An anomaly detection method to detect web attacks using Stacked Auto-Encoder

摘要

Network borne attacks are currently major threats to information security. Enormous efforts such as scanners, encryption devices, intrusion detection systems and firewalls have been made to mitigate these attacks. Web application firewalls use intrusion detection techniques to protect servers form HTTP traffic and, Machine learning algorithms have used based on anomaly detection in these firewalls. In this work, we proposed a method based on the deep neural network as feature learning method and isolation forest as a classifier. We compared this method with the methods does not include feature extraction models on CSIC 2010 data set. Additionally, we applied different activation function and learning for our deep neural network. Results show that deep models are more accurate than methods do not have feature extraction.