[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <C21807915A8D20A60347F757@paul-schmehls-powerbook59.local>
Date: Tue, 06 Jun 2006 20:05:18 -0500
From: pauls@...allas.edu
To: bugtraq@...urityfocus.com
Subject: Re: Squirrelmail local file inclusion
--On June 6, 2006 4:32:02 PM -0400 "Steven M. Christey" <coley@...re.org>
wrote:
>>
>> Yet know we're getting "security advisories" warning, hey, if you
>> change the defaults and ignore all the warnings, you too can write
>> insecure code!
>
> In this sense, I agree. Default configuration is one thing, but
> active negligence is another.
>
> That said, Squirrelmail apparently thinks this issue is important
> enough to release an advisory:
>
> http://www.squirrelmail.org/security/issue/2006-06-01
>
> So maybe they know more about the implications on their consumers than
> we do.
>
Did you look at the patch?
It's almost 40 lines of code designed to *force* register_globals to "off"
no matter what the idiot behind the wheel does.
I understand that there are many shared hosting sites that have
register_globals set to "on". IMNSHO opinion they are putting their
customers at risk unnecessarily. The default setting should be "off", and
if someone needs it turned on in the application they're using, they can
turn it on there and assume the risk themselves.
It wouldn't take much for web hosting companies to educate their users. A
few lines of advice on a webpage would suffice. For example, for Apache
users, you can set the php_flag register_globals on/off in a VirtualHost
context (inside <Directory>.) So the hosting company can enable it on
request but use the proper default setting for most sites.
Or the individual webmasters can set php_register_globals = [on|off] in a
.htaccess file (Apache 1.3.x only!) or php_value register_globals [0|1]
(Apache 2.x only!) if they don't like the default setting the hosting
company uses. Or they can simply include a file with the setting in it and
add an include statement to the code of the app.
I'm sure there are other methods that could be used as well. Since there
are many options available, the default should be secure.
Paul Schmehl (pauls@...allas.edu)
Adjunct Information Security Officer
The University of Texas at Dallas
http://www.utdallas.edu/ir/security/
Content of type "application/pkcs7-signature" skipped
Powered by blists - more mailing lists