Auto Encoder System for Intrusion Detection

Abstract

Monitoring a web application for attacks and issuing alerts when one is detected is the job of an intrusion detection system. In contrast, existing implementations are time-consuming and require a thorough understanding of security domains. A web application is an easy target for cyber-attacks due to its vulnerability and network accessibility. To begin with, we examine the feasibility of an unsupervised/semi-supervised method for detecting web attacks based on the Robust Software Modelling Tool (RSMT), which monitors and characterizes web applications in runtime automatically. In the second step, we describe how the RSMT encodes and reconstructs the call graph using a stacked denoising auto encoder. Finally, both datasets were tested using the RSMT and the results were analyzed. Little labelled data can be used to detect attacks is efficient and accurate when we use the Long Short Term Memory algorithm.