Federated Learning Paradigms in Network Security for Distributed Systems

Abstract

In this work, we present a novel approach for enhancing network security within distributed systems through Federated Learning (FL). The proposed method integrates three fundamental algorithms: Federated Averaging (FedAvg), Secure Aggregation, and Adaptive Learning Rate. FedAvg allows collaborative training of a global model while preserving data privacy on decentralized devices. Secure Aggregation employs cryptographic techniques to maintain confidentiality during model updates' aggregation. Adaptive Learning Rate dynamically adjusts the learning rate, enhancing model optimization efficiency. The proposed work illustrates these algorithms through detailed equations and flowcharts. The proposed approach combines these techniques to achieve secure and collaborative model training across distributed devices, thereby enhancing network security.