Landscape of learning techniques for intrusion detection system in IoT: A systematic literature review

Abstract

The IoT has interconnected devices that collaborate via the Internet. Yet, its widespread connectivity and data generation pose cybersecurity risks. Integrating robust intrusion detection systems (IDSs) into the architecture has become crucial. IDSs safeguard data, detect attacks, and ensure network security and privacy. Constructing anomaly-based intrusion detection systems using artificial intelligence methods, often termed learning techniques, has gained significant traction lately. In this context, this study undertakes a systematic literature review to comprehensively analyze the current landscape of research concerning IoT security, explicitly employing learning techniques. These techniques fall under four primary categories: machine learning, deep learning, transfer learning, and federated learning. From a pool of 646 papers published between 2018 and 2023, we have selected 36 papers encompassing all these techniques based on the keywords of the study. These chosen studies were then categorized based on their respective learning techniques, with an additional hybrid classification that combines federated learning and transfer learning. Moreover, the paper provides a comparative analysis of the studied articles across different dimensions. The research outcomes demonstrate the effectiveness of each learning technique, shed light on the datasets and metrics employed, and conclude with a discussion on open challenges and future recommendations in this domain