Network Intrusion Detection System Using Deep Learning to Analyze Network Traffic

This paper presents a network intrusion detection system based on deep learning using a recurrent neural network model (RNN-IDS) to analyze network traffic. The experiments were conducted using the NSL-KDD dataset, which involved preprocessing steps such as encoding non-numeric features and normalization. The evaluations were performed under two scenarios: binary classification (normal and anomaly) and five-class classification. In binary classification, the model achieved strong performance, with a test accuracy of 83.28% and a training accuracy of 99.81%. In the five-class classification task, the model achieved an accuracy of 81.29% on the test set, outperforming other common machine learning algorithms. Furthermore, comparison with an RNN-based dimensionality reduction method showed that the proposed model exhibits stronger modeling capabilities and faster detection speed. The model's training time without GPU acceleration was approximately 1765 seconds. The results indicate that employing deep learning in network intrusion detection can significantly enhance accuracy and efficiency. Future research can explore more advanced recurrent neural network architectures, such as Long Short-Term Memory (LSTM) networks, which excel at capturing long-term dependencies within sequential data. Incorporating these architectures is expected to further boost classification performance and robustness, thereby strengthening defenses against increasingly complex cyber threats.