Deep reinforcement learning based automatic intrusion detection and diagnosis approach for cloud computing attacks

Abstract

The expansion of cloud computing settings is undergoing a major boom in the contemporary period, receiving widespread support. The rising ubiquity of Internet-connected gadgets adds to the construction of a complex network in which data is automatically collected from the environment via sensors and easily relayed over the internet without direct human interaction. Because of the adaptability of cloud computing, this networked environment enables seamless interactions between humans and real-world applications. Cloud computing technologies serve a critical role in addressing numerous demands, enriching both social and business worlds, by operating across varied areas such as healthcare, education, agriculture, and commerce. Nonetheless, the vulnerability of cloud computing to attacks remains a major worry, owing mostly to resource constraints and inherent weaknesses.We provide a novel technique for automated intrusion detection and diagnosis in cloud computing assaults that makes use of deep reinforcement learning (DRL). The first phase involves analyzing the cloud server using the modified bird swarm optimization (MBSO) method. The goal of this method is to identify and optimize aspects that are critical for intrusion detection and diagnosis. Furthermore, we use deep reinforcement learning algorithms to identify and diagnose intrusions within the cloud computing infrastructure. The future method's performance is then evaluated and connected with existing techniques utilizing benchmark cloud datasets from UNM (University of New Mexico). The findings demonstrate the efficacy of our MBSO-DRL method, with a maximum detection rate of 97.373% and a minimum error rate of 10.871%. This demonstrates the resilience and efficiency of our proposed deep reinforcement learning-based solution to improving cloud computing environment security against attacks.