Deep Neural Network-Based Intrusion Detection in Internet of Things: A State-of-the-Art Review

Various security threats are faced by the Internet of Things (IoT) as it enriches people’s daily lives. Intrusion detection is employed as an effective method to mitigate these threats, encompassing Botnet, DDoS, and Scan attacks. Due to the rapid development of machine learning technology in recent years, deep neural networks (DNNs) emerge as powerful models utilized to significantly enhance the accuracy performance of intrusion detection systems (IDSs) and to increase their adaptability to dynamic networks. In this paper, related works proposed in the last three years are collected and selected, considering both traffic-based and behavior-based intrusion detection. Subsequently, a study and analysis of these related works is conducted. Additionally, we compare their techniques utilized, results, advantages, and disadvantages. Finally, we analyze the existing challenges and open issues and suggest some insightful future research works.