Enhancing Intrusion Detection with Autoencoder Based Classifier and Statistical Feature Selection

Abstract

Translate Article

In today's digital landscape, the rapid expansion of computer networks and the increasing reliance on information technology have made network security a paramount concern. With the growing sophistication of cyber threats, traditional intrusion detection systems (IDS) face significant challenges in effectively identifying and mitigating security breaches. To address these evolving threats, novel approaches that combine cutting-edge technologies are required. This paper explores the fusion of autoencoder based classifier to training and classifying the attacks of IDS. This approach is applied on the most meaningful feature that selected based on the pearson correlation (for continues vales) and chi-square test (for binary values). The benchmark NSL-KDD database is utilized to assess the validity of the suggested IDS. The experimental outcomes demonstrate that the designed Intrusion Detection System (IDS) attains superior classification accuracy at 92.27%, outperforming both prior research efforts and alternative classifier methods.