On Tuesday 23 June 2015 01:37:35 Carsten Bormann wrote:
> Thiago Macieira wrote:
> > Indeed. The security team should figure out a way so that discovery can be
> > secured against eavesdropping.
> 
> We don't have a standard way to do this (at the CoAP level).
> Of course, if your network uses encryption, then you already have some
> protection.

Right, but not against eavesdropping from someone who did join the network. 
Most people these days give their network passwords to guests when they come 
to their places, instead of creating a separate network for guests.

And this is for people who did set a non-trivial password for their WiFi 
networks...

> > reply should be secure. Either way, the client sending the discovery needs
> > to list which port number it's listening for DTLS unicast replies on.
>
> That doesn't work either:  The DTLS endpoint is separate from the UDP
> endpoint the multicast was sent from -- the response wouldn't match the
> request.

Understood. That's why the discovery request needs to include the secure port 
number in the payload.

> http://tools.ietf.org/html/rfc7252#section-9.1.2
> 
> You could use two-step discovery: Only expose a single discovery
> resource on a DTLS-secured discovery endpoint in the unsecured
> /.well-known/core; then do discovery on that via DTLS.  The only thing
> you really expose is the IP address, and that is already visible in the
> packet.

This is something that John suggested: making an unencrypted discovery for a 
directory server, then securely discover services from that directory. This 
way, we solve two problems with one stone: we won't flood the network with 
discovery and wake up every single device when looking for a particular one.

> All this doesn't really answer the question where the keys for the DTLS
> connection would come from.  If you are still doing discovery, you may
> not know the other guy well enough (and v.v.) to talk to them securely.
> (DCAF might help here.)

Provisioning should happen during the onboarding process.

-- 
Thiago Macieira - thiago.macieira (AT) intel.com
  Software Architect - Intel Open Source Technology Center

Reply via email to