Anti-CSRF measures and XSS
During an assessment of an application, I and my colleague Ronnie were discussing about a scenario in the application. The application had login section behind which there were few pages that were vulnerable to Reflected XSS. Application was also vulnerable to CSRF.Needless to say that we suggested anti-CSRF measures for the application. Although we also suggested anti-XSS measures but the anti-CSRF measures were good enough to mitigate any attempt to exploit the reflected XSS flaws on the pages behind authentication. The application was rejecting any external request.
So any attempt to exploit the reflected XSS will bear no fruit in scenario like this.
Anyways we had recommended fixing both flaws independently but I wanted to have a discussion over the issue.
Lots of people responded to that. All were with the same suggestion- do fix both issues, don't take chance.
But what I found most convincing were these arguments from MustLive and Lava:
My suggestion to you and all people in such cases - always fix all holes.
There were similar cases in my practice, mentioned by you, when developers
was trying to argument to not fix XSS due to CSRF protection. Like in 2006
when I found many holes in WordPress 2.0.3 and developers trying to tell me
regarding one XSS that they have anti-CSRF tokens at that page, so no need
to fix that hole, but I said them that all holes must be fixed, and they
fixed. So fixed hole it's fixed exactly this hole (not some other hole).
Nilesh, you must understand, that there are methods which allow to bypass
anti-csrf filters, so if XSS will be left unfixed, then sometime it can be
used for attack.
Best wishes & regards,
Administrator of Websecurity web site
ClickJacking can be used to bypass Anti-CSRF measures in many instances.
Tomorrow we might have a new technique to bypass CSRF countermeasures.
And everytime that happens the application would be open to two attacks CSRF as well as XSS.
Moreover, if the attacker can perform a Session fixation attack and use his session's Anti-CSRF tokens to perform XSS, the user would still be in trouble.
True, I agree with both of them. If any new technique comes tommorow which can bypass anti-CSRF filters (even today Clickjacking is there); then XSS would get exploited.
We had also similar concerns so we recommended developers to deal with both issues separately. But we just wanted to make our arguments more strong and asked people to provide us their valuable comments.
Thanks to all!