Integrated IoT ‐Cloud Security Through Markov Decision Processes, Blockchain Authentication and Reinforcement Learning

Abstract

The pace of the IoT devices growth and their connection with cloud services posed considerable security threats such as DDoS attacks, data breach, and unauthorized access, which undermine the integrity and privacy of the system. The current security solutions are not usually able to deliver adaptive, low‐latency, and reliable security in dynamic IoT‐cloud systems, and a more robust and smarter security model is required. The purpose of this research was to create an end‐to‐end IoT‐cloud security system that is based on markov decision processes, reinforcement learning, and blockchain‐enhanced authentication in order to achieve better attack detection, false alarms, and safe device management. The framework simulated dynamic network states using MDP in the optimal choice, applied RL to keep on enhancing detection policies and used blockchain in the authentication of devices in a non‐tampered manner. The reinforcement learning agent in this framework operates through MDPs by monitoring network states to choose security actions which lead to policy updates based on state‐transition rewards for ongoing threat reduction improvements. Performance was measured using the Bot‐IoT dataset. Experimental findings revealed that it achieved a detection accuracy of 98.00% and reduced the false positive rate to 1.50%, which was much better than the conventional ML and single RL‐MDP model. There was an efficiency improvement of 52.79% reduction in system latency and 48.15% improvement in throughput. The framework offers a scalable, adaptable, and safe system of IoT‐cloud networks to guarantee the integrity of data and resilience of operation.