Review of Machine Learning Intrusion Detection Systems in Internet of Things Environment

Abstract

The use of Internet of Things (IoT) technology entails incorporating numerous devices into heterogeneous networks, which is done in order to provide intelligence as well as automation in services. Yet, increased connectivity also poses various issues relating to the cyber-security of these networks, and resource-constrained nature of IoT devices hampers the effectiveness of conventional cyber-security techniques due to evolving threats. Intrusion Detection Systems (IDSs) has become an integral part of IoT network security. The purpose of this research is to analyze Machine Learning (ML)-based approaches to detect cyber-attacks in the environment of IoT. The review includes the types of learning approaches that were used in studies on existing IDSs, classifiers, databases, the types of attacks and anomalies detected, techniques to select features, as well as methods for evaluation of IDS. Thus, the analysis demonstrates that using machine learning enables to detect a variety of cyber-attacks and anomalies. Still, there are some barriers, such as imbalanced and small-size database, feature redundancy, computational constraints, false positives, and identification of new attacks. Thus, the review highlights the benefits and draw backs of currently existing IDS based on machine learning and proposes possible areas for future studies.