Security of Copy-Control Watermarks

Digital watermarks are typically used in copy-control applications to communicate copy management information across analog interfaces. Alternatively they are used across unencrypted digital interfaces or for non-credentialed encryption systems. Compliant devices extract copy management information and prevent illicit use of copyrighted content. Attacker may try to remove watermarks, distort them beyond recognition or bypass extractor. Security challenges are so big that some researches believe that it is naïve to relay on copy-control watermarks. In this paper we will show how to improve security of watermarking system and make attacks unreliable, attacked content unappealing, and/or attack procedure too burdensome for typical consumer. First we will review our experience with audio watermarks for DVD-Audio and Secure Digital Music Initiative (SDMI) architectures. Next we will consider major attack classes, such as circumvention attacks, blind attacks, and informed attacks. Then we will propose a number of security enhancement measures. One technique is based on uncoordinated selection of hiding places between embedders and extractors, with statistical analysis of expected matches. This approach reduces the repeatability of attacks and opens up the possibility of system renewal. Other security measures are based on weak watermark detections, time diversity extraction, weight accumulation, soft decision decoding, and extraction expansion or focusing. Finally, we propose methods to conceal watermark presence, location and properties using random embedding, extraction and reporting, and random host signal modification techniques designed to counter differential analysis or scrambling attacks.