Stegdetect 0.4 released and results from USENET search available

Harald Koch chk at
Sat Dec 22 12:14:31 EST 2001

I'm nervous about (one of) the conclusions being drawn from your work.
The work itself is important, both for the stated goal (to attempt to
refute the USA Today article) and to advance the state of the art of
detection (which will in turn advance the state of the art of hiding).

While it is certainly important to try, I don't think there is any way
to disprove the rumour started by the USA Today article (and continued
in the popular press). People will believe what they want to believe, no
matter how much evidence you pile in front of them, and the media is
notorious for taking advantage of this.

Anyway, let me rephrase my concern.

Statistical research usually expends significant effort proving that the
sample data is *good* sample data. Your paper assumes it. Twice you
claim, without justification, that (paraphrased) "even if the majority
of messages were undetectable, we should still find some messages".

In my opinion, you need discuss the issue, and either support the claim
or mention it as a source of potential error.

Of course, I'm just an amateur...

Harald Koch     <chk at>

"It takes a child to raze a village."
		-Michael Fry

The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at

More information about the cryptography mailing list