Re: A Media Distribution Problem

From: Michael Brown (see_at_signature.below)
Date: 05/31/04 

Date: Mon, 31 May 2004 18:29:58 +1200

Simon Johnson wrote:
> "Mok-Kong Shen" <mok-kong.shen@t-online.de> wrote in message
> news:c9dllf$p84$05$1@news.t-online.com...
>>
>>
>> Simon Johnson wrote:
>>
>> [snip]
>>> I'm fairly confident that even with quite drastic changes the
>>> watermark could survive. Remember that our watermark isn't war and
>>> peace.
>> [snip]
>>
>> Do you have references about really robust 'practical' watermarking
>> techniques? From the (very) little that I happen to read about
>> the topic, I have the impression that the state of the art is
>> yet not very satisfying and hence researchers are continuing to
>> do their work.
>>
>> M. K. Shen
>>
>
> Nah.. quite the opposite in fact. If you look at the 1st reference in
> my initial post it gives a proof of security of a secure stego
> channel that is resistant to changes introduced to the stegotext. I
> simply propose using this construction to watermark copyrighted
> material.

There's a difference between stego and watermarking for this case. In the
stego case, you have one plaintext embedded in one file. In this case, you
have multiple (different) plaintexts embedded in the same file. It's
concievable that an attacker will have access to several versions of the
video stream with different watermarks, at which point classical stego
doesn't work any more. I haven't looked in detail at this particular paper,
but at a brief glance it does not appear to offer protection against this
attack.

There's actually a provable (information theoretic) upper limit on the
number of bits of watermarking you can embed in an <x> bit plaintext such
that there's no gauranteed method of removing the plaintext if the attack
has <y> different watermarked sources. For example, if the plaintext is 32
bits, and the watermark is 8 bits, then if the attacker has <z> watermarked
plaintexts they have sufficient information for them to remove or alter the
watermark regardless of the watermarking method used. I can't remember the
exact formula or reference, but it was something close to exponential (ie:
if the attacker has 2<z> copies, then you can only embed 4 bits of
watermarking). Of course, the upper bound on the durability of current
watermarking schemes is significantly below this theoretical value. 

--
Michael Brown
www.emboss.co.nz : OOS/RSI software and more :)
Add michael@ to emboss.co.nz - My inbox is always open