[VOIPSEC] Using SRTP for University project

[Randell Jesup]

"Hadriel Kaplan" <HKaplan at acmepacket.com> writes:
>> Transparent proxies can be detected by using a tunnel to somewhere outside
>> of the provider's network.  Paranoid users might connect via SBC, and do a
>> verification connection/message over a tunnel link to check response
>> headers.
> 
>I'm curious what you mean by "transparent" SBC?  And who this tunnel link
>would connect to, and what it would check?  Enterprise SBCs work differently
>than service-provider ones, in general. (if by "transparent" you mean inline
>like a firewall) 

I mean transparent in the same manner as a transparent HTTP proxy, which
intercepts port 80 transactions.  This would have to exist close to the
end-user, such as associated with or in a cable CMTS or DSL/etc router, etc
-- or it could exist on the service-providers' connection, which would be
hard (but not impossible) to do without the agreement of the service
provider.  It would have to act as a kind of B2BUA (with IP spoofing), and
it would have to be in a position to filter all traffic in both directions.

> The person can call someone they know through the local
>"proxy" and see if it's a B2BZA by doing a verbal check.  But it means
>nothing about the other people they're going to call - they could be the
>ones with the local B2BZA.  Wouldn't such a paranoid user just do the verbal
>check with the person they really want to call, every time?  

The problem with the "B2BZA" intercepting everything is that while it would
work at first, if the user ever verifies verbally via ZRTP the keychain,
they'll discover the MITM.  One could assume they would quickly discover
that _all_ transactions were being MITMed.  The reality would probably be
that a B2BZA would only insert itself into first-time connections that
matched some criteria worth risking (eventual) exposure of the B2BZA -
probably a combination of the caller and callee and perhaps otehr factors
(like a list of "suspicious" callers, patterns, destinations, web of
connections, etc).  A B2BZA (if monitored in realtime by a human) could
also partly hide it's existance in the face of a key-check by disconnecting
the call or causing sufficient packet loss to make the audio
non-understandable, possibly followed by making the network path
unreachable or by blocking ZRTP entirely.  This probably wouldn't work very
well or for very long, except for not-too-clued-in users (and many
criminals/etc are pretty dumb, especially about technology).

>-hadriel
>p.s. I'm not opposed to ZRTP - the original discussion was about s/mime.  I
>actually like many of ZRTP's properties - not the least of which being the
>IP addresses/ports of RTP at both ends aren't included inside the check (I
>think).  It's about time someone realized IP addresses need not be a
>component of end-to-end identity.  But it's too heavy for media gateways I
>think, and I've been told most sip calls go to media gateways today. 

That's since most are currently PSTN calls; that will change.  People don't
tend to worry about voice being intercepted (much) by random people, but
they do start worrying when they know images are being transmitted - that's
part of why we use SRTP for all video calls and why we have a physical
shutter on our camera as well.

-- 
Randell Jesup, Worldgate (developers of the Ojo videophone), ex-Amiga OS team
rjesup at wgate.com
"The fetters imposed on liberty at home have ever been forged out of the weapons
provided for defence against real, pretended, or imaginary dangers from abroad."
		- James Madison, 4th US president (1751-1836)