Extreme learning machines for Web layer anomaly detection

Abstract

The idea of service oriented architecture (SOA) and the wide adoption of the cloud computing cause the rapid advancement of web applications. Also the constantly increasing expectations of end-users concerning the usability of graphical interfaces have become a driving force for new information and communication technologies. However, as new technologies, frameworks and software solutions are created, it often happens that accidentally software flaws are introduced. In many cases, those flaws may have serious implications, such as privileges escalation, server and client sides infection with the malware or sensitive data leakage. Therefore, recent cyber incidents concerning web applications show that the new countermeasures are needed in order to protect the web layer. In this paper we propose the method that adapts the Extreme Learning Machine to solve the two class classification problem in the Web Layer Anomaly Detection domain. Our experiments give promising results proving that this technique can be used to effectively detect cyber attacks targeting web applications.