Comprehensive Study on Detecting Multi-Class Classification of IoT Attack Using Machine Learning Methods

Abstract

The proliferation of IoT devices has heightened their susceptibility to cyberattacks, particularly botnets. Conventional security methods frequently prove inadequate because of the restricted processing capabilities of IoT devices. This paper suggests utilizing machine learning methods to enhance the detection of attacks in Internet of Things (IoT) environments. The paper presents a novel approach to detect different botnet assaults on IoT devices by utilizing ML methods such as XGBoost, Random Forest, LightGBM, and Decision Tree. These algorithms were examined using the N-BaIoT dataset to classify multi-class botnet attacks and were specifically designed to accommodate the limitations of IoT devices. The technique comprises the steps of data preparation, preprocessing, classifier training, and decision-making. The algorithms achieved high detection accuracy rates: XGBoost (99.18%), Random Forest (99.20%), LGBM (99.85%), and Decision Tree (99.17%). The LGBM model demonstrated exceptional performance. The incorporation of the attack evaluation model greatly enhanced the identification of botnets in IoT networks. The paper displays the efficacy of machine learning techniques in identifying botnet assaults in IoT networks. The models generated exhibit exceptional accuracy and can be seamlessly integrated into existing cybersecurity systems.