[6399] in cryptography@c2.net mail archive
Blue Spike and Digital Watermarking with Giovanni
daemon@ATHENA.MIT.EDU (Eugene Leitl)
Sat Jan 15 21:14:26 2000
From: Eugene Leitl <eugene.leitl@lrz.uni-muenchen.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <14465.10158.755432.928553@lrz.uni-muenchen.de>
Date: Sat, 15 Jan 2000 18:06:38 -0800 (PST)
To: Eugene Leitl <Eugene.Leitl@lrz.uni-muenchen.de>
Cc: cryptography@c2.net
In-Reply-To: <14463.50929.123629.27211@lrz.uni-muenchen.de>
Eugene Leitl writes:
> This assumes the location and the amount of tweaked bits is fixed. As
> long as I stay under perception threshold, and mix the watermark bits
> with noise bits, using some wild steganographic/broadband encoding
> thing, you need a lot of copies to recover the pristine source.
>
> [Nope. First of all, you don't need to recover pristine source -- you
> just have to destroy the watermark. Second of all, good statistical
A watermark can include just a few ten bits, spread over the entire
stream using intrinsically redundant broadband encoding techniques. If
we're talking about a video, you can hide a lot there. Minute-long
audio is tougher, and single pictures very tough.
> arguments can be made for why you don't need to get that many texts
> even given no knowledge of the watermark system. I'll post more if
> pushed --PM]
Please do, this sounds interesting.
> I also haven't heard about any system which does work reliably. All I
> know is that Joe Sixpack won't be able to strip the watermarks, and
> hence be intimidated into not giving his buddies a few copies.
>
> [Joe Sixpack has nothing to to lose and almost no odds of being caught
> giving away two or three copies... --PM]
The point of watermarking is that you can personalize each piece,
linking it to the customer's identity. Percolation of warez through
buddy networks would eventually reveal original purchaser.
Of course this is unlikely to be implemented, but in theory it's doable.
