Enhancing IoT Network Security Through Intrusion Detection Using Machine Learning

With the increasing growth of Internet of Things (IoT) devices, network infrastructures have become more vulnerable to cyber-attacks. Traditional network security measures often fall short in detecting sophisticated intrusion patterns in real-time, highlighting the need for intelligent and integrated detection systems. This study proposes a machine learning-based approach to enhance IoT network security by leveraging advanced classification models. The process involves data preprocessing, normalisation, and feature selection using mutual information to identify the most impactful attributes. We evaluated several supervised learning algorithms specifically, Decision Trees, Random Forests, and XGBoost using metrics such as accuracy, precision, recall, F1-score, and ROC-AUC. To enhance detection performance further, we implemented a soft voting ensemble classifier that combines the strengths of the individual models. The study also focuses on binary classification by distinguishing benign from malicious traffic, simplifying the real-time detection tasks. The ensemble model demonstrates superior accuracy, robustness, and generalisation, making it a viable solution for modern IoT intrusion detection systems. All experiments and evaluations are conducted using the CIC IoT 2023 dataset, a comprehensive and up-to-date benchmark for IoT security research.