Academy & Industry Research Collaboration Center (AIRCC)

Volume 10, Number 07, June 2020

A Self-Attentional Auto Encoder based Intrusion Detection System


Bingzhang Hu and Yu Guan, Newcastle University, UK


Intrusion detection systems (IDSs) have received increasing attention in recent years due to the rapid development of Internet applications and Internet of Things. Anomaly based IDSs are preferred in many situations due to their capabilities of detecting novel unseen attacks. However, existing works have neither considered the intrinsic relationships within the network traffic data nor the correlations shared among the sub features (i.e. content feature, host-based feature, etc.). In this paper, we propose a self-attentional auto-encoder based intrusion detection system, namely the STAR-IDS, to effectively explore the intrinsic structures of network traffic data and evaluated it on the NSL-KDD dataset. The experimental results show that the proposed STAR-IDS has achieved state-of-the-art performances.


Intrusion Detection System, Auto Encoder, Anomaly Detection, Self-attentional.