Researchers Combat Terrorists by Rooting Out Hidden Messages

[Eugen Leitl]

On Tue, Feb 01, 2005 at 11:21:31PM -0800, Steve Schear wrote:
> At 02:07 PM 2/1/2005, Tyler Durden wrote:
>
> >Counter-stego detection.
> >
> >Seems to me a main tool will be a 2-D Fourier analysis...Stego will
> >certainly have a certain "thumbprint", depending on the algorithm. Are

Stego doesn't need to have a detectable (as telling apart from noise)
signature. If you show me how you test for stego I can show you a way to
package content that will pass that test. The problem space is similiar to
build good digital watermarks.

The difficulty is constructing a realistic-looking noise for a given set of
digital sources. Given that the tests take crunch, this will be limited to
forensics. (And one would wonder why the turdorrists smart enough to use
steganography wouldn't use really good cryptographic file systems).

And any idiot knows successful terrorists don't use crypto.

> >there certain images that can hide stego more effectively? IN other words,
> >these images should have a lot of spectral energy in the same frequency
> >bands where Stego would normally show.
>
> Images that ideal for hiding secret messages using stego are those that by
> default contain stego with no particular hidden content.  A sort of Crowds
> approach to stego.

If you have noise in the signal, can you substitute that noise with your
payload easily, or is it better to use synthetic low-noise signals, and add
your suitably encoded payload to it?

--
Eugen* Leitl <a href="http://leitl.org">leitl</a>
______________________________________________________________
ICBM: 48.07078, 11.61144            http://www.leitl.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE
http://moleculardevices.org         http://nanomachines.net

[demime 1.01d removed an attachment of type application/pgp-signature]