On 10/14/01 3:48 PM, "Harvey Newstrom" <mail@HarveyNewstrom.com> wrote:
> James Rogers wrote,
>> Actually, I'm having difficulty fitting the "adding noise makes
>> it easier to
>> eliminate noise" concept into the algorithm spaces that I am familiar with
>> that deal with these very things.
>
> I think you mistake my meaning. I am not talking about adding noise. I am
> talking about analyzing existing "noise" to see if it is a hidden message.
>
> When I speak of "more noise", I mean we have more of a sample to analyze.
> If the "noise" is a hidden message, the larger percentage it takes up, the
> easier it becomes to detect. Increasing this kind of "noise" makes it more
> detectable if it is a message, or less likely to be a message if it does not
> become more detectable.
>
> I am merely saying that having more data makes statistical analysis easier.
Oh, okay. *This* makes sense now. :^)
However, wouldn't a legitimate counter-measure to this be to dither all your
data except the ones that you hide data in? If you matched your
steganography package to a dithering package, it might be very effective at
hiding the data. There are many different dithering algorithms around these
days (mostly because the hardware is better than the data formats); I'll bet
you could piggyback data in the noise function for at least some of them.
Hmmm... <hamster wheels spinning>
Cheers,
-James Rogers
jamesr@best.com
This archive was generated by hypermail 2b30 : Sat May 11 2002 - 17:44:13 MDT