Resilience of luminance based liveness tests under attacks with processed imposter images

Abstract

Liveness tests are techniques employed by face recognition authentication systems, aiming at verifying that a live face rather than a photo is standing in front of the system camera. In this paper, we study the resilience of a standard liveness test under imposter photo attacks, under the additional assumption that the photos used in the attack may have been processed by common image processing operations such as sharpening, smoothing and corruption with salt and pepper noise. The results verify and quantify the claim that this type of liveness tests rely on the imposter photo images being less sharp than live face images.