Covertrack-Dqn: Enhancing Penetration Testing Efficiency With Deep Q-Networks for Trace Covering
Muhammed Karam Fathi, Khaled Metwally, Mohamed Sobh, Ayman Mohammad Bahaa-Eldin · 2025
Increasing the development of cybersecurity defenses has necessitated advanced techniques development to ethical hackers to keep their effectiveness. In this paper, we present CoverTrack-DQN, a novel framework Deep QNetwork (DQN) for improve track covering in ethical hacking process. Ethical hacking traditional methods rely on fixed instructions and tools, these methods poor adaptability for dynamic environments. CoverTrack-DQN authorizes attackers to learn and adapt their strategies for hiding traces across different attack stages continuously, comprehensive exploitation, privilege escalation, and persistence keeping. By using reinforcement learning, CoverTrack-DQN adapts its actions to maximize trace covering, reducing the risk of detection by intrusion detection systems (IDS) and log analysis tools. Experimental results describe that CoverTrack-DQN safely outperforms traditional approaches, offering better adaptability and efficiency in real penetration testing scenarios. This research highlights the potential of combining reinforcement learning with ethical hacking and sets a foundation for future advancements in AI-driven cybersecurity tactics. Furthermore, it underscores the importance of adaptive, self-learning systems in the evolving landscape of cybersecurity defenses.