We have gotten off topic, but to complete the thought:
Is it fair to assume by your last comments about working for ISS and
panning Tipping Point that ISS' Proventia appliance would detect the
scenario you outlined below? If so, is it blocked out of the box?
Can someone on the list test this with a Proventia box?
Back on topic (and soapbox)
Is there nothing else that the IPS/IDS user community wants from
future IPS appliances? I see this as an opportunity to really outline
the roadmap that vendors should be taking. It is the lack of good
input on forums like this that results in developers who aren't
security savvy creating products that aren't what the public wants.
Jack
On Tue, 21 Sep 2004 15:31:52 -0400, David Maynor <dmaynor@gmail.com> wrote:
I work for ISS btw.
On Mon, 20 Sep 2004 18:49:25 -0400, David Maynor <dmaynor@gmail.com> wrote:
No guess. It's a simple recipe. Take one Tipping Point box, one
machine that is vulnerable to MS03-026, and a bit on coding knowledge
of rpc. I chose 03-026 because due to Blaster it should at be the one
exploit it could block. You then manually generate the bind request
carefully making sure the RPC frag size is so small that the target
guid is in two different packets. Something funny happens. You would
get system access and the Tipping Point box does nothing. Nada. Zip.
Zilch. While you are issuing commands like 'net user /add hacker' the
Tipping Point box stays silent. I scratched my head over this till I
realized that it happens because TippingPoint doesn't do protocol
parsing, it's the only explanation. Hardly the behavior for an award
winning IPS that writes its sigs for the vulnrebility and not the
exploitâ.supposedly.
On Thu, 16 Sep 2004 23:30:30 -0400, Tony Carter <tcarter@entrusion.com>
wrote:
David,
Can you back your claim that IPS can easily be evaded by fragging
packets? Have you actually tested this or is it your guess?
-Tony
On Sep 12, 2004, at 12:29 AM, David Maynor wrote:
Yeah....I am gonna go ahead and disagree with you on some of these.
I have seen a lot of discussion about the differences between IDS,
IPS, and firewalls and the potential for convergence, but I do not
recall a discussion on the primary features that an IPS should have
out of the box.
I am thinking of:
- Flow Control - limitations on flooding, unused connections, etc...
Most of this should be handled by the signature base.
- Robust, ACURATE signature base
Only way to do this and not create tons of false postives is true
protocol parsing. This knocks out most IPS vendors like Tipping Point.
- Packet capture - no debate on how much before, as that has been
covered
- Pre-deployment network analysis tools to accelerate deployment
- Anomaly detection
Why? I have yet to see a system that is more than a parlor trick.
Anomaly based system are even easier to evade than sig based systems
that don't do protocol parsing.
What I would add is better tools for testing. Almost nobody grabs a
copy of Canvas from Immunity or Impact from Core and actually checks
what attacks are caught. Further more an even fewer number use modded
copies of public exploits to see if the claims made by vendors are
actually true. How many vendor's IPS implementation would actual catch
a MS03-026 exploit if you frag at the RPC layer at a size like 8
bytes?
-----------------------------------------------------------------------
---
Test Your IDS
Is your IDS deployed correctly?
Find out quickly and easily by testing it with real-world attacks from
CORE IMPACT.
Go to
http://www.securityfocus.com/sponsor/CoreSecurity_focus-ids_040708 to
learn more.
-----------------------------------------------------------------------
---
--------------------------------------------------------------------------
Test Your IDS
Is your IDS deployed correctly?
Find out quickly and easily by testing it with real-world attacks from CORE
IMPACT.
Go to http://www.securityfocus.com/sponsor/CoreSecurity_focus-ids_040708 to
learn more.
--------------------------------------------------------------------------
