Payload-Based Web Attack Detection Using Deep Neural Network

Web attack is a major security challenge in cyberspace. As web applications are usually hosted by the HTTP protocol, which is an application layer protocol, payload-based attack detection is proved to be quite effective. The payloads in a typical HTTP packet are text. Therefore, techniques such as deep neural network developed in the field of text processing can be adopted to extract the key features and detect web attacks. In the paper, we try to apply two kinds of deep neural networks, which are AutoEncoder and RNN, to figure out payload-based web attacks. Experiment results show that both networks have a very promising performance in this field.