A Unsupervised Learning Method of Anomaly Detection Using GRU
Zhaowei Qu, Lun Su, Xiaoru Wang, Shuqiang Zheng, Xiaomin Song, Xiaohui Song · 2018
With the development of the Internet, malicious acts of illegal access to Internet services are also growing, detection and prevention of these acts become particularly important. In order to detect illegal access, existing research not only uses rulebased and statistical-based techniques, but also uses machinelearning techniques. In the field of security, these studies have been effective in detecting abnormal traffic. However, few studies focus on detecting behavior patterns based on malicious users. Malicious behavior from users tends to have a certain pattern. In this paper, we propose an improved neural network model that uses web log data to identify user anomalies (INNAD). Our experimental evaluation shows that our detection method is more effective than traditional lstm method and svm model in anomaly detection.