AHT-QCN: Adaptive Hunt Tuner Algorithm Optimized Q-learning Based Deep Convolutional Neural Network for the Penetration Testing

• Published in: Cybernetics and information technologies : CIT Vol. 24; no. 3; pp. 182 - 196
• Main Authors: Railkar, Dipali, Joshi, Shubhalaxmi
• Format: Journal Article
• Language: English
• Published: Sciendo 01-09-2024
• Subjects: Adaptive hunt tuner algorithm; Deep convolutional neural network; Deep learning; Penetration testing; Q-learning
• ISSN: 1314-4081; 1314-4081
• DOI: 10.2478/cait-2024-0032

Penetration Testing (PT), which mimics actual cyber attacks, has become an essential procedure for assessing the security posture of network infrastructures in recent years. Automated PT reduces human labor, increases scalability, and allows for more frequent evaluations. Real-world exploitation still challenges RL-based penetration testing because the agent’s many possible actions make it hard for the algorithm to converge. To resolve these shortcomings, a deep learning- model named Adaptive Hunt Tuner algorithm optimized Q-learning based deep Convolutional neural Network (AHT-QCN) is developed for efficient PT. Specifically, the Q-learning employed in this model improves its efficiency by enabling optimal policy learning for decision-making. In addition, the Adaptive Hunt Tuner (AHT) algorithm enhances the model’s performance by tuning its parameters with reduced computational time. The experimental outcomes demonstrate that the developed model attains 95.25% accuracy, 97.66% precision, and 93.81% F1 score.