Comparative Analysis of Machine and Deep Learning Models for Intrusion Detection in Fog-Enabled IoT Networks

IoT is rapidly being integrated across various domains, with fog computing enhancing cloud services by extending them to the network’s edge. This combination is enabling smart environments to be adopted in different areas. However, security risks are significantly expanding the attack surface for intruders, thereby enabling more devastating cyber-attacks. Intruders target IoT network resources, aiming to deplete them through malicious activities. Developing new techniques and detection algorithms for IoT networks necessitates a benchmark assessment of various machine learning and deep learning algorithms. This paper aims to identify a stable and accurate model for effective intrusion detection by investigating various machine learning and deep learning models. It compares and contrasts the ways in which nine various models operate. The paper utilizes the IoTID20 dataset as a basis for contrasting different intrusion detection techniques in IoT networks and ToN_IoT for validation. As both IoTID20 and ToN_IoT datasets are highly imbalanced, SMOTE techniques are utilized for data balancing. The experiments are compared using confusion matrix parameters and ROC curve analysis. The deep learning models outperform the performance of machine learning models in regards of the confusion matrix.