On Thu, Feb 12, 2009 at 9:26 AM, Scott Lawrence
<[email protected]> wrote:
>
> On Wed, 2009-02-11 at 17:59 -0500, M. Ranganathan wrote:
>>
>> Currently, if I know your public IP address, I can simply send an
>> INVITE to  num...@sipxbridge-public-ip-address and right now (i.e. in
>> sipxbridge today), that call will go through. You would not need to
>> have an account with any ITSP or even be calling from a phone to make
>> a call to sipxbridge. We will need to disallow this case as you state.
>
> I consider the fact that that works to be an acceptable side effect, but
> not an essential feature.


It might in fact be an undesirable side effect ( given the fact that
it exposes the PBX to DoS attacks ) :-)


>
>> We are, however, missing some configuration soupport. We would need to
>> add some additional support in sipxconfig  for cases that do not
>> require registration but which do send INVITE from an  Inbound Proxy (
>> not the same as the  ITSP Registrar field that we have today).
>>
>> I have come across ITSPs that support redundancy where the IB proxy
>> can be one of a list of addresses (an example of such a setup would be
>> AT&T) which would require you to know all inbound proxy server
>> addresses so this would imply we need to support a list of such
>> addresses in the configuration field (AT&T does not support DNS SRV -
>> they directly use IP addresses for configuration).
>
> I assume that by "Inbound Proxy" you mean "the Service Provider proxy
> that sends calls to the PBX".
>
> Yes, _if_ you were to do this using source IP addresses as the filter,
> you'd need to know the whole list (or, better, a name that could be
> resolved via DNS to the whole list, which would let the provider change
> the list without reconfiguring the PBXs).


Correct. In particular this covers the AT&T requirements.


>
> As I said, it would be even better to insist on mutual TLS - don't
> accept an INVITE unless it comes on a TLS connection whose peer has a
> certificate that names it as the Service Provider.
>

Indeed! Given that most (all that I have tested in fact) do not even
support TCP, I think we have some way to go on that score.


>
>



-- 
M. Ranganathan
_______________________________________________
sipx-dev mailing list
[email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-dev
Unsubscribe: http://list.sipfoundry.org/mailman/listinfo/sipx-dev

Reply via email to