Hi, Reflecting on this discussion and the adoption of the related draft, if the draft is adopted I would suggest that before it goes forward an auditing section should be added to security considerations so that appropriate advice is given in the case a member of a group is compromised. I could envision several mechanisms to address that concern. Here are a few examples that would need to be considerably fleshed out:
* An audit system is admitted to the group and records IP addresses
and times of messages received. This may be compared to IPFIX
records or packet logs to determine the veracity of sources.
* Lower level source validation may be employed to determine the
veracity of a source IP address. This validation might take the
form of IPSEC-AH.
* Endpoints might log transmission and receipt of messages so that
information may be compared.
Other approaches may be employed as well.
On additional authorization methods, one approach to protect group
membership would be to use a well known L3 multicast address and a
separate application port for communications, thus permitting normal
network filtering based on pre-authorization of the device at lower layers.
The combination of these mechanisms should at least limit potential risk
and the threat surface.
Eliot
On 7/28/16 4:02 PM, Michael Richardson wrote:
> Michael StJohns <[email protected]> wrote:
> > On 7/27/2016 5:56 PM, Michael Richardson wrote:
>
>
> > Kathleen Moriarty <[email protected]> wrote:
> >>> Mohit Sethi <[email protected]> wrote:
> >>> > designed/developed/specified for their use-case. I could definitely
> >>> > see some IoT startup building a solution that switches on the lights
> >>> > in a room as soon as you unlock the door (thus keeping them in the
> >>> > same group).
> >>>
> >>> Or perhaps more usefully, turning the lights (and the oven) off when
> you
> >>> leave the house.
>
> >> Good points, but you could do this without them being in the same
> >> group with some controller that managed the interactions with each.
> >> This would be a good set of examples for the security considerations
> >> sections, providing guidance to use a controller rather than group
> >> keys to perform useful functions like these.
>
> mcr> I agree.
> mcr> Perhaps we could convince Mike St.Johns to write that section?
>
> msj> Probably not - as long as symmetric key group communications is used
> msj> as a control protocol.
>
> Well, who other than Nixon could go to China?
>
> mcr> And ACE has the right mechanisms to make this work well.
>
> msj> I agree - public key systems. But that seems to be out of scope here.
>
> I'm not convinced. What I have taken home is that people think they want to
> use symmetric keys, and perhaps it might be safe among completely equivalent
> devices. I take your point (strongly) that this bubble will be broken, with
> catastrophic results. We need asymmetric methods between bubbles, and we
> need to define that early.
>
> mcr> We should also consider whether we can use hash-chains, like S/KEY
> did, to
> mcr> authenticate messages that should only go out in emergencies. Such
> messages
> mcr> would clearly *need* to be multicast, but once used, they can never
> be
> mcr> reused. They can't be originated by more than one sender though....
> so it's
> mcr> really the message stored by the "EMERGENCY STOP" button.
>
> msj> That's an interesting idea - but AIRC, hash chains could be
> originated
> msj> by anyone that held the signing/verification key.
>
> Yes, provided they distributed the initial (asymmetrically signed) h^n(k)
> value out in advance, and gave all the devices enough time to verify that
> signature. Plus flash to store the n-th hash.
>
> For the single sender/controller, multiple receiver/actuator situation this
> would be almost as fast as the symmetric group key, yet much more secure.
>
> For the multiple sender situation, you need to replicate things n-times.
> But it's still not n*m keys.
>
> msj> Let's do this the right way. Let's not bow to the "tyranny of the
> msj> light switch" in accepting solutions that are NOT secure, even for
> the
> msj> limited scope they are proposing.
>
> +1.
>
> --
> Michael Richardson <[email protected]>, Sandelman Software Works
> -= IPv6 IoT consulting =-
>
>
>
>
>
> _______________________________________________
> Ace mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/ace
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Ace mailing list [email protected] https://www.ietf.org/mailman/listinfo/ace
