Optimizing Convolutional Neural Networks for Web Security With a Hyperparameter Tuning to Enhance Model Performance

Web applications are prime targets for cyber-attacks due to their network accessibility and inherent vulnerabilities. Traditional web security mechanisms such as rule-based intrusion detection systems and conventional machine learning models are mostly work on predefined rules and extensive labelled datasets which makes them less effective against evolving threats. This paper explores an optimized hyperparameter tuning approach to enhance the performance and efficiency of convolutional neural networks (CNNs) for web security. Our contributions are threefold. First, we trained the CNN model on pre-processed of unstructured and imbalanced datasets collected from open repositories. Second, hyperparameters of CNN are optimized on different constant batch sizes like 32 and 16 to improve model accuracy of CNN models. Third, the proposed models are trained again on the optimized hyperparameters values for better results of accuracy. This research highlights the potential of optimized CNN model in revolutionizing web security to provide an intelligent and autonomous solution to counteract cyber-attacks.