AAPDEM: An Intelligent Model for Automated Attack Path Discovery and Exploitation in Modern Networks
Tamer Mohamed Abdellatif, Salih Rashid Majeed, Ahmed Hamed, Ahmed Ali Seyam · IEEE Access · 2025
Penetration testing plays a pivotal role in proactively identifying security vulnerabilities within modern networks. Traditional approaches, including Deep Reinforcement Learning (DRL)-based models such as Deep Q-Networks, often suffer from high computational cost and poor scalability, limiting their applicability in large and dynamic environments. This paper introduces AAPDEM, an Automated Attack Path Discovery and Exploitation Model, which integrates real-time network scanning, structured vulnerability analysis, and heuristic pathfinding to address these limitations. AAPDEM constructs attack graphs based on discovered vulnerabilities and employs a cost-aware A* search algorithm to identify optimal attack paths. This eliminates the need for model training while significantly reducing memory and processing overhead. The framework adapts efficiently to network changes by incrementally updating only the affected portions of the graph, maintaining high accuracy and responsiveness. Comprehensive experiments were conducted across various network scales using real-world vulnerability data. AAPDEM was benchmarked against state-of-the-art methods, including Random Walk, Greedy Search, Genetic Algorithms, and DRL-based approaches. Results show that AAPDEM consistently achieves higher path accuracy—improving up to 12%—while reducing execution time by 35% on average. Furthermore, it demonstrated strong resilience in dynamic environments by maintaining stability across fluctuating vulnerability conditions. Overall, AAPDEM offers a scalable, efficient, and accurate solution for automated penetration testing, equipping cybersecurity professionals with a practical tool for attack path discovery and risk assessment in evolving network infrastructures.