Hi Jonathan,

Thank you very much for your comments.

On Sat, Mar 14, 2009 at 1:38 PM, Jonathan Rosenberg <[email protected]> wrote:

> Firstly, it seems that the easiest solution is just 'use tcp'. This is one
> of many reasons (and growing) why we need to be pushing implementations to
> tcp.

absolutely, or even better, [D]TLS.

problem is, in reality, UDP is not going away overnight.  we're far
more likely to get vendors to add a mechanism in that doesn't cost too
much in terms of implementation than expect everyone stops using UDP.

> Secondly, if you want to do this for UDP, you could instead use anonymous
> authentication. Then, the existing nonce takes on the role of a cookie. I
> suspect that anonymous authentication is not widely implemented, but neither
> is a new protocol. That said, I think the right answer still is 'use tcp'.

I did consider that.  The problem then is that it requires end-to-end
support, where as this is hop-by-hop, and as such means it doesn't
have HERFP problems or require every endpoint to upgrade - only the
hops that talk to public servers.   Via cookies also work over
multiple hops, where as anonymous would require a request-per-hop,
each request getting one hop further.  This could be very expensive in
both network load and call setup time.

> Finally, the media portion of this attack, as you point out, is far more
> disruptive. That attack does not require spoofing even; just listing the IP
> address/port of the target in the SDP of the INVITE. We were calling this
> the 'voice hammer' attack, originally documented here:
>
> http://tools.ietf.org/html/draft-rosenberg-mmusic-rtp-denialofservice-00

excellent.  i'd not come across that before.  i'll have a read tonight.

> and described in Section 18.5.1 of ICE. The via cookie mechanism you
> propose, SIP-over-TCP, or anonymous authentication, none of them fix this
> attack.

indeed - i wanted to solve the signalling side before i moved onto the
(albeit more serious) media problem

> AFAIK, ICE is the only remedy.

Yes, i came to the same conclusion; it's a shame that an ICE lite
implementation wouldn't itself fix it. (could it? :-))

There is also the problem of targeting a dialog at the victim (think
SUBSCRIBE with Contact of the victim), which is the 3rd issue.

 ~ Theo
_______________________________________________
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