On Wed, 2009-02-11 at 15:25 -0500, M. Ranganathan wrote:
> Hello!
> 
> I have been thinking about measures that can mitigate DOS attacks in 
> SIPXBRIDGE.

You're thinking of "DoS" == "Denial of Service".  
"DOS" is an old operating system :-)

> Problem Statement
> ==============
> 
> Sipxbridge is a Session Border Controller. It sits on the edge between 
> the public network and the private network and as such is a good place 
> to add DOS mitigation features. Because of the nature of the service ( 
> i.e. being able to take calls from any provider ) and the nature of SIP 
> and the function being provided by sipxbridge. 100% DOS attack 
> prevention is impossible unless we constrain inbound calls to only be 
> from trusted sources. This is clearly too restrictive for most 
> purposes.  

Clearly how?  I don't think that's clear at all...

> The purpose of this thread is to explore what can be done and 
> how to approach the problem.
> 
> Here then are *some* vulnerabilities in SipXbridge :
> 
> 1. SipXbridge will accept calls from any IP address just like your phone 
> will accept a call from anywhere. SipX will not authenticate calls 
> originating from a foreign domain. Hence anybody can call into your 
> network (just like anybody can call your phone number). This is fine if 
> that is the way you want it but unlike a PSTN network which is 
> regulated, there is no control over inbound calls. So you can have DOS 
> attacks from sources that simply flood sipx with INVITEs, or even worse, 
> INVITEs with various errors and unresolvable host names etc. and 
> SipXbridge currently will not be able to do anything about it.
> 
> 2, RTP relay can be subject to DOS attack by simply calling in, 
> transferring to the park server and never hanging up.
> 
> 3. Take over of sipx services such as call park etc. can be easily 
> accomplished if one knows what extensions  Park Orbits are  assigned to 
> for example, and an AA is configured ( a common configuration ), one can 
> call the AA, issue DTMF for the park extension and fill up the park orbit.

Question for Woof - does the AA allow that?  It probably shouldn't.

> 4. One can also flood a user's voice mail box (possibly with obscene 
> language) and do other nasty things.

Yes, but that's possible even from legitimate sources using reasonable
calling patterns.

> Some possible starting points :
> 
> 1. Accept calls only from configured ITSP accounts.
> 
> Not a very good solution. The problem of course is you have no control 
> over the ITSPs that your callers will pick. You will effectively 
> restrict inbound calls to only those ITSPs you know about and have 
> accounts with. Unfortunately we cannot authenticate inbound requests 
> from foreign domains ( defeats the whole purpose of unrestricted inbound 
> calling).

I don't understand why you think that's unreasonable.  If I get my phone
service from BT, I wouldn't expect that interface to get calls from
NTT.  

Remember: it is NOT the job of sipXbridge to be a general purpose
Internet Calling interface - it is the job of sipXbridge to be the
interface to an ITSP.  Why would I ever get a call from an ITSP that I
had no relationship with?  

If I allow direct SIP calls from the internet (to sip:[email protected]),
then those do not need to go through sipXbridge.

>  2. Black list of bad IP addresses for inbound INVITE from foreign domain.
> 
>    However, a "good ITSP" can have "bad subscribers". Can filter by 
> domain. 
> 
> 3. Rate limit INVITE from IP addresses.
> -- How to select rate?
> 
> 4. Observe the activity pattern from an IP address/user name and ban on 
> that basis.
>   -- What pattern to observe?
> 
> It is a pity that ITSP providers do not provide signed 
> P-Asserted-Identity header for calls headed to the PBX. Otherwise we 
> could do some level of filtering on that basis. Perhaps that is 
> something we need to push to SIPCONNECT 1.1

Better - all SIP should be over mutually-authenticated TLS.

> Please respond to this thread with ideas so we can discuss 1. Is this is 
> a problem worth looking at or is this all moot because it can be handled 
> at as firewall rules distinct from sipxbridge. Do we need SIP Specific 
> support in SipXbridge ? and 2.  Assuming the answer to 1 is YES, what 
> can be reasonably accomplished in the near term? What guarantees or 
> assurances can be provided to our customers?

Bear in mind that you've not touched on the most basic of DoS attacks -
ones that happen at the IP level and just hammer your interface (or the
router in front of it) so badly that the SIP can't get through.



_______________________________________________
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