FedHENet: A Federated Learning Framework Using Homomorphic Encryption for Efficient Image Classification
By
[Submitted on 13 Feb 2026]
2d ago· 2 min readenInsight
75/100
Toasty
Bagelometer↗
Right out the toaster. Reliable, with some real depth.
Score75TypeanalysisSentimentpositive
Summary
This paper introduces FedHENet, a federated learning framework for image classification that extends the FedHEONN approach. It uses a fixed pre-trained feature extractor and learns only a single output layer, avoiding costly iterative deep network optimization. The method analytically aggregates client knowledge in a single communication round using homomorphic encryption (HE). Experiments show competitive accuracy compared to iterative FL baselines, with superior stability and up to 70% better energy efficiency. The approach is hyperparameter-free, eliminating the carbon footprint from hyperparameter tuning.
Key quotes
· 3 pulledBy using a fixed, pre-trained feature extractor and learning only a single output layer, we avoid costly local fine-tuning.
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.
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, w
