Re: [Anima-bootstrap] anima-bootstrap: Bootstrap proxy discovery options

["Toerless Eckert (eckert)" <[e-mail]>]

On Wed, Dec 09, 2015 at 01:32:36PM +1300, Brian E Carpenter wrote:
> I *really* don't see how the first step can be anything other than a
> recognizable unencrypted multicast probe, whether it is sent by the joiner or
> by some node that is already trusted. You could require that the unencrypted
> probe includes a public key so that a unicast response can be encrypted; but
> neither end is authenticated at that stage.

Brian:

Max'es original argument was to minimize the invention of new secure association
setup. So when you need a secure association for bootstrap, we rely on TLS for EST.
When we need a secure association for ACP/GRASP, we rely on (d)TLS (still have to
decide in detail how to do this).

If this argument makes sense to the team, then i think we do not need crypto
info in the multicast discovery, because the response will directly be the first
packets of (d)TLS for EST or ACP/GRAP - eg: nothing new needed.

The only idea i could think of was to include crypto information in the multicast
announcement not for the benefit of the security association itself, but for the benefit
of a third party observer that tries to ensure no attacker sends discoveries to prey
on defenseless unconfigured greenfield devices.

I am happy about any new valuable crypto we can put into the discovery, i just
want to make sure we can be precise about the attack vector it helps to protect
against.

Cheers
    Toerless