Using GANs in cyberattacks and offensive security
GAN and Generative AI’s ability to understand the underlying distribution and create realistic data can be harnessed for purposes other than to create deepfakes to spread misinformation. This includes helping with black-box and brute-force attacks, which inherently rely on guessing data.
Let’s start with evading face verification.
Evading face verification
We have already discussed the ability to create composites of images of an attacker with the facial elements of the victims. This relies on face alignment and landmarks to create images that bypass popular face detection algorithms such as FaceNet and DeepFace. You can read more about these algorithms in two seminal papers:
- Facebook’s research paper in 2014, For a detailed discussion of face alignment and facial verification, provides a more in-depth review of state-of-the-art face alignment and how it works. The paper can be found at https:/...
