Artificial Intelligence-Based Intrusion Detection Systems for Enhancing Firewall Security and Preventing Cyber Threats

The study proposes a hybrid CNN-GRU model for intrusion detection in networks using the NSLKDD dataset, which includes various cyber-attack features. Several models, including Naive Bayes (NB), K-Nearest Neighbors (KNN), Deep Neural Networks (DNN), Recurrent Neural Networks (RNN), and the CNN-GRU hybrid, were employed for classification. The dataset was split 80:20 for training and testing, with evaluation based on accuracy, precision, recall, F1-score, and confusion matrices. The CNN-GRU model outperformed the others with an accuracy 99.69%, precision 99.65%, recall 99.69%, and F1-score 99.70%. In contrast, KNN achieved 91% accuracy, while NB and DNN performed moderately, and RNN showed the least effectiveness. Further optimization is needed to increase generalization and resilience in real-world circumstances since the CNN-GRU model exhibited potential overfitting despite its excellent performance.