From: rishab@dxm.ernet.in
To: cypherpunks@toad.com
Message Hash: f2333156c2d7c3953daf9681f65b259fc66072d4832cbd46542611c6adef00fc
Message ID: <gate.DN7qic1w165w@dxm.ernet.in>
Reply To: N/A
UTC Datetime: 1994-03-06 09:13:35 UTC
Raw Date: Sun, 6 Mar 94 01:13:35 PST
From: rishab@dxm.ernet.in
Date: Sun, 6 Mar 94 01:13:35 PST
To: cypherpunks@toad.com
Subject: No Subject
Message-ID: <gate.DN7qic1w165w@dxm.ernet.in>
MIME-Version: 1.0
Content-Type: text/plain
sdw@meaddata.com (Stephen Williams):
> on pictures with few colors and a large amount of a small number of
> colors (like cartoons, etc. with a backround), you could use the
> unused color table entries. You set these to be the same colors as
> the correspondingly largest used colors and use them when juxtaposed
> with their color.
This, and other similar methods will of course only work with GIFs; the
problem is that GIFs compress, and the loss of compression caused by adding
to the color table could be noticed, as could the fact that there are multiple
entries with the same value.
It's also somewhat more complicated for stego users to figure out where the
data is - unless you use a 'standard' header or format, which is as known to
the NSA as to cypherpunks.
What worries me more about many stego methods is their assumption that
'stegging' a picture will not be noticed. I am not so sure, even if you stick
to the LSB of an 8-bit blue channel of a 24-bit hi-res scan.
Things like the abnormal frequency of 1s as mentioned earlier, are almost as
easy to build into your stego program as to detect. I've worked a lot with
images, particularly image recognition, and have seen many surprising
similarities among pictures of some category that are human-noticable only
after considerable statistical analysis (or filtering through sophisticated
neural networks). It may well be possible to develop highly CPU-intensive
techniques that can detect, or indicate suspicion towards, 'stegged' images.
Unless, of course, equally complex stego techniques are developed, which will
probably have the extremely low message-to-image ratio of the old 'letter to
granny' codes. (Message in third letter of alternate words, etc., in innocuous,
plausible letter.)
This whole cat and mouse game with the authorities reminds me of the reaction I
had to the story in Wired 2.02 on the Stealth watchers. OK, intrepid renegades
take zoom-lens photos and listen in to secret frequencies, but only because
US authorities don't REALLY mind. Come on, if Big Brother were really serious,
Phil Zimmerman (and most of us on this list) would have been shot long ago,
democracy be damned. As it so happens, PEOPLE still vote, not SPIES.
Return to March 1994
Return to “rishab@dxm.ernet.in”
1994-03-06 (Sun, 6 Mar 94 01:13:35 PST) - No Subject - rishab@dxm.ernet.in