Neural network 'in-jokes' could pass secrets

[Pete Chown]

Will Knight wrote:

> I'd be interested to know what people think of this story and whether
> anyone is aware of any similarly unusual encryption systems.

Sounds a bit reminiscent of the steganographic spam:

http://spammimic.com/

The current implementation is not keyed so it would be very easy to try
all received spam to determine whether or not it decrypts to something
useful.  Adding keying should be easy though.  It's pretty neat IMO --
if you hid messages in images, say, you might have to explain why you
had received so many.  But no one has to explain why they receive spam.

Making realistic sounding "in-jokes" might be more difficult, but if
they can come up with something convincing it will work well too.

I've been thinking about something similar in connection with the
Napster-style file sharing networks.  If I want to share an MP3, that's
easy, I just set up a web server and put the MP3 on it.  The problem is
enabling people to find it.  Napster had a centralised model, and
various decentralised search systems have followed.

My idea is slightly different.  Instead of providing a new search
facility, simply piggy-back on existing ones.  Convert the details of
the MP3 and its location to obscure phrases, put them on a website, and
get it indexed by the search engines.  People who want the MP3 do the
same thing and just locate it with a regular search engine.

-- 
Pete


---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at wasabisystems.com