Message-ID: <023e01c2fdfc$c95741c0$8701a8c0@satan>
Date: Tue, 8 Apr 2003 11:29:26 -0700
From: "Geoff Shively" <gshively@...x.com>
To: "Nicolas Gregoire" <ngregoire@...probe.com>
Subject: Re: False-negatives in several Vulnerability Assessment tools


Nicolas,
Though I do agree with your analysis of this slew of assessment
tools I must interject.

Any administrator or security tech that relies solely upon a VA tool
to secure a hubed LAN or a corporate enterprise network is asking
for trouble right there. I even believe that a few of the vendors you
refer to recommend that their tool not be used as an all-in-one security
solution.

Running a VA tool or automated pen testing app should be preformed
by a seasoned security professional who will not take the output as the
definitive guideline to what needs securing. Other tools need be included
to diversify and solidify any good security tech's tool-belt.

With that said be mindful of the false positives though they shouldn't be
a big problem if the security team is more than a single VA app and a
basic sys-admin.

Cheers,
Geoff Shively, CHO
PivX LABS

http://www.pivx.com



----- Original Message -----
From: "Nicolas Gregoire" <ngregoire@...probe.com>
To: <bugtraq@...urityfocus.com>
Sent: Monday, April 07, 2003 3:06 AM
Subject: False-negatives in several Vulnerability Assessment tools


------------------------------------------------------------------------
Title : False-negatives in several Vulnerability Assessment tools
Released : April 7th 2003
Location : http://www.exaprobe.com/labs/advisories/esa-2003-0407.html
------------------------------------------------------------------------

General overview
================

Numerous Vulnerability Assessment (VA) tools are available for security
engineers, pen-testers and network administrators.  Their results are
mostly trusted by users since they don't have time nor competences to
validate that output.

More and more softwares are currently implementing some banners and
error messages that depend on the language. Especially for commercial
softwares, like Microsoft SQL Server or the Windows operating system.

Some VA tools don't integrate this localization feature and so generate
false-negatives. It can thus lead to a false sense of security. Some
exploit work on the English as well as on some non-English versions, it
then constitutes a security breach.

We chose to demonstrate those security exposures on Microsoft SQL Server
with the "SQL Server blank password" vulnerability.

Please note that this is not the only issue :

- Some problems were found when VA tools began to detect the IIS/Unicode
vulnerability, like the unicoder.pl script of HD Moore, which is looking
for the localizable string "Directory of" [1].

- The admin account on Windows operating systems depends on the
localization. On English-speaking versions, the name is "Administrator",
whereas on French version (for example), it is "Administrateur". This
leads to issues on brute-force attacks.

A pratical example
==================

Introduction
============

Microsoft SQL Server is a perfect choice to test VA tools about
localization issues because it is widely deployed, it depends on
the localization and it is vulnerable to some well-known
security flaws.

Testing conditions
==================

First, we set up default installations of Microsoft SQL Server
2000 on Win2K SP3, in the following languages :
- English
- French
- German
- Japan
The "sa" admin account was set with a blank password.

We tested every VA tools from our panel on the English version
looking for the vulnerability CAN-2000-1209 ("MS-SQL blank
password").  Products which found this breach were then tested
on the other languages.

Tested VA tools
===============

- ISS Database Scanner
- Vigilante SecureScan NX
- eEye Retina Network Scanner
- eEye Spida Scanner (dedicated to find blank "sa" accounts)
- Nessus
- Sensepost senseql

Untested (or untestable) VA tools
=================================

- ISS Security Scanner (doesn't do this check)
- Symantec NetRecon (doesn't do this check)
- NetIQ (doesn't do this check)
- GFI LANGuard (unreliable results)

Results
=======

        +----------------------+-----------------+------------------+
        |       VA Tool        | English version | Others languages |
        +----------------------+-----------------+------------------+
        | ISS Database Scanner |       OK        |        OK        |
        +----------------------+-----------------+------------------+
        | Vigilante Secure NX  |       OK        |  False-negative  |
        +----------------------+-----------------+------------------+
        | eEye Retina Scanner  |       OK        |  False-negative  |
        +----------------------+-----------------+------------------+
        | eEye Spida Scanner   |       OK        |  False-negative  |
        +----------------------+-----------------+------------------+
        | Nessus               |       OK        |  False-negative  |
        +----------------------+-----------------+------------------+
        | Sensepost senseql    |       OK        |  False-negative  |
        +----------------------+-----------------+------------------+

Notes about the above results
=============================

- The eEye Retina Scanner was tested on this point some time
ago. Amazingly, it used to detect this vulnerability on
non-English versions of Microsoft SQL Server.

- Informal discussions with nCircle developpers conclude that
their VA tool shouldn't be affected by this problem.

- The exploit code nammed SQLpoke [2] (used in the
Worm.SQLSpida.A malware [3]) succeeds to compromize every
localized Microsoft SQL server. This implementation operates at
the application level.

Editors status
==============

- Vigilante Secure NX :
Work in progress on the editor side ...
- eEye Retina Scanner :
Work in progress on the editor side ...
- Nessus :
We provided the Nessus team with some patches which were
integrated to the related plugins
- Sensepost senseql :
A new release is available at [4]

Conclusion
==========

In our opinion, it's now up to VA tools editors to take into account the
localization issues when developping pattern matching signatures. Of
course, security engineers and consultants should review every scan
reports for false-positives. They should also run several tools in order
to better detect false-negatives. A good way to avoid these problems
would be to check vulnerabilies at an application level, like the
SQLpoke exploit code.

Credits
=======

Nicolas Gregoire, security engineer
        - initial discovery of the MS-SQL localization bug
        - testing and redaction

Philippe Conchonnet, security consultant
        - testing of Windows-based VA tools

Christophe Briguet, technical manager
        - review of the document

References
==========

[1] : http://packetstormsecurity.org/NT/scanners/Sqlpoke.zip
[2] : http://lists.insecure.org/lists/pen-test/2001/Jun/0128.html
[3] : http://www.avp.ch/avpve/worms/sqlspida.stm
[4] : http://www.exaprobe.com/labs/downloads/tools/senseql-1.1.tgz

--
Nicolas Gregoire ----- Consultant en Sécurité des Systèmes d'Information
ngregoire@...probe.com ------[ ExaProbe ]------ http://www.exaprobe.com/
PGP KeyID:CA61B44F  FingerPrint:1CC647FF1A55664BA2D2AFDACA6A21DACA61B44F

Powered by blists - more mailing lists