TR-IIS-03-012    PDF format

Abstract

Robustness is a critical requirement regarding the practicability of a watermarking scheme. Current watermarking methods usually claim a certain degree of robustness against those attacks that aim to destroy the hidden watermark at the expense of degrading the quality of media data. However, there exists a kind of watermark-estimation attack (WEA) such as the collusion attack that can remove watermarks meanwhile the attacked data can be made further transparent to its original. Another kind is the copy attack that can create the protocol ambiguity problem within a watermarking system. The motivation of this paper is dedicated to cope with the WEA that is clever at disclosing hidden information for unauthorized purposes. To this end, we begin by gaining an insight into the WEA that leads to formal definitions of optimal watermark estimation and perfect cover data recovery. Subject to these definitions, content-dependent watermark (CDW) is proposed to resist watermark-estimation attack. The key point is to introduce a media hash as a constituent component of the CDW. Mathematical analyses and experiment results have consistently verified the effectiveness of the content-dependent watermarking scheme. To our knowledge, this anti-disclosure watermark is the first study that has taken resistance to both the collusion and the copy attacks into consideration.

¡@

Keywords: Collusion attack, Copy attack, Content-dependent watermark, Robustness, Watermark estimation