Co-Evolutionary Defence of Active Directory Attack Graphs via GNN-Approximated Dynamic Programming
Diksha Goel, Hussain Ahmad, Kristen Moore, Mingyu Guo
TL;DR
This work addresses defending Active Directory (AD) attack graphs against adaptive attackers by framing attacker–defender interactions as a Stackelberg game and introducing a co-evolutionary framework that combines GNN-approximated dynamic programming (GNNDP) for attacker policy learning with Evolutionary Diversity Optimisation (EDO) for diverse defender blocking plans. A Fixed-Parameter Tractable (FPT) NSP-based graph reduction condenses large AD graphs while preserving strategic structure, enabling scalable learning. The attacker is modeled as an MDP and approximated with a graph neural network, while the defender generates diverse edge-blocking strategies guided by the GNN-based fitness oracle; these components are trained in a closed loop to prevent overfitting and improve generalisation. Experimental results on synthetic AD graphs show near-optimal performance on small graphs (within 0.1% of DP-optimal) and robust improvements over baselines on larger graphs, demonstrating scalability and practical value for proactive AD defence.
Abstract
Modern enterprise networks increasingly rely on Active Directory (AD) for identity and access management. However, this centralization exposes a single point of failure, allowing adversaries to compromise high-value assets. Existing AD defense approaches often assume static attacker behavior, but real-world adversaries adapt dynamically, rendering such methods brittle. To address this, we model attacker-defender interactions in AD as a Stackelberg game between an adaptive attacker and a proactive defender. We propose a co-evolutionary defense framework that combines Graph Neural Network Approximated Dynamic Programming (GNNDP) to model attacker strategies, with Evolutionary Diversity Optimization (EDO) to generate resilient blocking strategies. To ensure scalability, we introduce a Fixed-Parameter Tractable (FPT) graph reduction method that reduces complexity while preserving strategic structure. Our framework jointly refines attacker and defender policies to improve generalization and prevent premature convergence. Experiments on synthetic AD graphs show near-optimal results (within 0.1 percent of optimality on r500) and improved performance on larger graphs (r1000 and r2000), demonstrating the framework's scalability and effectiveness.