Yes, and Fourier transforms are well understood. So it is easy to do the transform to convert the data representation domain and then analyze to remove the 'treble' frequencies that constitute the watermark.

This is the problem with any type of data encryption (which is what this type of watermarking boils down to in essense). Anything that is mathematically stable (solvable) can be decrypted. It may take a long time. 64-bit encryption was once considered unbreakable. That was until they threw only several hundred thousand computers running as a pseudo-supercomputer at it...it only took four years. Think about how fast it would be it the number of processors were in the millions (several months maybe). 128-bit encryption will eventually be broken. It will not take 'trillions and trillions' of years - only enough parallel processing power to run through the astronomical number of choices.

When it comes to brute-force, nothing beats the human ego like a computer processor executing millions/billions/trillions of instructions per second. See Gasparov... :) That said, the best encryption scheme would be fractal. Well, some fractals are stable (Mandelbrot, Sierpinski, etc.) whereas others truly chaotic (non-linear). The problem with truly chaotic systems is that the same input yields widely varying results. Therefore, they are unstable. Encryption keys couldn't be guaranteed to be stable under such circumstances, but they'd be unbreakable (for anyone of course).

Message edited on: 11/07/2005 03:02

C makes it easy to shoot yourself in the foot. C++ makes it harder, but when you do, you blow your whole leg off.

 -- Bjarne Stroustrup

Contact Me | Kuroyume's DevelopmentZone