Exploring Machine Learning Models : For IoT Network Intrusion Detection : A Literature Review and Comparative Analysis

Abstract

The Internet of Things (IoT) is a complex network of low-powered interconnected devices that communicate and exchange data over the internet. IoT devices are a popular target for cybercriminals because of the user data they handle and the lack of security updates most IoT devices receive. As cyber defenses mature, so does the complexity of cyber-attacks. Sophisticated attackers can stay undetected inside of a network for long amounts of time because of the complexity of modern networks. Although Machine Learning based Intrusion Detection Systems (IDS) show promise, they are not as effective as traditional manual monitoring by network security specialists. This research identified several existing network activity databases and their trained Machine Learning models. It aims to implement new machine learning algorithms with existing network activity databases to improve intrusion detection effectiveness. The datasets included in this literature review simulate popular IoT cyber-attacks within virtual environments. These databases include attributes of a live network, including both malicious and benign connections. They contain rich metadata which allows ML models to classify connections as potentially malicious. Overall, this research conducted a literature review of ten studies including twenty-eight algorithms and nine datasets.