Abstract

The proliferation of Internet of Things (IoT) devices has created a vast attack surface, making these networks increasingly vulnerable to cyberattacks. Traditional Intrusion Detection Systems (IDS) often struggle to cope with the resource constraints of IoT devices, the dynamic nature of IoT traffic, and the need for real-time threat detection. This paper presents a novel Adaptive Intrusion Detection System (A-IDS) designed specifically for IoT networks. A-IDS employs a hybrid approach that combines federated learning (FL) and edge computing to achieve distributed, adaptive, and efficient intrusion detection. Edge devices perform local anomaly detection using lightweight machine learning models trained collaboratively via FL. This minimizes latency and conserves bandwidth. A centralized server aggregates and refines the global model, enabling the system to adapt to evolving threats. The proposed A-IDS is evaluated using a simulated IoT environment with realistic traffic patterns and attack scenarios. The results demonstrate that A-IDS achieves high detection accuracy, low false positive rates, and minimal resource consumption compared to traditional IDS approaches. This research highlights the potential of FL and edge computing to enhance the security of IoT networks by enabling adaptive and distributed intrusion detection.