Traitor-tracing aided by compressed SURF image features

Abstract

Traitor-tracing (aka fingerprinting) has received much attention as a possible solution for protecting media copyrights. However, all the current schemes for images and video lack robustness against geometric attacks. We propose a novel semi-blind algorithm that can cope with such attacks. The algorithm uses compressed SURF features as side information in order to estimate and invert the geometric attack. Moreover the algorithm is highly robust against cropping and additive white noise.