Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Mon, 29 Nov 2010 10:06:50 -0500 (EST)
From: "Steven M. Christey" <>
To: Jon Oberheide <>
        "Steven M. Christey" <>
Subject: Re: Linux kernel address leaks

On Tue, 23 Nov 2010, Jon Oberheide wrote:

> Great. There's plenty of precedent for CVE assignment to vulnerabilities
> that leak information that can assist an attacker in exploitation.
> In particular, I'm thinking about the handful of ASLR information leaks
> (eg. CVE-2009-2691 [1]), where userspace addresses are leaked via /proc,
> assisting an attacker in exploiting a vulnerability in a setuid
> userspace binary.
> In this case, we have an analogous situation, except that we're leaking
> kernel address via /proc, assisting an attacker in exploiting
> vulnerability in the kernel.
> I don't see any reason why it would not be entirely appropriate to
> assign a CVE here.

I agree that it would be appropriate.

There is information that leaks from one entity to another entity that 
should not have access to it, in the "intended" security model of the 
Linux kernel (documented security model may be a different story... 
anybody have any pointers?)  No matter how small the leak is.

As Dan pointed out, our original definition of "exposure" (the "E" in 
"CVE") is that it "allows access to information or capabilities that can 
be used by a hacker as a stepping-stone into a system or network." 
(  Kernel address leaks are 
useful for attacking a protection mechanism.

The ultimate CVSS score, while likely low, is not relevant - it's still 

The difficulty of creating a solution is not relevant, either.

There are some practical considerations from the larger CVE perspective, 
e.g. if there are hundreds or thousands of kernel address leaks (which 
wouldn't surprise me given what seems to be happening with struct 
initialization) then there is a risk of having the CVE namespace being 
overwhelmed by the sheer volume of these issues and making CVE IDs more 
difficult to manage.  But that's not necessarily a reason to stop 
assignment (the large number of recent library-loading issues for 
Windows/Linux come to mind), and reporting these in large groups instead 
of bug-by-bug would help keep the CVEs manageable.

If there is some formal or semi-formal position taken by, say, the Linux 
kernel devs or the hardening team that declares "userspace is allowed to 
know about a kernel address," then this becomes part of the 
documented/intended security model, and CVE assignment wouldn't be 
necessary.  (Just trying to be complete here...)

- Steve

> Regards,
> Jon Oberheide
> [1]
> -- 
> Jon Oberheide <>
> GnuPG Key: 1024D/F47C17FE
> Fingerprint: B716 DA66 8173 6EDD 28F6  F184 5842 1C89 F47C 17FE

Powered by blists - more mailing lists

Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.