Abstract:

With the widespread use of the Internet, network technology is used in a large amount in daily life, and the Internet and network are currently suffering from severe threats of network attacks. Network anomaly detection is one of the most significant issues in network security, and it is a core method to prevent cyber-attacks because it monitors network traffic data to figure out whether they are normal or abnormal. A variety of research frameworks have been proposed for network anomaly detection, and nowadays, deep learning-based methodologies are in the spotlight. For this reason, this research employed three deep learning models, i.e., Deep Neural Network (DNN), Convolutional Neural Network (CNN), and Recurrent Neural Network (RNN) models, with the public dataset, which is CICIDS 2017 dataset to examine their effectiveness for network anomaly detection. After evaluating the three deep learning models with the CICIDS 2017 dataset, the experimental results show that all three deep learning models show satisfactory results and have high detection accuracy, precision, recall, and F1 Score. This means that they could facilitate a more in-depth analysis of network data and identify anomalies faster. Besides, we observed that the DNN model outperformed the other two deep learning models, which achieved 98.14% of the overall detection accuracy. It proves that deep leaning models seem to be a robust potential tool for network anomaly detection in the cybersecurity field.

Keywords: