Intrusion Detection using Federated Learning with Application of Federated Dropout

The massive proliferation of IoT devices in recent years creates new challenges in securing distributed systems, especially when it comes to intrusion detection and mitigation in real-time. Most traditional cloud-based intrusion detection approaches encounter issues ranging from increased latency to bandwidths and even centralized vulnerabilities. This paper introduces an innovative IDS framework based on federated learning (FL) combined with a novel federated dropout mechanism and edge computing. Federated dropout updates model parameters selectively during communication rounds, minimizing bandwidth usage considerably without compromising model performance. This work resolves key challenges such as communication overhead and encrypted traffic processing, opening the door to strong, real-time, and decentralized security solutions.