Re: CERT vulnerability note VU# 539363

["Paul D. Robertson" <proberts () patriot net>]

On Wed, 16 Oct 2002, Daniel Hartmeier wrote:

> On Wed, Oct 16, 2002 at 09:36:06AM -0400, Paul D. Robertson wrote:
>
> > If you're hosting public resources behind the same firewall that's 
> > protecting everything else in your enterprise, you've probably made a 
> > questionable architectural decision.  If you're keeping state on say 
> > inbound SMTP traffic, the question is "Why?"  If the 'Net as a whole can 
> > connect to something, the state itself isn't going to do much good.  If 
> > you're trying to rewrite sequence numbers because of a host that talks to 
> > the public with high predeictability, again you're probably made a 
> > questionable architectural decision.
>
> Keeping state can have performance benefits. Depending on your rule set,
> associating a packet with a state entry is cheaper than evaluating the
> rules. Keeping state does not 'just' increase the quality of filter
> decisions.

Ok, I can see that if you're handling less stateful entries than you have 
rules, but with good rule ordering, or a busy site, I'm not sure it's a 
gimme.  Do you have any way to measure which is better, or threashold 
information?

> > Public-talking hosts should be protectable with simple non-stateful packet 
> > filtering rules- *especially* those which allow the untrusted side to 
> > initiate connections.
>
> In my experience, allowing to specify a maximum for the number of states
> created by a filter rule is very useful in this case (if you want to
> keep state on all connections, and everything passes through the same
> firewall). While an attacker can exhaust the individual maxima for
> incoming connections to different services, other kinds of connections
> (like outgoing connections, or connections the attacker can't establish)
> are not affected.

If you can limit the connection rate- I've never been in a position where 
that was overly necessary- I kept Web servers away from firewalls behind 
screening routers, and tuned the stack of my SMTP gateway to handle 
whatever it could without dropping legitimate connections- you could rate 
limit services with QoS as well- that just moves the issue from the 
stateful filter and its buffers to the router's buffers though.  Unless 
you can just reject the traffic.

Rate limiting is an interesting application of a state engine though, and 
certainly one I hadn't thought about much.  The issue here however is that 
rate limiting creates a DoS window.

How likely is an exhaustion attack which doesn't turn into a complete 
flood which brings down the other services?

I think for non-malicious stuff, rate limiting by state may be 
interesting, but I think in the face of a malicious attack, it's probably 
ultimately less useful than it seems on the surface (assuming a 
relatively normal architecture, and not a hydra of connections and 
address spaces.)

Paul
-----------------------------------------------------------------------------
Paul D. Robertson      "My statements in this message are personal opinions
proberts () patriot net      which may have no basis whatsoever in fact."
probertson () trusecure com Director of Risk Assessment TruSecure Corporation

_______________________________________________
firewall-wizards mailing list
firewall-wizards () honor icsalabs com
http://honor.icsalabs.com/mailman/listinfo/firewall-wizards