocol supporting dynamic client exits and leader-failure switching enhances the system stability. Experimental results show BPFL maintains comparable model accuracy to state-of-the-art solutions with better computational efficiency and communication overhead, especially stronger robustness under high client dropout rates.