Re: Re: CVE Request - Zope / Plone: Multiple vectors corrected within 20121106 fix

[Matthew Wilkes <matthew.wilkes () plone org>]

Hi all,

> > It looks like some of these can be CVE merged, e.g. 14 and 15, 1 and
> > 5, can you confirm that these should not be merged?
>
> Thanks for constructing this comprehensive table, but please do not
> merge 14 and 15, or 1 and 5.

I agree. It's our opinion that these are distinct flaws, and we would 
prefer to keep them tracked independently.

> CVE assignment by MITRE most often has merges when the available
> information suggests one of these two situations:
>
>    A. Flaw types that have been used for many years and are thought to
>       be well understood. At present, a large fraction of our merges
>       are for XSS, SQL injection, CSRF, buffer overflows, integer
>       overflows, use-after-free issues, and directory traversal.
>       However, a merge can occur correctly for any flaw type.

Indeed, we do like to keep similar flaws with very different causes 
separate, though. In the particular case of 1/5 and 14/15 we don't see 
any similarities, but in general we've tried hard here to request CVEs 
in a way that accurately reflects discrete vulnerabilities in the stack.

> At this point in the history of CWE, a discloser's choice of the same
> CWE identifier for two different bugs might not be a strong indication
> that a CVE merge should occur.

The CWE dictionary is huge, I would very much appreciate any feedback 
anyone can give me on the appropriateness of my choices here. I have 
been working on a Plone specific CWE dictionary but it's slow going.

> For example, some CVE consumers don't like situations in
> which a vendor publishes multiple disclosure documents that explain
> different aspects of the same CVE.

We will bow to your advice here. If you tell us that our merge 
recommendations are poor we will make sure that we don't issue multiple 
guidance documents in future. For now, our list reflects our best 
understanding of the CVE guidance documents.

> 14 and 15: One might argue that these are different because 14 is
> about algorithmic complexity but 15 isn't.

Indeed, I did consider CWE-407 here, but 749 is one of our go-to choices 
as it's a common error in Zope. I think, if I had to come down one side 
or another, then 15 is complexity whereas 14 is allowing users to 
circumvent caching on an expensive function. They are both similar 
outcomes, in that they're expensive pages, but for very different reasons.

> 01 and 05: One might argue that these are different because 05 is
> about incomplete security declarations but 01 isn't.

Indeed. Also, 05 is a AC:H Au:S whereas 01 is AC:L Au:N. 05 allows an 
escalation of privileges in the sandbox whereas 01 allows unauthorised 
authoring of code which happens to only ever be run in a higher 
privilege set. 01 is a write-once persistent single statement whereas 05 
is an editable file with multiple statements available. The privilege 
set that 01 escalates to is the un-escalated set for 05.

Matt