Neural Analysis of HTTP Traffic for Web Attack Detection

Abstract

Hypertext Transfer Protocol (HTTP) is the cornerstone for information exchanging over the World Wide Web by a huge variety of devices. It means that a massive amount of information travels over such protocol on a daily basis. Thus, it is an appealing target for attackers and the number of web attacks has increased over recent years. To deal with this matter, neural projection architectures are proposed in present work to analyze HTTP traffic and detect attacks over such protocol. By the advanced and intuitive visualization facilities obtained by neural models, the proposed solution allows providing an overview of HTTP traffic as well as identifying anomalous situations, responding to the challenges presented by volume, dynamics and diversity of that traffic. The applied dimensionality reduction based on Neural Networks, enables the most interesting projections of an HTTP traffic dataset to be extracted.