Robust embedding and detection for multimedia fingerprinting

For the fingerprinting system, both collusion attacks and geometrical distortions are very harmful to its performance on tracing traitors. In this paper, we propose a novel multimedia fingerprinting scheme for images and videos with high robustness to geometrical distortions as well as collusion attacks. First of all, the full inverse DCT (IDCT) transform is applied on the fingerprint, and then the result is embedded into the space domain of the multimedia. This process is equivalent to embedding the fingerprint in the full DCT domain directly. Furthermore, the computing cost decreases significantly. In the detection, we propose the image correction method based on the scale invariant features transform (SIFT) feature for recovering suspicious multimedia suffering from geometrical distortions. With the proposed fingerprint embedding and robust detection, the experimental results show that the system not only resists a variety of geometrical distortions, but also can resist collusion attacks.