IoT Security Fortification: Enhancing Cyber Threat Detection Through Feature Selection and Advanced Machine Learning

The extensive use of IoT devices has revolutionized various domains, but the connectivity and data interchange in these systems expose them to numerous security vulnerabilities, necessitating robust techniques for promptly and efficiently detecting cyber-attacks. This article commences the use of ensemble-based machine learning (ML) models to detect various categories of attacks against IoT devices. Using the benchmark dataset CICIoT2023, ensemble models are trained with the most impactful features - selected by a combination of two feature selection models to enhance cyber threat detection. Random Forest demonstrated superior accuracy, precision, recall, and F1-score values than the existing implemented models across three different types of classification of the attacks and benign traffic exceeding 99.7%. Implementation of SHAP and LIME explainable AI provides a comprehensive insight into the model's prediction and identifies the most impactful factors for detecting cyber threats.