Evolutionary LightGBM-Based Intrusion Detection System for IoT Networks

Abstract

With the rapid growth of the Internet of Things (IoT), securing interconnected devices is becoming increasingly critical. This paper introduces the LightShield intrusion detection system (IDS) to enhance intrusion detection in IoT environments using high-performance computing. LightShield features preprocessing of IoT data, ReliefF algorithm for feature selection, and a novel detection model based on LightGBM, a gradient boosting framework. The system leverages GPU acceleration for faster model validation, enabling real-time monitoring. By adapting to IoT characteristics, LightShield provides flexible, scalable defense against evolving cyber threats. Results show its potential to improve security in IoT ecosystems, offering valuable insights into anomaly-based intrusion detection and the future of secure IoT networks. The binary classification model displayed exceptional precision with a 99.82% accuracy in detecting potential attacks, and the multiclass classification model achieved a commendable 97.25% accuracy in classifying distinct attack types.