Game-Theoretic and Graph-Based Artificial Intelligence Approaches for Proactive Cyber Defense: A Contemporary Literature Review

Abstract

The rapid progress of AI-based cyber-attacks has made conventional reactive security measures less effective. There has been a growing trend in recent research work (2020-2025) to combine game theory, artificial intelligence (machine learning, deep learning, and reinforcement learning), and graph theory to develop proactive and adaptive cybersecurity systems. This literature review systematically analyses 25 high-impact studies published in the last five years, categorized into game-theoretic models, AI-enhanced cyber defense, graph-based attack modeling, and integrated approaches. Results show that there have been major breakthroughs in Stackelberg and Bayesian games, multi-agent reinforcement learning, graph neural networks, and hybrid models, but there is a definite research gap in scalable and real-time frameworks that can effectively utilize all three paradigms. This paper will identify the research gap and provide future research directions.