An Adaptive Cyber Deception Approach for Active Cyber-attack Defense Method based on Deep Reinforcement Learning

Abstract

The diverse landscape of network models, including Software-Defined Networking (SDN), Cloud Computing (C2), and Internet of Things (IoT), is evolving to meet the demands of flexibility and performance. However, these environments face numerous security challenges due to cyber-attack complexity. Traditional defense mechanisms are no longer effective against modern attacks. Therefore, Defensive Deception (DD) is proposed as an active defense approach for deceiving attackers. Despite the optimized resource deployment of both Machine Learning (ML) and Deep Learning (DL), they necessitate the usage of pre-existing datasets that have been labeled. Our paper combines Deep Reinforcement Learning (DRL) and SDN technology to establish a novel strategic deception deployment method. This combination creates a powerful security solution that generates deceptive targets and resources to attract attackers, as a result, it provides improved visibility, threat detection, response capabilities, and threat intelligence. Our experiments are implemented on a simulated SDN-based network. The experimental results show that our approach gives significant effectiveness for deception resource allocation compared to random strategies.