Deep Learning Based Network Intrusion Detection System: A Deep Abstract Networks (DANets) Model Approach

Abstract

The need for frequent updates on Network Intrusion Detection Systems (NIDS) is crucial for safeguarding enterprise data against cyberattacks due to the advancement in the attackers' arsenal resulting from technological advancements. Conventional signature-based NIDS often struggle to keep up with newer types of intrusions due to their sophisticated nature. This has forced the advancement of machine learning-based NIDS, which have shown promising results. There is much recent research on machine learning and deep learning-based network intrusion detection system. Also, there is few important research datasets like UNSW-NB15, KDD CUP dataset, etc published for the use of research community. In this paper, we aimed to assess the effectiveness of a Deep Abstract Networks (DANets) model in network intrusion detection. We used UNSW-NB15 dataset. The accuracy and F1 score of the proposed model were very promising.