Smart IoT Security: Lightweight Machine Learning Techniques for Multi-Class Attack Detection in IoT Networks

The swift expansion of the Internet of Things (IoT) has significantly increased the demand for effective network security capable of addressing a wide spectrum of new threats. This research addresses existing gaps in multi-class attack detection for IoT devices by introducing innovative, lightweight ensemble methods grounded in robust machine learning principles. By leveraging the CICIoT 2023 dataset-which contains 34 unique attack types categorized into 10 groups-we conducted a thorough evaluation of several state-of-the-art machine learning models to determine those best suited for enhancing IoT security. Our investigation focuses on how various machine learning classifiers respond to the challenges brought by the everchanging and intricate landscape of IoT attack vectors. Notably, the Decision Tree classifier excelled, achieving an accuracy of $\mathbf{9 9. 5 6 \%}$ and an F1 score of $\mathbf{9 9. 6 2 \%}$, proving its dependability in identifying threats both precisely and consistently. Similarly, the Random Forest model delivered impressive results, with $\mathbf{9 8. 2 2 \%}$ accuracy and a $\mathbf{9 8. 2 4 \% F 1}$ score, underscoring the capability of machine learning methods to handle the high-dimensional data prevalent in IoT settings. These findings highlight the promise of integrating machine learning-based classifiers into IoT security infrastructures, paving the way for further research into scalable and efficient detection systems, such as those based on keystroke analysis. Our proposed methodology introduces a novel approach for developing advanced yet lightweight algorithms specifically designed for the limited resources of IoT devices, maintaining an optimal trade-off between detection accuracy and computational efficiency. In summary, this study enriches the IoT security field by presenting a versatile and adaptive framework for intelligent security deployment across interconnected devices.