Web application modeling and anomaly detection

Widespread use of Web applications and its distribution and complexity, makes web application performance analysis and anomaly detection become urgent and challenging. Previous web application anomaly detection, was more focused on intrusion detection, and made a lot more sophisticated algorithms and methods, but the performance anomaly detection is mostly achieved with the current monitoring software, which greatly limits the scope of its application. The innovation of this paper is to extract the timing sequence from web logs using stepwise regression method and establish the time consumption model of the server-side, dynamic baseline is also used to determine whether there is an exception, and to analyze whether anomaly is caused by the workload. It's proved the web log records can be obtained via web server with little cost.