A time series intra-video collusion attack on frame-by-frame video watermarking

Abstract

Current methods of digital watermarking for video rely on results from digital watermarking for images. However, watermarking each frame using image watermarking techniques is vulnerable to an intra-video collusion attack because it provides grounds to make statistical inference about information within consecutive frames. An algorithm using bootstrapped time series is proposed to exploit this vulnerability. Experimental results demonstrate that this algorithm produces a video with a significantly lower similarity to the given watermarked video using the standard watermark detector. © Springer-Verlag Berlin Heidelberg 2009.