RoPAD: Robust Presentation Attack Detection through Unsupervised Adversarial Invariance

For enterprise, personal and societal applications, there is now an\nincreasing demand for automated authentication of identity from images using\ncomputer vision. However, current authentication technologies are still\nvulnerable to presentation attacks. We present RoPAD, an end-to-end deep\nlearning model for presentation attack detection that employs unsupervised\nadversarial invariance to ignore visual distractors in images for increased\nrobustness and reduced overfitting. Experiments show that the proposed\nframework exhibits state-of-the-art performance on presentation attack\ndetection on several benchmark datasets.\n