Multi-Objective Game Theoretic Optimization Model for Autonomous Cyber Attack Mitigation in Heterogeneous Network Architectures

Abstract

The introduction of cloud, edge, Internet of Things (IoT), software-defined networking (SDN), cyber-physical systems (CPS), and wireless infrastructures has enlarged the cyber threat landscape, which requires new defense techniques to respond to adaptive and coordinated attacks. A Multi-Objective Game-Theoretic Optimization Model (MGTOM) is proposed in this paper for the problem of autonomous cyber-attack mitigation in heterogeneous network environments. The attacker-defender interaction is represented as a dynamic non-cooperative game and both sides simultaneously optimize attack impact, mitigation latency, resource utilization, defense effectiveness and network resilience. The utility-based payoff function and adaptive, strategy-selection algorithm allow autonomous defense responses based on attack severity and available computer resources. The accuracy of controlled simulation-based evaluation is high, with an accuracy of 95.08% and a macro-average AUC score of 0.969. The proposed model achieves an average success rate of mitigations of 93.23% and an effectiveness of 87.8% under the critical level of attack severity. The model takes 55.8 ms to make mitigation decisions at the network level of 800 nodes, and has lower computational and resource overhead than the evaluated baseline methods. The Pareto-based analysis also shows that there is a satisfying compromise between effectiveness of mitigation and time for decision-making. The results show that MGTOM offers a complete and scalable framework to mitigate cyber-attacks autonomously in a heterogeneous network environment.