AI-Enhanced Attack Graphs Using Markov Decision Processes for Proactive Threat Hunting and Risk Forecasting
Jagdish Makhijani, Yashwant Vishnupant Pathak, Soumya Bajpai · 2025
The increasing sophistication of cyber threats and the expanding attack surface of modern networks necessitate advanced methodologies for proactive risk assessment and threat mitigation. Traditional attack graphs provide a structured representation of potential attack paths but often struggle with scalability, adaptability, and real-time threat intelligence integration. To address these limitations, this chapter explores the integration of AI-enhanced attack graphs with Markov Decision Processes (MDPs) for proactive threat hunting and cyber risk forecasting. AI-driven techniques, including graph neural networks (GNNs), reinforcement learning, and Bayesian inference, are leveraged to enhance attack graph performance, automate risk assessment, and optimize cybersecurity decision-making. The incorporation of MDPs provides a probabilistic framework for modeling adversarial behavior, enabling predictive analytics for threat evolution and automated mitigation strategies, hybrid AI models improve attack graph scalability by integrating deep learning for pattern recognition, evolutionary algorithms for optimization, and federated learning for distributed security intelligence. The proposed framework shifts cybersecurity from reactive defense mechanisms to a proactive, adaptive, and intelligence-driven approach. Case studies and experimental evaluations demonstrate the efficacy of AI-enhanced attack graphs with MDPs in large-scale, dynamic environments, reinforcing their potential for real-time cyber defense applications. This chapter contributes to advancing risk-aware cybersecurity strategies, fostering automation in cyber risk profiling, and enhancing resilience against emerging threats.