AI enhanced intrusion detection system for IoT networks using lightweight machine learning model

Abstract

The rapid growth of the Internet of Things (IoT) has created significant security challenges because most IoT devices have limited processing power, memory, and energy. This study addresses these issues by developing a lightweight, AI-enhanced intrusion detection system (IDS) tailored for resource-constrained IoT environments. The system emphasizes efficiency, scalability, and real-time detection using lightweight machine learning methods. Using the CIC-IoT 2023 dataset, the researchers built and evaluated models based on Decision Tree and Random Forest algorithms. Data preprocessing, training, and evaluation were performed in Python, and a web-based interface was developed to display detection outputs. Results show that the lightweight IDS achieved 80.61% accuracy, a very small model size of 26.33 KB, and a low detection latency of 0.533 ms-making it suitable for IoT edge deployment. Although traditional models achieved higher accuracies, their computational and memory demands make them unsuitable for constrained devices. The study concludes that lightweight AI-driven IDS solutions can effectively balance accuracy and resource efficiency, making them ideal for IoT environments such as smart homes, healthcare, and industrial applications