Collusion detection using multimedia fingerprints

Abstract

The large-scale distribution of digital multimedia over the Internet has seen steep increases in the numbers of criminal cases involving the unauthorized sharing and duplication of copyrighted multimedia content. Consequently, it is important to design reliable investigative techniques to combat unauthorized duplication and propagation, and to provide protection in the form of theft deterrence. Several fingerprint embedding schemes have been developed to combat single-user attacks. However, a new breed of attacks known as "collusion attacks" can defeat these schemes. Collusion attacks use the combination of multiple fingerprinted copies to create a new version of the multimedia artifact in which the underlying fingerprint is attenuated to render the colluders untraceable. This paper proposes a wavelet-based fingerprinting scheme and a clustering algorithm for collusion attack detection and colluder identification. Experimental results show that the scheme can identify colluders while maintaining low miss rates and false accusation rates.