Re: [stir] DKIM-like key mgmt approach - MITM

[Hadriel Kaplan <[e-mail]>]

On Jun 12, 2013, at 6:14 AM, Stephen Farrell <[e-mail]> wrote:

> One concern I have though is whether or not it'd be ok to leave
> open potential spoof calls launched from e.g. a zombied home router
> to a callee on the home network. If the final model had exactly
> the same security properties as DKIM, then that would be an issue.
> If the final model were based on Brian's 5280-based ideas that
> threat would probably not be an issue. Or there may be ways to
> look up a signer public key over a TLS channel that'd work too.

I understand what a "zombied home router" is and what a "callee on the home network" is, but I don't understand the case you're talking about.  Can you describe the scenario for how the attack would work/behave?

One type of attack that we might have to worry about, that might be called a MITM attack because the attacker becomes a type of MITM during the attack: a malicious callee receiving a call and using the received signature to initiate an outbound call during the short time the signature is valid, spoofing the original caller.  

This is described in:
http://tools.ietf.org/html/draft-kaplan-sip-baiting-attack-00

Basically it boils down to us needing to distinguish legitimate call-forwarding from malicious call-forwarding.  Unfortunately none of the possible solutions for it are very palatable, imho.

-hadriel