Re: [rtcweb] SDES vs DTLS-SRTP revisited

[Harald Alvestrand <[e-mail]>]

On 03/20/2012 04:44 PM, Hadriel Kaplan wrote:
> On Mar 20, 2012, at 6:30 AM, Harald Alvestrand wrote:
>
>> I don't get this scenario. If Alice calls Bob using two different gateways, won't she go through a credentials and fingerprint exchange with the gateways?
>> In that case, wouldn't the fingerprint belong to the gateway?
> Yes, but the statement being made was in the context of Alice calling Bob, them seeing their browsers claim DTLS-SRTP "secured" or whatever, and them being super-geeks and checking the detailed info of what the actual DTLS fingerprints were, and finding they don't both see the same fingerprints... and that they would thus believe there was either a software bug or a malicious middleman.
So we've uncovered a requirement here: When a fingerprint is shown, it 
should clearly identify what it is the fingerprint of - in this case, 
the gateway that relays the call to Bob, not Bob.
> So my point was that's not a good conclusion to jump to, since both caller and called parties can see DTLS-SRTP "secured" mode lock-icon, but with different fingerprints, and yet it being neither a software bug nor a malicious middleman.  I realize no distinction can be made between a PSTN-gateway and a malicious MitM, by design, but that's not a good thing - because it means people will simply learn to ignore the warnings generated by the browser, because from a user perspective they'll all be false positives.
So in this case, the browser needs to not warn; it needs to say 
"connected to gateway in order to reach Bob".