To solve the problem of publishing digital images, researchers have come up with digital image watermarking. This method allows the owner of an original image to add an invisible watermark to the digital image before publishing it. The watermark serves to claim copyright on the image. The owner protects the watermark with a cryptographic secret key, inhibiting anybody that does not possess the secret key from reading or even detecting the watermark. The watermark is also supposed to be robust against image tampering. Therefore anybody who wants to distribute the image further will also distribute the watermark with it, violating the copyright on the image. If the copyright holder can detect the fraud, he can prove ownership by showing that the image contains his proper private watermark.
The scheme described works well in a ``perfect'', clean-room research environment. But in the ``real world'' people play by different rules. The systems are not used as intended in the research environment and hidden back-doors are exploited. This report shows the problems that today's current digital image watermarking schemes face.
For the rest of this paper we will use the characters Alice, Bob and Mallory [Sch96] to describe copyright protection scenarios. Alice in our case is the image creator and copyright holder and Bob is the Buyer of the image. In some other settings Alice and Bob always represent the righteous characters. Mallory on the other hand is the malicious attacker and tries to forge, steal, etc.
The convention for describing the watermarking procedure is that Alice inserts a watermark into the original image. To check the image for a watermark, she can extract it.