FedHENet: A Frugal Federated Learning Framework for Heterogeneous Environments

TL;DR

FedHENet introduces a privacy-preserving, energy-efficient federated learning framework that uses a fixed feature extractor and single-round aggregation, achieving competitive accuracy without hyperparameter tuning.

Contribution

It extends FedHEONN to image classification, enabling single-round, hyperparameter-free federated learning with homomorphic encryption for improved efficiency and stability.

Findings

Achieves up to 70% energy efficiency improvement.

Maintains competitive accuracy with iterative FL methods.

Operates without hyperparameter tuning.

Abstract

Federated Learning (FL) enables collaborative training without centralizing data, essential for privacy compliance in real-world scenarios involving sensitive visual information. Most FL approaches rely on expensive, iterative deep network optimization, which still risks privacy via shared gradients. In this work, we propose FedHENet, extending the FedHEONN framework to image classification. By using a fixed, pre-trained feature extractor and learning only a single output layer, we avoid costly local fine-tuning. This layer is learned by analytically aggregating client knowledge in a single round of communication using homomorphic encryption (HE). Experiments show that FedHENet achieves competitive accuracy compared to iterative FL baselines while demonstrating superior stability performance and up to 70\% better energy efficiency. Crucially, our method is hyperparameter-free, removing the carbon footprint associated with hyperparameter tuning in standard FL. Code available in https://github.com/AlejandroDopico2/FedHENet/