Privacy Preserving Machine Learning with Homomorphic Encryption and Federated Learning

TLDR

Privacy protection is a key concern in the era of successful machine learning. The paper proposes PFMLP, a multi‑party privacy‑preserving machine learning framework that combines partially homomorphic encryption with federated learning. PFMLP trains by having all parties transmit encrypted gradients via homomorphic encryption, and the paper discusses trade‑offs in key length, network structure, and client number. Experiments show PFMLP achieves accuracy within 1% of baseline while the improved Paillier algorithm speeds training by 25–28%.

Abstract

Privacy protection has been an important concern with the great success of machine learning. In this paper, it proposes a multi-party privacy preserving machine learning framework, named PFMLP, based on partially homomorphic encryption and federated learning. The core idea is all learning parties just transmitting the encrypted gradients by homomorphic encryption. From experiments, the model trained by PFMLP has almost the same accuracy, and the deviation is less than 1%. Considering the computational overhead of homomorphic encryption, we use an improved Paillier algorithm which can speed up the training by 25–28%. Moreover, comparisons on encryption key length, the learning network structure, number of learning clients, etc. are also discussed in detail in the paper.