Authors
Assistant Professor, Faculty of Information Technology, University of Bisha, Kingdom of Saudi Arabia
Abstract
Cyber-attacks have grown more sophisticated, posing critical threats to national infrastructure and data integrity. Deep learning methods, particularly Artificial Neural Networks (ANNs) and Reinforcement Learning (RL), offer promising capabilities in cyber threat detection and response. However, existing approaches face limitations in scalability, data sufficiency, and real-time adaptability. This study conducts a comparative analysis of ANNs and RL models to assess their effectiveness in analyzing cyber-attacks within the Saudi National Cybersecurity Authority. Using a descriptive analytical approach, data were gathered from 105 experts via questionnaires and 10 specialists through interviews. Results reveal that both models exhibit high performance in detection accuracy, adaptability, and operational efficiency. Nevertheless, challenges persist, including data dependency and interpretability. The study contributes by (1) evaluating the current implementation of ANNs and RL in Saudi cybersecurity infrastructure, (2) identifying practical and technical limitations of each model, and (3) recommending adaptive, hybrid strategies for enhanced cyber defense using deep learning.