A Performance Evaluation of Collusion Attacks in Multimedia Fingerprinting

Digital fingerprinting is an emerging technology that provides a method for protecting multimedia content from unauthorized redistribution. Collusion attack is a cost-efficient attack for fingerprinting schemes where classes of users combine their fingerprinted copies for the purpose of attenuating or removing the fingerprints. In this paper, we evaluate the performance of different collusion attacks in multimedia fingerprinting. Firstly, we use a uniform model to represent them. Then, we select a number of typical fingerprinting schemes, which have previously shown a good performance in collusion resistance, and implement a set of collusion attacks on them. Furthermore, we compare the practical attack performance using the same evaluation criteria and the same test data. Our contribution rests on that we are first to make an integrated comparison of the typical collusion attacks with practical fingerprinting schemes rather than the assumption of Gaussian-distribute signals to our knowledge.