A novel approach for intrusion detection using intelligent agents technology for network security management

Detecting and preventing cyber attacks is of vital importance for computer network security. However, today's cyber attacks can occur so fast, complex, AI-supported, and even capable of being autonomous that they cannot be prevented by traditional Intrusion Detection Systems (IDS). Therefore, there is a need for autonomous IDSs that can perform automated reasoning. In this context, leveraging intelligent agent technologies for network security management is one of the most promising approaches. This thesis proposes an innovative approach to cyber attack detection using intelligent agent technologies for network security management. The study developed various unique agent-based attack detection models through a four-phase development method. The performance metrics of these models were calculated, and the methods, model features, performance criteria, and complexity values were compared, thereby analyzing each model. The results of this thesis indicate that the novel Deep Q-Networks (DQN)-based model (DQN-IoT) proposed in the fourth and final phase of the research stands out compared to other proposed unique agent-based cyber attack detection models.