Neural Network Based Web Log Analysis for Web Intrusion Detection

With the increased attacks of web servers and web applications, it is urgent to develop a system to detect web intrusions. Web log files are stream data recording users’ clicks behavior during surfing the Internet. By carefully analyzing these log files, we can reveal some potential anomalies or attacks so as to reduce the loss of property. A method, that applies neural network method to web intrusion detection based on web server access logs, is proposed in this paper. Before feeding the raw log files into neural network algorithms, we need to preprocess these text files and make sure processed logs are of good quality with less noisy and errors. At the result part, our evaluations also demonstrate that the proposed method is superior to decision tree classifier, which shows neural network method can be transplant to web intrusion detection effectively.