Watermarking Scheme Capable of Resisting Sensitivity Attack

摘要

In sensitivity attacks, attempts across the boundary between "watermarked" and "not watermarked" using an available detector can provide adequate information about an embedded watermark for clearing it without causing serious distortion. This letter proposes a novel watermarking system capable of resisting such sensitivity attacks, which contains a tailor-made embedding algorithm and a corresponding detecting method. Although an attacker can still find the most effective way for destroying the watermark in a constructed signal near the boundary, it is impossible to move a watermarked product away from the "watermarked" region with low distortion. Therefore, the watermark detector can be widely distributed without fear of being used for sensitivity attacks