Neural network 'in-jokes' could pass secrets

cryptography at summitsecurity.org cryptography at summitsecurity.org
Mon Mar 25 11:05:13 EST 2002


IMO there are a lot of tricks and traps to get from this idea to a workable
cryptosystem.  For the networks to converge they must still share some
characteristic - think of -this- characteristic as the secret key, and it
seems you have the same key distribution problem you're trying to solve.
Guessing from the sketchy details, the net's initial weights must be
logically identical for them to converge so exactly, so then anyone having
a copy of the same net should be able to decrypt the same messages as any
other.  I think the balloon's being squeezed but no breakthrough here.

At 06:45 PM 3/24/02 +0000, Will Knight wrote:
>Hello
>I'd be interested to know what people think of this story and whether
>anyone is aware of any similarly unusual encryption systems.
>Will.
>
>----------------
>http://www.newscientist.com/news/news.jsp?id=ns99992067
>Neural network 'in-jokes' could pass secrets
>
>Artificial brains could use "in-jokes" to deliver secret messages,
>according to computer scientists.
>
>The technique relies on neural networks, computer systems designed to
>mimic the brain. Just as the brain's nerve cells are wired together in a
>complex mesh, neural nets consist of a web of electrical switches, or a
>computer simulation of these connections.....
>----------------
>
>Will Knight
>http://www.newscientist.com
>151 Wardour Street London W1F 8WE
>Tel +44 (0)207 411 2688
>Mob +44 (0)7905 863 625
>PGP Key ID: 7F28A285
>
>
>---------------------------------------------------------------------
>The Cryptography Mailing List
>Unsubscribe by sending "unsubscribe cryptography" to
majordomo at wasabisystems.com
> 

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



More information about the cryptography mailing list