A Lightweight and Federated Machine Learning Based Intrusion Detection System for Multi Attack Detection in IoT Networks

The quick growth of Internet of Things (IoT) devices has increased the amount of cybersecurity threats, requiring the creation of increasingly complex intrusion detection systems (IDS). The available IDS concentrate on specific threats, rely on restricted datasets, or fail to take into account the resource-constrained of IoT networks. A unique IDS that is lightweight, scalable, and real-time is suggested to detect many types of attacks, including distributed denial of service (DDoS), botnet, and denial of service (DoS) attacks. The suggested approach, which combines hybrid feature selection methods, is used to optimize feature sets. These techniques include Genetic Algorithm (GA), Mutual Information, and Principal Component Analysis (PCA). In addition, federated learning is used for anomaly detection that is responsible for individuals' privacy. Lightweight supervised machine learning models are constructed and assessed using multiple datasets, including IoTID20 and other publicly available benchmarks. The scalability, low latency, and energy efficiency of the proposed system are evaluated by real-time testing in an environment that simulates the Internet of Things for testing purposes. From the simulation results, the proposed approach does a better job than conventional IDS in terms of detection accuracy, computing efficiency, and flexibility to a wide variety of IoT scenarios.