ACC device certificates RE: NAT Traversal

["Hallam-Baker, Phillip" <[e-mail]>]

> > nodes on your home network when one talks to another. That 
> would be a bit
> > hard to manage: a certificate for each of your machines at 
> home, possibly
> > including home appliances which are on the network.
> 
> Why is a certificate-per-machine hard to manage?

It has been made hard because people have been insisting on applying PKI
techniques designed to support authentication of humans to authenticate
devices.

With a device you can embed a private key during manufacture that is unique
to the device. We already do this with cable modems (and no the economics
are not prohibitive).

I just published a White paper on this subject on the VRSN research web
site:
http://www.verisignlabs.com/Papers/ACC1.html

The basic idea is to embed a private key in the device during manufacture,
tie the public key to the serial number of the device using a certificate
and use the combination for the SOLE PURPOSE of authenticating the device
when it applies to authenticate application keys that are generated in the
device during initialization.

If the device is decomissioned the applications keys are cleared but the ACC
key remains so that the next purchaser can initialize it. The genuinely
paranoid (i.e. the military) might have the option of paying a lot more to
install their own ACC keys

The objective is plug and play for cryptography. Every device should
initialize with the absolute minimum of fuss. This needs to be simple enough
that your granny can install it.

This does not remove the need for certs to authenticate humans. But that is
a separate layer of authentication.

		Phill