Harvey Newstrom wrote:
> Remember that standard graphic algorithms, such as GIF or JPEG, use
industry
> standard mathematical algorithms to compress data and smooth insignificant
> variations. It is easy to use these algorithms to predict how color
should
> have been smoothed and to detect bits that deviate from the expected
values...
####How about using high-level manipulation of the picture (before
compression) - e.g bending blades of grass, shifting the shape of clouds,
done with a graphics program. Such modifications would not (if skillfully
done) produce statistical aberrations at the level where compression occurs.
By comparing the modified picture to an unmodified original (initially
transmitted by a secure channel - a courier with CD-ROM) you could read the
message, according to some predefined rules.
----If you did not PGP encode the string which told the positions in the picture where the desired objects were found, it would be a variation on a book cipher. In the old days the spy and his home base would share a book, and send messages which were the positions of the desired words in the book. You have done the same thing with a picture.
### At last I know the name of this trick (which I reinvented when I was musing about unbreakable codes).
With a CD-ROM of pictures and the high-level modifications this (call it the wedding-album cipher ;-) ), would be not only unbreakable but also undetectable.
----
Anders Sandberg wrote:
Another approach that might work is realtime games. What if the two persons trying to communicate play Quake across the net?
### Very, very cool. Will the Big Brother start looking intently over teenagers' shoulder one day?
Based on the above, there is no way the governement could prevent terrorists from communicating, short of physically destroying the Internet. And yet 72% of the (so uninformed) public opinion apparently believe encryption controls could help detect the evil ones.
It's so sad that so many people think so little and do so much.
Rafal Smigrodzki
This archive was generated by hypermail 2b30 : Fri Oct 12 2001 - 14:40:53 MDT