Adversary malware generation: the end of detection?
Generative adversary networks (GANs) are architectures based on neural networks that can generate adversary samples. These samples are a form of malware modified to appear as harmless software. As a result, classifiers often fail to detect them, allowing malicious code to bypass security controls and operate undisturbed in networks and devices.
This emerging technology presents significant challenges for both the research and industry communities involved in malware analysis. It also complicates forensic investigations following cyber attacks or intrusions, making it more difficult to track and analyze malicious activity.
In this talk, we will explore the current state of the art of adversary technologies, discuss possible defense strategies, and consider future scenarios they may generate.