Advances in Deepfake Detection and Face Presentation Attack Prevention

The field of cybersecurity is witnessing significant developments in the detection and prevention of deepfakes and face presentation attacks. Researchers are exploring innovative techniques to analyze and understand the inner workings of generative models, such as StyleGAN, to better detect and prevent synthetic media. The application of Vision Language Models and the Lottery Ticket Hypothesis are showing promise in improving the detection of physical and digital attacks. Furthermore, the use of foundation models and conditional GANs are being investigated for zero-shot face presentation attack detection and facial demorphing. Noteworthy papers include:

  • Uncovering Critical Features for Deepfake Detection through the Lottery Ticket Hypothesis, which applied the LTH to identify key features for deepfake detection and achieved high accuracy with substantially pruned models.
  • Are Foundation Models All You Need for Zero-shot Face Presentation Attack Detection?, which demonstrated the effectiveness of foundation models in achieving high performance in zero-shot face presentation attack detection with minimal effort.

Sources

Tackling fake images in cybersecurity -- Interpretation of a StyleGAN and lifting its black-box

In-context Learning of Vision Language Models for Detection of Physical and Digital Attacks against Face Recognition Systems

Uncovering Critical Features for Deepfake Detection through the Lottery Ticket Hypothesis

Are Foundation Models All You Need for Zero-shot Face Presentation Attack Detection?

DOOMGAN:High-Fidelity Dynamic Identity Obfuscation Ocular Generative Morphing

Facial Demorphing from a Single Morph Using a Latent Conditional GAN

Built with on top of