Char-Level Neural Network for Network Anomaly Behavior Detection

Abstract

With the rapid development of the Internet, various attacks against network servers have been increasing. At present, most of the network protection measures are mainly aimed at attacks on the network layer and the transport layer. There is almost no protection against attacks at the application layer, but more and more attacks against the web are completed through the application layer. Traditional intrusion detection methods rely too much on rule matching, and there is a problem of high false positive rate. In view of the shortcomings of traditional network intrusion detection, this paper introduces char-level neural network method into the field of Network anomaly behavior detection, and the experimental data is the http requests parsed from the collected web logs. The experiment results show that, compared to traditional machine learning models, the char-level neural network performs better in term of detecting anomaly intrusion.