Generative Adversarial Networks in Security: A Survey

Abstract

In the Information Age, the majority of data stored and transferred is digital; however, current security systems are not powerful enough to secure this data because they do not anticipate unknown attacks. With a growing number of attacks on cybersecurity systems defense mechanisms need to stay updated with the evolving threats. Security and their related attacks are an iterative pair of objects that learn to enhance themselves based upon each others' advances - a cybersecurity "arms race." In this survey, we focus on the various ways in which Generative Adversarial Networks (GANs) have been used to provide both security advances and attack scenarios in order to bypass detection systems. The aim of our survey is to examine works completed in the area of GANs, specifically device and network security. This paper also discusses new challenges for intrusion detection systems that have been generated using GANs. Considering the promising results that have been achieved in different GAN applications, it is very likely that GANs can shape security advances if applied to cybersecurity.