Deep Learning Based Anomalous Traffic Detection Research

Abstract

Since the popularization of the Internet, network viruses and various types of network traffic attacks have emerged in an endless stream, resulting in many security incidents due to malicious intent, while technological innovation and the complexity of network topology have also put forward new requirements for network intrusion detection systems. In the case of network intrusion detection facing high challenges, traditional techniques such as statistical analysis and signal processing are difficult to meet the demand for analyzing complex network structures and cannot achieve efficient detection speed and robustness standards. In order to meet the anomaly detection needs in advanced network workplaces, this paper will introduce the structure implemented by LetNet as the main attack module and Long Short-Term Memory (LSTM) algorithm. It is designed and implemented through an end-to-end deep network approach with good gains, with a view to providing certain implications in network traffic detection.