eWeek: Cryptography Guru Paul Kocher Speaks Out

[Peter Wayner]

At 3:58 PM -0700 5/5/03, Bill Frantz wrote:
>It occurred to me that one way to attack a watermarking system that I
>haven't heard proposed before is instead of removing the watermark, bury it
>in other watermarks.  While this approach would degrade the quality, it
>might not degrade it enough to bother the people who trade video camera
>pointed at the screen copies of movies.


Some watermarking systems deliberately avoid this problem. For 
instance the early work (circa 95) of Ingemar Cox et al at NEC 
described a system where multiple watermarks could be written without 
overwriting each other. There were other limitations to the system, 
but it's possible to build a system where the watermarks don't "write 
over each other".

The easiest way to see this is to think about radio stations. As long 
as they're broadcasting on different frequencies, they don't 
interfere too much. If you put your watermark at one frequency, then 
another frequency won't interfere.

The problem is that the detector generally needs to know the 
"frequency". That means an attacker can often find it out and write a 
new watermark at exactly that "frequency". So there's something to 
your idea, it just requires a bit more work.



---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at metzdowd.com