Utilizing a Cross-Silo Federated Learning Approach with Wireless Backhaul VPN in Computer-Aided Healthcare System
Atif Mahmood, Saaidal Razalli Azzuhri, Zati Hakim Azizul, Miss Laiha Mat Kiah · 2025
Federated Learning (FL) is rapidly becoming a popular cooperative and distributed approach, utilized by edge devices to develop machine learning models. In this research, we present a high-efficiency FL network designed for analyzing healthcare data, leveraging VPN technology and implementing a cross-silo methodology across a wireless backhaul network. Our detailed evaluation revealed that the FedProx algorithm combined with mmWave technology. It significantly enhances accuracy and decreases convergence time from 55 to 38 seconds, highlighting the benefits of high-bandwidth communication links. Furthermore, we established a comprehensive three-tier security strategy. This strategy starts with the integration of a private network into the telecom framework. Powered by licensed frequency channels and reinforced by VPN-based protections which provides the extensive security for the FL network and its critical data.