IoT Intrusion Detection: A Review of ML and DL-Based Approaches

Abstract

The world is becoming more and more tied to the Internet of Things (IoT) technology. It is present in almost every era: healthcare, agriculture, education, industry, etc. This technology has many advantages, but also many challenges. The characteristics of IoT networks, such as heterogeneity, mobility, multi-tenancy, make the security of these networks a critical challenge. Many solutions have been proposed to ensure the security of IoT networks. One of the very interesting solutions is the intrusion detection system (IDS). Its main role is to continuously monitor the network traffic and detect any malicious activity. There are two intrusion detection strategies: signature-based detection (SIDS) and anomaly-based detection (AIDS). SIDS has demonstrated its effectiveness at detecting known attacks with low false alarms. However, it fails at detecting unknown and new attacks known as zero-day attacks. In contrast, AIDS has demonstrated effective results in detecting zero-day attacks. This article focuses on reviewing recent researches that present intrusion detection systems based on machine learning (ML) and deep learning (DL) techniques in IoT networks. The challenges faced by the reviewed articles are identified, and recommendations are suggested to give a clear path for future work.