FEDNET - A Federated Learning Platform

Abstract Federated Learning (FL) is a decentralized approach to training machine learning models along with pre- serving data privacy. This paper introduces FedNet, a secure and efficient FL platform designed for model training across distributed clients. FedNet enhances the traditional FL framework by integrating Fernet encryption for secure communication, HTTP and Socket.IO-based real-time client-server interaction, and a JavaFX-powered user interface for ease of deployment. The platform supports multiple FL aggregation techniques, including FedAvg, FedProx, FedNova, and Zona, providing optimized global model updates. FedNet allows machine learning clients to upload, encrypt, and distribute their models securely while allowing clients to train them on local data without the risk of exposing sensitive information. The system provides authentication through model and client ID verification, encrypts model weights before transmission, and aggregates securely the received weights to improve performance. Experimental analysis demonstrates FedNet’s efficiency, scalability, and security in FL-based model training. The proposed platform bridges the gap between privacy-preserving machine learning and real-world deployment, making it suitable for applications in healthcare, finance, and other data-sensitive industries. Keywords: federated learning, machine learning, data privacy, artificial intelligence, data security