TY - CONF T1 - Colluding Fingerprinted Video using the Gradient Attack T2 - Acoustics, Speech and Signal Processing, 2007. ICASSP 2007. IEEE International Conference on Y1 - 2007 A1 - He,Shan A1 - Kirovski,D. A1 - M. Wu KW - attack;multimedia KW - attacks;digital KW - content KW - data;video KW - distribution;fingerprint KW - effort;gradient KW - fingerprinted KW - fingerprinting;disproportional KW - fingerprints;colluding KW - fingerprints;Laplace KW - Gaussian KW - identification;multimedia KW - of KW - processing; KW - protection;unauthorized KW - signal KW - spectrum KW - spread KW - systems;security KW - video;collusion AB - Digital fingerprinting is an emerging tool to protect multimedia content from unauthorized distribution by embedding a unique fingerprint into each user's copy. Although several fingerprinting schemes have been proposed in related work, disproportional effort has been targeted towards identifying effective collusion attacks on fingerprinting schemes. Recent introduction of the gradient attack has refined the definition of an optimal attack and demonstrated strong effect on direct-sequence, uniformly distributed, and Gaussian spread spectrum fingerprints when applied to synthetic signals. In this paper, we apply the gradient attack on an existing well-engineered video fingerprinting scheme, refine the attack procedure, and demonstrate that the gradient attack is effective on Laplace fingerprints. Finally, we explore an improvement on fingerprint design to thwart the gradient attack. Results suggest that Laplace fingerprint should be avoided. However, we show that a signal mixed of Laplace and Gaussian fingerprints may serve as a design strategy to disable the gradient attack and force pirates into averaging as a form of adversary collusion. JA - Acoustics, Speech and Signal Processing, 2007. ICASSP 2007. IEEE International Conference on VL - 2 M3 - 10.1109/ICASSP.2007.366197 ER -