|
Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com |
From: Chris Scott (chris
hostorlando.com)Date: Wed Feb 06 2002 - 10:13:21 CST
Interesting. I've seen the same behavior when generating GUIDs from ASP but
from SQL 7.0 I get:
7D62E3A7-E808-4AA5-B1C4-2F71DD64442A
3C2E0FCC-912E-4000-A351-19E60AEE1D00
B627455F-5674-4BA0-8439-64086618FA68
3B4D1DC5-9474-4F1F-9B24-A790C159E29D
Using the following SQL:
SET NOCOUNT ON
DECLARE
COUNTER NUMERIC
SELECT
COUNTER = 1
WHILE (
COUNTER < 5)
BEGIN
SELECT NEWID()
SELECT
COUNTER =
COUNTER +1
END
Chris
> -----Original Message-----
> From: Neil [mailto:webapp1
geekshanty.com]
> Sent: Wednesday, February 06, 2002 10:04 AM
> To: Innes Fisher
> Cc: webappsec
securityfocus.com
> Subject: Re: SQL GUID as a session authentication token
>
>
> GUID's created using NEWID() are *not* a secure identifier. They are
> unique, but extremely predictable and should not be used as an
> authentication method.
>
> SELECT NEWID() (done 5 times)
> B41B2D23-1923-11D6-90EC-00E0291DE14E
> B41B2D24-1923-11D6-90EC-00E0291DE14E
> B41B2D25-1923-11D6-90EC-00E0291DE14E
> B41B2D26-1923-11D6-90EC-00E0291DE14E
> B41B2D27-1923-11D6-90EC-00E0291DE14E
> ^^
>
> As you can see, most of the GUID is the same each time. From what I
> remember, the GUID is based in part on your MAC address and then an
> ever increasing number. This is even worse because the value isn't
> seeded based on time! To get the last connection you simply modify
> the GUID counter part down a number or 2 and you now have a valid
> identifier.
>
> In short, GUID's are about as predictable as they come and should not
> be used as an identifier for a security token.
>
> Hope this helps,
>
> Neil
>
>
> On 02-06 (08:14), Innes Fisher wrote:
>
> >
> >
> > I'm reviewing the security design on a web app which
> > uses a GUID generated by a backend MS SQL
> > database as the session token for the authenticated
> > user.
> >
> > While using GUIDs will theoretically guarantee
> > uniqueness of the tokens, I have serious reservations
> > about it being used in this way. If GUIDs can be
> > predicted in advance it would be possible to appear to
> > be an authenticated user.
> >
> > How predictable are GUIDs and has anyone seen
> > them used in this way before?
> >
> > Is this as crazy as I think?
> >
> > Thanks,
> > Innes
> >
>
>
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]