Paper 2021/1614

PEPFL: A Framework for a Practical and Efficient Privacy-Preserving Federated Learning

Yange Chen, Baocang Wang, Hang Jiang, Pu Duan, Benyu Zhang, Chengdong Liu, Zhiyong Hong, and Yupu Hua

Abstract

As an emerging joint learning model, federated deep learning is a promising way to combine model parameters of different users for training and inference without collecting users’ original data. However, a practical and efficient solution has not been established in previous work due to the absence of effcient matrix computation and cryptography schemes in the privacy-preserving federated learning model, especially in partially homomorphic cryptosystems. In this paper, we propose a practical and efficient privacy-preserving federated learning framework (PEPFL). First, we present a lifted distributed ElGamal cryptosystem that can be applied to federated learning and solve the multi-key problem in federated learning. Secondly, we develop a practical partially single instruction multiple data (PSIMD) parallelism scheme that can encode a plaintext matrix into single plaintext to conduct the encryption, improving effectiveness and reducing communication cost in partially homomorphic cryptosystems. In addition, a novel privacy-preserving federated learning framework is designed by using momentum gradient descent (MGD) with a convolutional neural network (CNN) and the designed cryptosystem. Finally, we evaluate the security and performance of PEPFL. The experiment results demonstrate that the scheme is practicable, effective, and secure with low communication and computational costs.