Adaptive Intrusion Detection System (A-IDS) for IoT Networks: A Hybrid Approach Leveraging Federated Learning and Edge Computing

Abstract

The explosion of Internet of Things (IoT) devices has opened up an enormous attack surface, leaving such networks extremely susceptible to cyberattacks. Conventional Intrusion Detection Systems (IDS) tend to have problems handling the resource limitation of IoT devices, the nature of IoT traffic being dynamic in nature, and real-time detection of threats. This paper introduces a new Adaptive Intrusion Detection System (A-IDS) with a focus on IoT networks. A-IDS uses a hybrid method that integrates federated learning (FL) and edge computing to provide distributed, adaptive, and efficient intrusion detection. Local anomaly detection is carried out on edge devices with lightweight machine learning models that are collaboratively trained through FL. This reduces latency and saves bandwidth. A central server collects and fine-tunes the global model so the system can learn from dynamic threats. The recommended A-IDS is tested with a simulated IoT network with realistic traffic generation and attack models. The experiments show that A-IDS attains high accuracy detection, low rates of false positives, and low resource usage relative to conventional IDS methods. The research shows that FL and edge computing have the potential for improving the security of IoT networks by providing adaptive and distributed intrusion detection.