A Review of AI – Driven Security Frameworks for Autonomous IoT Networks

Today’s cyber physical world witnesses the growth of Autonomous IoT Networks in most vital sectors such as healthcare, manufacturing, transportation and energy. Unlike traditional IoT networks which are used to predefined data communication flows and perimeter – based defenses, the A-IoT networks are highly adaptive, open – ended and adhoc in structure. These autonomous systems operate with minimal human intervention. The mechanisms which enable the autonomy of the A-IoT networks such as self-learning models, continuous interdevice communication and collaborative behavior make them vulnerable to sophisticated attacks and threats. Threats such as firmware tampering, data spoofing and insider compromise are further increased by the decentralized nature and heterogeneity of A-IoT systems. Further the autonomy of IoT networks make them highly vulnerable to adaptive cyber threats including DDoS attacks, routing manipulation and data poisoning. The traditional IoT security solutions are static and reactive, which focus on prevention rather than sustained operation under attack. The lack of centralized control and unified security policy in A- IoT networks makes real time threat detection and their mitigation strategies challenging. While the deep learning and machine learning models exist for intrusion detection, most of them lack integration with autonomous decision making and recovery, leading to the IoT networks becoming more vulnerable to cyber-attacks. Cyber resilience underlines the important ability of a system to detect, absorb, adapt, recover and learn from the cyberattacks. Thus it becomes imperative today to develop AI – based, context – aware, cyber resilient framework to safeguard autonomous IoT networks where machine learning and reinforcement learning techniques will help in real time threat detection, adaptive mitigation and recovery. The autonomous IoT networks currently lack the self - adaptive mechanisms which can respond to evolving cyber threats, integrated AI – based resilience decision engines, quantitative evaluation of resilience metrics such as recovery time and survivability. Thus this research focusses on the design and simulation of an AI – driven self - adaptive, cyber resilience framework for autonomous IoT networks. The proposed framework would be designed, implemented and evaluated using NetSim providing a realistic simulation environment. We implement multiple cyber-attack scenarios such as DDoS flooding attacks, sinkhole attacks targeting RPL routing, Sybil attacks using false identities and data injection attacks.