A Comparative Study of Using Deep Learning Algorithms in Network Intrusion Detection

Abstract

This study introduces a deep learning approach for network intrusion detection (NIDS), which excels in both binary and multi-classification tasks. This approach combines the strengths of six distinct deep learning algorithms: DNN, CNN, RNN, LSTM, GRU, and a Hybrid CNN-LSTM architecture. The NSL-KDD dataset, a widely recognized benchmark for intrusion detection research, was utilized for implementation and evaluation. In binary classification, the approach demonstrates exceptional capabilities, with the GRU approach outperforming others. Similarly, the DNN, LSTM, CNN, and RNN approaches exhibit robust performance, showcasing their efficacy in detecting anomalies within network data. In the multi-classification setting, the DNN approach stands out with outstanding performance. While other approaches, including RNN, CNN, LSTM, GRU, and the Hybrid CNN-LSTM approach, also maintain commendable results, the DNN approach proves to be the most effective in handling complex network patterns. This research provides valuable insights into the application of deep learning approaches using the NSL-KDD dataset for network anomaly detection, emphasizing their versatility and reliability across different classification scenarios. The findings lay the groundwork for further exploration and utilization of deep learning methodologies in enhancing network security.

Keywords

• Network security
• anomaly detection
• NIDS
• deep learning algorithms
• NSL-KDD dataset
• binary classification