Deep Learning Techniques for Anomaly based Intrusion Detection System: A Survey

Information security has become one of the significant concerns with the advancement of technology and digital assistance. An Intrusion Detection System(IDS) plays a substantial role in guarding the systems from security threats. However, existing IDS frameworks have faced challenges such as high false alarm rate, low detection rate, raw and huge dataset handling, etc. The Deep Learning techniques has grown as a reliable methodology to address such issues. This paper presents a taxonomy of anomaly based IDS frameworks. It also includes a detailed analysis of Deep Learning algorithms used in IDS frameworks and their comparison based on different characteristics. In addition, this study indicates critical challenges of the anomaly based IDS frameworks followed by possible future directions to improve their performances.