On Apr 5, 2008, at 12:08 PM, Dean Willis wrote too quickly:
>
> So we either change need to change RFC 4474 (by making a signature
> "mean less" OR by allowing one to make a disclaimer on the signature),
> or we change DTLS-SRTP framework to make it nod depend on RFC 4474 for
> end-to-end communication of the signature. We do have two drafts
> (Wing, Fischer) that would accomplish this.

Sorry, I wrote too quickly and left out a bridge-construct, making the  
above more confusing than my usual rambling. And I misphrased a word  
or two. Sorry.

Let's retry:

So we either change need to change RFC 4474 (by making a signature   
"mean less" OR by allowing one to make a disclaimer on the  
signature),  or we change DTLS-SRTP framework to make it not depend on  
RFC 4474 for  end-to-end protection of the fingerprint.

We also have the issue of detecting fingerprint changes midstream,  
meaning that we've had a MITM on the media (that has replaced the  
Identity header). We do have two drafts  (Wing, Fischer) that would  
accomplish this.

The consequence of the second para above, if I understand it  
correctly, is that DTLS-Framework is relying on the RFC 4474 Identity  
header to protect the integrity of the key fingerprint which is  
transmitted in the SDP.  If the Identity header is UA-signed, this is  
pretty much rock-solid. However, if it is signed by an intermediate  
node, then it is possible that the intermediate node has altered the  
SDP (and fingerprint), and that it may be colluding in interception of  
the media. And that's just the way RFC 4474 works, so anything that  
relies on it (like DTLS-SRTP framework) has this limitation.

The point is that there may be a possibility to fix all these flaws in  
one pass. I'm not sure what solution would do this, but I have the  
feeling there's one out there that we just haven't quite identified yet.

The logic goes this way: DTLS-SRTP framework uses RFC 4474 for  
integrity protection of the fingerprint. By doing so it inherits  
weaknesses of RFC 4474. We have noted two of these as issues:

1) RFC 4474 cannot legitimately be used for calls originating on the  
PSTN and transiting a gateway to SIP because it lacks a way to inform  
the recipient that the From: field was not actually authenticated by  
the authentication service.  Otherwise said: as written, 4474 promises  
that the From: field was authenticated by the authentication service,  
and since in this case it wasn't authenticated, the service shouldn't  
be saying it was authenticated.

2) RFC 4474 headers may be subject to replacement by intermediates.  
While this might be required for certain uses of intermediates (such  
as SBCs that replace SDP because they're relaying it), it can also be  
used to insert an intermediary that intercepts the cleartext of the  
media (taps the call) without the consent of the endpoints.

We could work around by fixing these problems independently -- for  
example, by adding the From: disclaimer proposed by Adam and the  
assorted SBC-traversal and fingerprint preservation/verification  
techniques proposed by Dan, Kai, and probably others.

Or we could fix it by either fundamentally changing RFC 4474, or by  
providing a different fingerprint integrity mechanism for DTLS-SRTP  
framework and thereby breaking the dependency cycle.

--
Dean

_______________________________________________
Sip mailing list  https://www.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use [EMAIL PROTECTED] for questions on current sip
Use [EMAIL PROTECTED] for new developments on the application of sip

Reply via email to