On 4/15/12 3:12 PM, Alexey Melnikov wrote:
>> Dear WG participants,
>> draft-gondrom-frame-options-02.txt and
>> draft-gondrom-x-frame-options-00.txt (and their earlier version) were
>> discussed in a couple of WG face-to-face meetings. I believe both of the
>> documents are in scope for the WG Charter and I think there is support
>> for working on them.
>>
>> As both of these documents are co-edited by Tobias, I will be judging
>> consensus on them. draft-gondrom-frame-options is targeted to become a
>> Proposed Standard, while draft-gondrom-x-frame-options will be an
>> Informational document (as it documents existing header field). So I
>> would like to start a one week acceptance call for these documents as
>> WebSec WG documents. Please send me your objections or statements of
>> support directly to me or to the mailing list before midnight GMT+1 on
>> April 23rd.
I have one nit with this document - the way it brings Cross Site Request 
Forgery (CSRF) into the discussion and positions X-Frame-Options as a 
"sometimes" mitigation for CSRF.  We could probably agree that wording such as 
"In some forms of Clickjacking and CSRF an attacker tricks a user into 
clicking..." sufficiently limits the scope of the context.  However I feel that 
CSRF in general is just completely out of scope - there are many attacks that 
might leverage framing after all.  I also worry that some people may be led to 
believe that the X-Frame-Options header provides general protection from CSRF 
when that's absolutely not the case.

When Microsoft released this functionality with IE8, it was positioned as a 
protection against framing attacks, or Clickjacking, and not CSRF.  I believe 
that's how many of us in the security community still view it - it's only a 
protection against CSRF in cases where framing is required to execute the CSRF 
attack.  The two (Clickjacking and CSRF) can only be linked in that and similar 
contexts.  

I suggest that references to CSRF protection be removed to avoid confusion or 
that the wording reflect this narrow scope of the CSRF-protection (preferably 
the former).  After all, CSRF is just one example of many other attacks that 
could leverage framing - e.g. we could include answering CAPTCHAs, certain 
forms of self-side XSS, and even information disclosure as equivalent forms of 
attack (like CSRF) that might leverage framing.

Can we keep the document focused on the primary design goal - 
controlling/preventing framing - and avoid lumping in other forms of attack 
that might piggyback on such framing?

Best regards,
Chris Weber

_______________________________________________
websec mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/websec

Reply via email to