Worst Case Morphs

Worst-case morphs represent the most challenging examples of image manipulations designed to deceive biometric systems, particularly face recognition. Current research focuses on generating these challenging morphs using various techniques, including generative adversarial networks (GANs), diffusion models, and methods leveraging template inversion and Wasserstein distances to optimize for maximal deception. This research is crucial for improving the robustness of biometric systems against sophisticated attacks and for developing more effective detection and de-morphing methods, ultimately enhancing security and privacy in applications relying on facial recognition.