Bugtraq mailing list archives

Re: Checkpoint SYN DoS Vulnerability


From: Jim Clausing <clausing () ieee org>
Date: Thu, 18 May 2006 00:07:44 -0400 (EDT)

Bojan actually makes a good point here.  Is it possible you are filling up 
the connection table during the scan?

--
Jim Clausing
GCFA, GCIA, GCFW, GSIP, GSOC, GREM, CISSP, CCSA
GPG fingerprint = EBD0 F967 3B1C 9EA6 79AD  8939 978A 079C 8BAB F921

On or about Wed, 17 May 2006, Bojan Zdrnja pontificated thusly:

Sanjay,

On 5/17/06, sanjay naik <sanjaynaik () hotmail com> wrote:
Pawel,

We have done a complete test using TCPdump on the checkpoint side and
Tethereal on the scanner side. We have tested this on atleast 3 dfferent
firewalls and found the same issue with our scans.

SYNdefender is disabled on the Nokia/Checkpoint firewall. Nokia's
response
after seeing the results of the scan has been that SYNdefender is still
functional even if we disable it and valid authorized scans won't be
allowed
from the firewall as that is a product limitation!

I don't agree this is a feature as that would be absurd. SYN Attack
Protection is not enabled on the firewalls. The scans are being done from
the Internal interface of the firewall and not the external interface.
The

From my experience with Checkpoint NGX, the Smart Defense module (at
least some of it's parts) basically can't be turned off. I've seen
numerous problems due to this, even when the only rule was to allow
all traffic to the destination IP address. The problems I've seen were
caused by the Smart Defense module incorrectly interpreting traffic as
invalid and dropping it, while in fact it was legitimate traffic.

That being said, and as one other poster wrote, it looks to me like
you are triggering the SYN flood DoS protection on the firewall, after
which it answers to every SYN packet. Once the connection is
established, it will pass this through to the destination IP address,
otherwise it gets deleted from the stateful connection table.

This indeed can cause some performance problems, but, if I'm not
wrong, you can increase the table size so it shouldn't drop any new
connections (if the table is big enough).

firewall has a rule to accept ANY services for the scanner. The scans are
sometimes successful and sometimes they get garbaged and how does that
make
it a feature?

Now, this looks like a potential problem. Can you reproduce this? Does
it always works ok first time and then in sequential scans you get all
the ports open?
If yes, then the firewall looks to be properly working. If not, you
should check what's going on with the connection table on the
firewall.

Cheers,

Bojan



Current thread: