Hunting IoT Cyberattacks With AI - Powered Intrusion Detection

• Published in: 2023 IEEE International Conference on Cyber Security and Resilience (CSR) pp. 142 - 147
• Main Authors: Grigoriadou, Sevasti, Radoglou-Grammatikis, Panagiotis, Sarigiannidis, Panagiotis, Makris, Ioannis, Lagkas, Thomas, Argyriou, Vasileios, Lytos, Anastasios, Fountoukidis, Eleftherios
• Format: Conference Proceeding
• Language: English
• Published: IEEE 31.07.2023
• Subjects: Computer crime; Decision making; Intrusion detection; Productivity; Q-learning; Real-time systems; Software defined networking
• DOI: 10.1109/CSR57506.2023.10224981

The rapid progression of the Internet of Things allows the seamless integration of cyber and physical environments, thus creating an overall hyper-connected ecosystem. It is evident that this new reality provides several capabilities and benefits, such as real-time decision-making and increased efficiency and productivity. However, it also raises crucial cybersecurity issues that can lead to disastrous consequences due to the vulnerable nature of the Internet model and the new cyber risks originating from the multiple and heterogeneous technologies involved in the loT. Therefore, intrusion detection and prevention are valuable and necessary mechanisms in the arsenal of the loT security. In light of the aforementioned remarks, in this paper, we introduce an Artificial Intelligence (AI)-powered Intrusion Detection and Prevention System (IDPS) that can detect and mitigate potential loT cyberattacks. For the detection process, Deep Neural Networks (DNNs) are used, while Software Defined Networking (SDN) and Q-Learning are combined for the mitigation procedure. The evaluation analysis demonstrates the detection efficiency of the proposed IDPS, while Q- Learning converges successfully in terms of selecting the appropriate mitigation action.