[general_re]

I think I still see a hole in it. Wanna put it to the test? ;)

[VeritatisSplendor]

What's the hole?

Remember, you have to destroy the watermark, which is designed to be just below the threshhold of aural detectibility (or just above it, but below the threshhold of artistic distinguishability), without degrading the sound quality, when you don't know exactly how the watermark is encoded. If you're thinking of "averaging" a whole bunch of different versions from different physical CDs, remember that you have both AM and FM encoding to deal with as well as other tricks like phase modulation between the right and left channels....good luck synching those up.

Theoretically, the fact that you are just trying to destroy the watermark rather than read it helps you, but I don't have to make your job impossible -- I just have to make it too much trouble compared to buying the CD or settling for a pulled-off-the-radio quality product.

[VeritatisSplendor]

A final mathematical point -- if I steganographically encode 400 bytes rather than 4, that means you would have to "average" more than 100 physical CDs to obscure the origin of the information -- otherwise I'll just be able to determine ALL the individual watermarks. This requires fancy error-correction, but information theory guarantees it's possible.