Privacy-Preserving Federated Learning via Homomorphic Adversarial Networks

TL;DR

This paper introduces Homomorphic Adversarial Networks (HANs), a neural network-based protocol for privacy-preserving federated learning that maintains high accuracy and significantly improves encryption efficiency.

Contribution

The paper presents the first neural network-based protocol for PPFL using HANs and an Aggregatable Hybrid Encryption scheme, addressing key sharing and collaboration issues.

Findings

HANs are robust against privacy attacks.

Negligible accuracy loss (up to 1.35%) compared to non-private FL.

HANs increase encryption speed by 6,075 times over traditional schemes.

Abstract

Privacy-preserving federated learning (PPFL) aims to train a global model for multiple clients while maintaining their data privacy. However, current PPFL protocols exhibit one or more of the following insufficiencies: considerable degradation in accuracy, the requirement for sharing keys, and cooperation during the key generation or decryption processes. As a mitigation, we develop the first protocol that utilizes neural networks to implement PPFL, as well as incorporating an Aggregatable Hybrid Encryption scheme tailored to the needs of PPFL. We name these networks as Homomorphic Adversarial Networks (HANs) which demonstrate that neural networks are capable of performing tasks similar to multi-key homomorphic encryption (MK-HE) while solving the problems of key distribution and collaborative decryption. Our experiments show that HANs are robust against privacy attacks. Compared with non-private federated learning, experiments conducted on multiple datasets demonstrate that HANs exhibit a negligible accuracy loss (at most 1.35%). Compared to traditional MK-HE schemes, HANs increase encryption aggregation speed by 6,075 times while incurring a 29.2 times increase in communication overhead.