IA Applied to IIoT Intrusion Detection: An Overview

Abstract

Over years, the Industrial Internet of Things (IIoT) has evolved rapidly, offering increased connectivity and benefits in terms of efficiency and productivity as well as significant business opportunities. It is used in various fields such as transportation, production, supply chain management, the oil and gas sector, mining and metallurgy, energy services, aviation, etc. However, this increased connectivity also exposes industrial systems to greater security risks, including intrusion attempts and cyberattacks. The multitude of sensors present in these networks generates a considerable amount of data, attracting the attention of cybercriminals worldwide. To protect Industrial Internet of Things networks, applications against these attacks and intrusion detection systems play a crucial role. However, they have limitations in terms of detection accuracy and false alert management. By using machine learning and deep learning techniques, it is possible to mitigate the multiple security threats and enhance the ability of intrusion detection systems to identify complex attacks patterns and to adapt configurations against new threats. Indeed, such artificial intelligence methods can analyze vast amounts of real-time data, detect anomalies, identify known attack signatures and even predict potential attacks. This article constitutes a bibliographical overview of artificial intelligence based intrusion detection approaches as well as the different datasets on which they have been tested. Additionally, it aims to identify current limitations and challenges in ongoing and existing researches and solutions, while providing some directions for further scientific works.