Re: squirrelmail CSRF vulnerability

[Josh Zlatin-Amishav <josh () ramat cc>]

On Fri, 11 May 2007, Tim Newsham wrote:

> > 1.4.8-4 is vulnerable to a XSS vulnerability, so an attacker could use the
> > XSS vector to grab the session token ("CSRF token") and continue the CSRF 
> > attack.
>
> This might just be semantics:  I wouldn't consider the XSS attack to be a 
> CSRF attack.

The point is, if the application is vulnerable to an XSS vulnerability
then having a CSRF token wont protect you from a CSRF attack. The
attacker could use the XSS vector to steal the CSRF token, much like the
Samy worm worked.

> The XSS script runs in the same context that the user or any 
> legitimate script running on behalf of the user runs.  When it makes a 
> reference, it has access to things like the CSRF token.

Exactly, thus the CSRF token wont be much help in protection you from a
CSRF attack, if the attacker can just parse out that token and use it in
CSRF attack.

--
 - Josh