[Download postscript version]

Digimarc Software Distribution

We believe that one of the critical weaknesses of the Digimarc system is its provision of a watermark detection program. The presence of attacks that can take advantage of this was pointed out in section 3. It is simple for a user, given a watermarked image, to apply various small transformations until the watermark is no longer detected.

This problem exists because Digimarc has two conflicting goals for its watermarking system. It is used both to detect copyright images on the web, by means of a search engine, and to provide viewers of the image with information about its author.

If the purpose of the watermark is solely to detect images that are being used without permission, there is no need for anonymous users to be able to detect the watermark's presence or its contents; indeed, security is strengthened if they cannot. Instead, the user of the watermarking software should only be able to search for watermarks they have embedded with their own key. This allows them to check their own images to ensure the watermark is still present after editing, without allowing attackers to do the same. Unfortunately, because Digimarc also wants to use the watermark to provide author identification, it must distribute free software to make watermark detection and retrieval possible for anyone, thus severely compromising security.