A Practical Comparison of Deep Learning Methods for Network Intrusion Detection
Tran Hoang Hai, Le Hai Nam · 2021 International Conference on Electrical, Communication, and Computer Engineering (ICECCE) · 2021
Cybersecurity is essential nowadays due to the vast development of Internet leading to miscellaneous attacks to the cyber systems. Numerous measures have been proposed to prevent those attacks including Intrusion Detection System (IDS). In response to the rise of both quantity and quality of the invasion, the IDS has been constantly enhanced, and integrating knowledge base into existing IDS is the solution of interest. Among the implementations of such solution, Deep Learning is highly appreciated and has the potential to expand in the future, showing higher accuracy detecting network anomalies in comparison with Machine Learning. In this paper, a comparison is drawn between the efficiency of three novel Deep Learning approaches: RNN (Recurrent Neural Network), LSTM (Long short-term memory) and GRU (Gated Recurrent Unit), tackling the problem of anomaly detection on two datasets CICIDS2017 and CSE-CICIDS2018.