Re: [fw-wiz] CERT vulnerability note VU# 539363

From: Paul D. Robertson (proberts@patriot.net)
Date: 10/16/02 

From: "Paul D. Robertson" <proberts@patriot.net>
To: Mikael Olsson <mikael.olsson@clavister.com>
Date: Wed Oct 16 09:21:05 2002

On Wed, 16 Oct 2002, Mikael Olsson wrote:

> Although this is something that people need to keep in mind when
> picking / designing a firewall, I'd argue that anything north of
> a stateless packet filter is going to be vulnerable to these sort
> of attacks.

So will anything south of a firewall- hosts aren't immune to flooding
attacks either, with our without state, nor are routers...

> If you keep state, you will be vulnerable to state table overflows.

I don't know that "overflow" is the right word here, "exhaustion" seems
more fitting.

When I first looked at this, I kind of shrugged and said "So what?" the
firewall is doing its job- stopping packets when there's an attack-

The issue is more of how easy that is, than the fact that you can do it.
So the real message here is know the failure mode and capacity limits of
the firewalls you use.

If you're being attacked, the firewall not allowing new traffic is
probably not a bad thing. For most folks, the ability to create a new
state table entry is an "outbound traffic only" issue for firewalls that
aren't "protecting" external services like Web servers.

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.

Public-talking hosts should be protectable with simple non-stateful packet
filtering rules- *especially* those which allow the untrusted side to
initiate connections.

> Period. The only real question is: how much work does the attacker
> need to put in before it becomes painful for the networks that the
> firewall is protecting? Is being able to resist a 1 Mbps stream
> (~4500 pps) "Not vulnerable"? Is being able resist a 34 Mbps stream
> (~150 kpps) "Not vulnerable"? Or should every single firewall
> vendor report in and say "Vulnerable", and describe what the limit is?

That's also a scale thing- if a firewall is in front of 10 hosts, the
effort to protect them from floods might be scalable to 10x, but if it's
1000 hosts, the ammount of protection is amost certainly going be be less
than 1000x[1].

> And, yes, ALG-only firewalls can also be overloaded. It's just a
> different type of 'state'.

Anything without some sort of artificial rate limiting can be DoS'ed.

What this really says to me is "Don't keep state on stuff that doesn't
*need* it." Possibly combined with "if you have a large number of
untrused users, make sure your policies let you disconnect them if they
cause trouble, and have enough diagnostic infrastructure to be able to
figure out where an internal attacker is (personally, I prefer
losts of routers.)"

As far as ALG state goes, at least the ALG is the final determination
point for the traffic, so it can deal with many issues (such as the CRC
one detailed in the vulnerability note) immediately, rather than having to
rely on a network reply from a host, or in some cases, just not knowing.

Paul
[1] There are products which scale higher, but they tend to be the kinds
of things you put in front of DSLAMs and cost several hundred thousand
dollars each.
-----------------------------------------------------------------------------
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

Relevant Pages

  • RE: [fw-wiz] CERT vulnerability note VU# 539363
    ... So will anything south of a firewall- hosts aren't immune to flooding ... firewall is doing its job- stopping packets when there's an attack- ... aren't "protecting" external services like Web servers. ... one detailed in the vulnerability note) immediately, ...
    (Firewall-Wizards)
  • RE: [fw-wiz] Vulnerability Response
    ... >>two evolving solution spaces that solve real problems. ... > management effort scales with the number of hosts. ... change control is an _enemy_ when talking about rank and file ... but not even the mjr perfectly secure firewall will work ...
    (Firewall-Wizards)
  • Re: Using netmask ffffffff
    ... The most important thing these new hosts need is connection to the outside world, for internet browsing, webmail access, fetch some documents from remote sites they forgot to bring with them for the conference, etc. ... the new hosts should not be able to directly contact each-other or the majority of my internal network. ... The trouble is that even if I set-up firewall rules to filter their traffic, they can still communicate behind the firewall directly through the switch they are all connected to, as only their internet traffic will go through the firewall. ...
    (comp.unix.bsd.freebsd.misc)
  • Re: XP vulnerabilities?
    ... Note that I also questioned your use of the "Corporate Edition" of Windows. ... If you were indeed running a network of 5 or more hosts for which you ... firewall host running the firewall software through which all your intranet ... export their rules so you can migrate them easily to another host, but NIS ...
    (alt.computer.security)
  • Re: HELP ! ipfw et natd
    ... > So the problem for me was to remark that the DNS of my IPS (193.252.19.3 it ... I don't think the nameserver's IP changed because of the firewall. ... Propagation of the change to your LAN hosts is another thing. ... well) and pointing the LAN hosts to the FreeBSD box as their nameserver. ...
    (comp.unix.bsd.freebsd.misc)