ecrit@ietf.org
[Top] [All Lists]

Re: [Ecrit] Signed LoST mappings

Subject: Re: [Ecrit] Signed LoST mappings
From: Shida Schubert
Date: Mon, 23 Jul 2007 09:35:52 -0700

Richard,

Comments inline..

Richard Barnes wrote:
Another solution would be to sign a LoST mapping response (with a service boundary). The SIP UA would then add the signed LoST mapping into the SIP message together with the location information it used for retrieving the mapping. The VSP would verify that
a) the signature of the mapping is correct.
b) compare the identity of the mapping server with a a whitelist of valid mapping servers c) verify whether the location information is contained in the service boundary.

I would make a couple of modifications:

1. Remove item (c) from the verification steps, since the location isn't really relevant (just that the destination is a PSAP).

2. Generalize item (b). It's not practical to maintain a global whitelist of mapping servers; what you'd want instead is some sort of "LoST PKI" that you would use to verify identities. You could have PKI hierarchies that correspond to tree hierarchies, with root tree nodes used as trust anchors, or you could use some FGs as trust anchors.

I like this approach, because it only gives you one thing to credential, namely LoST servers. Chances are that the LoST infrastructure is going to require strong authentication anyway (to support peering and synchronization), so this is a good way to re-use that to support authentication of emergency calls.

To summarize, then:

Assertion:
a) Include a signed LoST <mapping> element in the SIP message
The fact that you get LoST <mapping> element indicates that UA has a location to query the LoST server to begin with. In which case, UA would naturally holds the location it used for mapping thus includes the location in the SIP request which can be utilized by VSP to verify the rigidness of emergency call by simply querying the LosT server itself.

I thought you were trying to address a problem when location is not available neither for VSP nor UA while VSP wishes to verify that call indicated as an emergency call is really
an emergency call.

Use of connected-identity (C-ID) has some issues as well. AFAIK the C-ID
requires calling party's device and called party's domain to support C-ID.
It also uses UPDATE to send the C-ID and doesn't mandate when this UPDATE is sent.

So the following issues may arise
1. I can call Hannes with indication that it's an emergency call, VSP forwards the call and I establish a call with Hannes and start talking to him as long as Hannes's
     device doesn't send UPDATE with connected-identity.

2. As far as I know C-ID although can be used by VSP to verify callee's identity, it is the UAC that actually set the option tag indicating the support for C-ID which is necessary information for UAS or callee's domain to trigger the C-ID process. So if UAC doesn't indicate that it supports C-ID (Although VSP supports it) using the defined option-tag,
     C-ID won't be sent from the callee.


Verification:
a) Verify the signature on the <mapping> element
b) Verify that the signer is trusted
c) Verify that the To: address for the dialog is contained in a <uri>

Trust model:
- Trust in LoST infrastructure and LoST credentialing authorities
- Call is an emergency call if the To: address is a PSAP (as verified by a signed LoST mapping)

Pros:
- No location required
- Very similar to the location+URN use case

Cons:
- Requires credentialing of LoST servers





_______________________________________________
Ecrit mailing list
Ecrit@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ecrit




_______________________________________________
Ecrit mailing list
Ecrit@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ecrit

<Prev in Thread] Current Thread [Next in Thread>