Re: [stir] current draft charter

[Stephen Farrell <[e-mail]>]

Hi Jon,

On 06/12/2013 03:28 AM, Peterson, Jon wrote:
> 
> 
> On 6/11/13 6:46 PM, "Stephen Farrell" <[e-mail]> wrote:
> 
>>
>> So based on many recent mails, I'd suggest adding "and threat model"
>> to the problem statement deliverable and s/certificate/key management/g
>> might be better so as not to preclude a non-certificate based
>> solution.
> 
> Fine with adding a threat model to the problem statement.

Lovely.

> 
> And non-certificate based key management solutions are in scope, sure.
> RFC4474bis lists in its action items studying how to point at DANE from
> the Identity-Info header, and this could extend to alternatives for
> credentials in the DNS. Ironically, back when we wrote RFC4474, we assumed
> there would be keys in public ENUM and that we'd solve the problem this
> way eventually. Lack of public ENUM (and to a lesser extent, tardiness of
> DNSSEC) back-burnered that plan. I'm personally less concerned about the
> syntax of the credentials than the authority structures behind them.

That's a fair point. However, I'd argue that the difference between
a DKIM-like and RPKI-like approach is far from just syntax. For
example, in an RPKI-like approach the callee has to verify who
delegated what to whom in order to check the signature (e.g. via
5280 NameConstraints). With a DKIM-like approach, that's done in
some un- or differently-specified way by some infrastructure (before
making the public key available) and isn't directly visible to the
callee. That's definitely not just syntax I reckon.

>> And one near-nit: I think the term "identity" generates quite a bit
>> of confusion and its generally better to talk about identifiers and
>> not identities. In other cases that's more of a deal but since we're
>> only dealing with phone numbers here its almost a nit.
> 
> I hesitate to back out the term "identity" given that it's what we've used
> for more than ten years for this. 

I'm fine with that in this case. It'll probably come up though
so just as well to try get it out of the way:-)

S.

> The header even in RFC4474 is called
> Identity. The RFC3325 header is P-Asserted-Identity. I hear what you're
> saying - this signature is over an identifier - but it's pretty clear that
> this identifier serves as an identity. The calling TN, or a derivative
> from it, is what telephony UIs show their users when alerting for an
> incoming call. Whether CNAM does it in the PSTN, or your address book does
> on your smart phone, applications translate callings TNs into proper names
> of various kind to represent callers. To secure calling TNs is to secure
> the identity of the caller.
> 
> Jon Peterson
> Neustar, Inc.
> 
>>
>> Cheers,
>> S.
>>
>> On 06/12/2013 02:02 AM, Peterson, Jon wrote:
>>>
>>> Below is the current draft of the charter, based on our prior
>>> discussions.
>>>
>>> Jon Peterson
>>> Neustar, Inc.
>>>
>>> ----
>>>
>>> Name: Secure Telephone Identity Revisited (stir)
>>> Area: RAI
>>>
>>> Chairs: TBD
>>> Area Advisor: TBD (Barnes)
>>>
>>> Mailing list: (current source-auth)
>>> To Subscribe: --
>>>
>>> Over the last decade, a growing set of problems have resulted from the
>>> lack of security mechanisms for attesting the origins of real-time
>>> communications. Many of these problems are familiar from our experience
>>> with email: bulk unsolicited commercial communications remain a
>>> challenge for the traditional telephone network largely because the
>>> source of calls can be hidden. Others are potentially more serious:
>>> voicemail hacking, impersonating banks and similar attacks are becoming
>>> commonplace. The technologies that obscure the caller¹s identity are
>>> frequently gateways between the telephone network and the Internet.
>>>
>>> SIP is one of the main VoIP technologies employed by these gateways. A
>>> number of previous efforts have attacked the problem of securing the
>>> origins of SIP communications, including RFC3325, RFC4474 and the VIPR
>>> WG. To date, however, true cryptographic authentication of the source of
>>> SIP calls has not seen any appreciable deployment. While several factors
>>> contributed to this lack of success, two seem like the largest culprits:
>>> 1) the lack of any real means of asserting authority over telephone
>>> numbers on the Internet; and 2) a misalignment of the integrity
>>> mechanisms proposed by RFC4474 with the highly interworked, mediated and
>>> policy-driven deployment environment that has emerged for SIP. The VIPR
>>> alternative, while promising, faced apparently unavoidable privacy
>>> problems and other significant deployment hurdles.
>>>
>>> Given the pressing difficulties caused by the lack of a useful identity
>>> solution, the problem of securing the origins of SIP communication must
>>> be revisited. Because SIP deployments are so closely tied to the
>>> telephone network, we moreover must investigate solutions that can work
>>> when one side of a call is in the PSTN ­ or potentially even both. This
>>> will require a two-pronged approach: one prong relying on information
>>> carried in SIP signaling; the other prong relying on forming out-of-band
>>> connections between IP endpoints that are participating in a call.
>>>
>>> The changes to the RFC4474 approach to SIP signaling must include a new
>>> capability for Identity assertions to cover telephone numbers, rather
>>> than domain names. To conform with realistic deployments, we must also
>>> study ways to rebalance the requirements for the scope of RFC4474¹s
>>> integrity protection to emphasize preventing third-party impersonation
>>> over preventing men-in-the-middle from capturing media.
>>>
>>> Finally, the solution must encompass an out-of-band means for endpoints
>>> to establish identity when there is no direct SIP signaling path between
>>> them available, due to interworking or similar factors. This working
>>> group will investigate a means for Internet endpoints to discover one
>>> another in real time to verify that a telephone call between them is in
>>> progress based on minimal evidence or configuration. This architecture
>>> will, to the degree that is possible, reuse the credentials defined for
>>> telephone numbers for the in-band signaling solution, and define a
>>> discovery mechanism that provides better than hop-by-hop security.
>>>
>>> The working group will coordinate with the security area on certificate
>>> management.
>>>
>>> The working group will coordinate with RAI area groups studying the
>>> problem of signaling through existing deployments, including INSIPID.
>>>
>>> Identity is closely linked to privacy, and frequently one comes at the
>>> cost of the other. This working group is not chartered to mandate the
>>> presence of identity in SIP requests, and to the extent feasible it will
>>> find privacy-friendly solutions that leak minimal information about
>>> calls to third parties.
>>>
>>> The working group will deliver the following:
>>>
>>> - A problem statement detailing the deployment environment and
>>> difficulties motivate work on secure origins
>>>
>>> - A revision to SIP¹s identity features to provide several fixes:
>>>     Changes to support certification for telephone numbers
>>>     Changes to the signature
>>>
>>> - A document describing the certificate profile required to support
>>> telephone numbers in certificates
>>>
>>> - A fallback mechanism to allow out-of-band identity establishment
>>> during call setup
>>>
>>> Milestones
>>>
>>> Sep 2013   Submit problem statement for Info
>>> Nov 2013   Submit RFC4474bis for PS
>>> Jan 2013   Submit TN cert profile for Info
>>> Mar 2014   Submit fallback for PS
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> stir mailing list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/stir
>>>
> 
>