On Sat, 21 Jul 2001 jamesd@echeque.com wrote:
In principle, it should be possible to write a stego program that is undetectable, provided your enemy has no better models of noise sources in the medium than you have. As far as I know, no one has done this.
This is a point I raised on a watermarking list a while back -- most of the stego work today is aimed at watermarking/content protection applications, since that's where the money is. Those applications do not have the sort of strict demands on deniability that stego used for secret communication has. Instead of statistical transparency, they aim at perceptual. Instead of strict deniability, they go for robust detection and difficult removal, which imply easily caught redundancy in the output. Hence, most of the steganographic algorithms out there are completely unsuitable for cypherpunkly use, even when information theory posits steganography squarely as the kind of race-in-statistics you describe. Sampo Syreeni, aka decoy, mailto:decoy@iki.fi, gsm: +358-50-5756111 student/math+cs/helsinki university, http://www.iki.fi/~decoy/front