Privacy?preserving federated learning based on multi?key homomorphic encryption

With the advance of machine learning and Internet Things (IoT), security privacy have become critical concerns in mobile services networks. Transferring data to a central unit violates sensitive data. Federated mitigates this need transfer local by sharing model updates only. However, leakage remains an issue. This paper proposes xMK-CKKS, improved version MK-CKKS multi-key homomorphic encryption protocol, design novel privacy-preserving federated scheme. In scheme, are encrypted via aggregated public key before with server for aggregation. For decryption, collaboration among all participating devices is required. Our scheme prevents from publicly shared resistant collusion between k < N ? 1 server. The evaluation demonstrates that outperforms other innovations communication computational cost while preserving accuracy.