Network anomaly detection technique based on LSTM network and UNSW-NB15 dataset

Abstract

Cyber-attack is a dangerous attack technique used by attackers to attack information systems. There are two methods to detect cyber-attacks that are based on the sign set and based on abnormal behavior of the network data. In particular, the detection method based on abnormal behavior analysis techniques is being highly effective due to strong changes in technologies and algorithms in machine learning and deep learning. In this paper, we propose a method for detecting anomalies in networks using the Long Short Term Memory (LSTM) network and the UNSW-NB15 dataset Accordingly, based on the UNSW-NB15 dataset that was analyzed and extracted features of the network traffic, we use the LSTM network to classify into abnormal behavior or normal behavior. To see the effectiveness of the LSTM network in the problem of classifying and evaluating network behavior, in this paper, we will change the structure and parameters of the algorithm. These experimental results are presented in the experimental section of the paper.