Re: hacksdmi?

[Steve Mosher <farq () KILN ISN NET>]

On Thu, 12 Oct 2000, aliver vilereal wrote:

> And here is a question for the list:
> If the watermark is hidden in the least significant bits, a program could
> set all these bits to zero and this should effectively disable the
> watermark, because I am guessing that to compute the watermark an _exact_
> match must be made with the output of some decoding algorithm.  If the
> watermark is all NIL, there is no way that this could possibly match the
> watermark.  Would this have been an effective method to win the HACKSDMI
> challenge, are were they looking for an exact replica of the original, even
> though the new one with NIL bits has no audible difference?

        Really, what matters is that if zeroing the low bits would
annhiliate the watermark while leaving the audible sound unchanged, then
SDMI is useless. If they want an exact replica then they're thinking too
narrowly, since MP3s themselves cannot recreate the original (MP3s
started the SDMI effort, IIRC.)