A deep auto-encoder based light GBM approach for network intrusion detection system

Abstract

With the development of the network in recent years, cyber security has become one of the most challenging aspects of modern society. Machine learning is one of extensively used techniques in Intrusion Detection System, which has achieved comparable performance. To extract more important features, this paper proposes an efficient model based Auto-Encoder and LightGBM to classify network traffic. KDD99 dataset from Lee and Stolfo (2000), as the benchmark dataset, is used for computing the performance and analyse the metrics of the method. Based on Auto-Encoder, we extract more important features, and then mix them with existing features to improve the effectiveness of the LightGBM (Ke et al., 2017) model. The experimental results show that the proposed algorithm produces the best performance in terms of overall accuracy.