[Download postscript version]
next up previous
Next: The Robustness of Digimarc Up: Digital Image Watermarking in Previous: Handling small images

Digimarc's shortcomings

  As we have shown in section 2 the UI of security software is very important for the overall security of the system. The article [Kat97] showed that the Digimarc UI had various shortcomings. For example the dialog box where the user can choose the watermark's robustness, the user sees immediately the relation between image quality and watermark strength. It was observed that most users chose the best image quality, which resulted in the least watermark strength. Another problem was that the users inserted a watermark and then later applied image transformations, such as scaling and compression. For obvious reasons, this weakens the watermark. The user was not aware that watermarking should ideally always be the last step before publication. For these reasons, most of the images published did not contain the watermark at all or it was too weak to extract. Therefore Digimarc failed in providing the user with a usable implementation for image copyright protection. We will see in section 5 how the system can be improved to solve these problems.

Digimarc proposed to set up a web-spider [Dig97] which would scan web-servers and search for watermarked images. This approach fails to work for the same reasons as stated earlier. The web-server simply refuses to send any stolen image to the Digimarc web-spider. It would be quite complicated for Digimarc to set up a large number of different spiders to make this scheme to work. Another problem with this approach is that it will be difficult for Digimarc to prove that the data really came from that particular web-server and was not ``injected'' into the Digimarc database by a corrupt Digimarc employee. Yet another shortcoming of the Digimarc web-spider is that payment or access controlled sites are not checked. These sites are especially interesting for the image creator as he would like to make sure that his intellectual property is not sold anywhere else. Finally Digimarc only sends a monthly report to the customer. In the worst case the image could already have been distributed one month earlier. On the Internet one month is almost an eternity.

Another issue is that any user that wants the web-spider protection has only one choice: fully trust Digimarc and give them the user's private key. There's also no alternative for which watermarking method to use. Digimarc has a considerable market advantage as their system comes directly bundled with Adobe Photoshop. The user is only left with the choice to ``use it or leave it''.

One of the greatest problems with commercial watermarking schemes is the difficulty in getting details on how these schemes work; for reasons of either competitive advantage or patent application, companies tend to keep their methods secret. This is ``security through obscurity'', a law called Kerkhöffs principle [Sch96], and means the companies are relying to some extent on the difficulty of discovering their algorithm to foil attacks, rather than on the strength of some cryptographic key. This has proved foolish in the past.



 
next up previous
Next: The Robustness of Digimarc Up: Digital Image Watermarking in Previous: Handling small images
Adrian Perrig
1/15/1998