A bio-inspired and AI-driven approach to DDoS detection

Abstract For implementing Internet Protocol version 6 (IPv6), a key protocol referred to as Internet Control Message Protocol version 6 (ICMPv6) is utilized for inherent IPv6 services. Hence, proper detection and mitigation techniques need to be implemented to monitor the security issues associated with ICMPv6 messages. One of the most targeted forms of attacks is the ICMPv6-based Distributed Denial of Service (DDoS) attacks. In order to attain our objective, we have proposed BioDQN, an advanced artificial intelligence-based approach that incorporates adaptive feature selection for ICMPv6 DDoS Detection using Reinforcement Learning (RL) and Genetic Algorithm (GA). Our approach comprises a bio-inspired feature selection module that incorporates an RL and GA mechanism for optimally selecting feature subsets from the input dataset. The experimental findings suggested that among the various classifiers, the transformer model exhibited the peak detection accuracy of 94.2% with an F1 and Area Under the Curve (AUC) score of 0.95 and 0.91, respectively.