FedHide: Federated Learning by Hiding in the Neighbors

TL;DR

FedHide introduces a privacy-preserving federated learning method that uses proxy class prototypes to conceal true class information while enabling effective embedding network training, with proven robustness and convergence guarantees.

Contribution

The paper presents a novel proxy prototype sharing technique for federated learning that enhances privacy without sacrificing model performance.

Findings

Effective discrimination achieved with proxy prototypes.

Robustness against gradient inversion attacks demonstrated.

Convergence properties theoretically analyzed.

Abstract

We propose a prototype-based federated learning method designed for embedding networks in classification or verification tasks. Our focus is on scenarios where each client has data from a single class. The main challenge is to develop an embedding network that can distinguish between different classes while adhering to privacy constraints. Sharing true class prototypes with the server or other clients could potentially compromise sensitive information. To tackle this issue, we propose a proxy class prototype that will be shared among clients instead of the true class prototype. Our approach generates proxy class prototypes by linearly combining them with their nearest neighbors. This technique conceals the true class prototype while enabling clients to learn discriminative embedding networks. We compare our method to alternative techniques, such as adding random Gaussian noise and using random selection with cosine similarity constraints. Furthermore, we evaluate the robustness of our approach against gradient inversion attacks and introduce a measure for prototype leakage. This measure quantifies the extent of private information revealed when sharing the proposed proxy class prototype. Moreover, we provide a theoretical analysis of the convergence properties of our approach. Our proposed method for federated learning from scratch demonstrates its effectiveness through empirical results on three benchmark datasets: CIFAR-100, VoxCeleb1, and VGGFace2.