[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] Security discussion: Summary of proposals and criteria (was Re: Security vulnerability process, and CVE-2012-0217)



On 07/12/12 19:00, Stefano Stabellini wrote:
> On Thu, 12 Jul 2012, Joanna Rutkowska wrote:
>> > On 07/12/12 18:34, Stefano Stabellini wrote:
>>> > > On Mon, 9 Jul 2012, Joanna Rutkowska wrote:
>>>>> > >> > On 07/09/12 15:51, Tim Deegan wrote:
>>>>>>> > >>> > > At 13:31 +0200 on 09 Jul (1341840671), Joanna Rutkowska wrote:
>>>>>>>>>>> > >>>>> > >> > If you're into security industry (going to 
>>>>>>>>>>> > >>>>> > >> > conferences, etc) you
>>>>>>>>>>> > >>>>> > >> > certainly know the right people who would be delight 
>>>>>>>>>>> > >>>>> > >> > to buy exploits
>>>>>>>>>>> > >>>>> > >> > from you, believe me ;) Probably most Xen developers 
>>>>>>>>>>> > >>>>> > >> > don't fit into this
>>>>>>>>>>> > >>>>> > >> > crowd, true, but then again, do you think it would 
>>>>>>>>>>> > >>>>> > >> > be so hard for an
>>>>>>>>>>> > >>>>> > >> > interested organization to approach one of the Xen 
>>>>>>>>>>> > >>>>> > >> > developers on the
>>>>>>>>>>> > >>>>> > >> > pre-disclousure list? How many would resist if they 
>>>>>>>>>>> > >>>>> > >> > had a chance to cash
>>>>>>>>>>> > >>>>> > >> > in some 7-figure number for this (I read in the 
>>>>>>>>>>> > >>>>> > >> > press that hot
>>>>>>>>>>> > >>>>> > >> > bugs/exploits sell for this amount actually)?
>>>>>>> > >>> > > I think the argument is that an exploit that's going to be 
>>>>>>> > >>> > > public (and
>>>>>>> > >>> > > patched) in the next couple of weeks would not fetch the same 
>>>>>>> > >>> > > kind of
>>>>>>> > >>> > > price as a unknown attack that can be kept for later.
>>>>> > >> > 
>>>>> > >> > Depending on the type of an exploit. For client-side exploits, 
>>>>> > >> > perhaps
>>>>> > >> > you're right. But for infrastructure attacks it's a different 
>>>>> > >> > story --
>>>>> > >> > having an exploit such the Rafal's one, I could have *silently* 
>>>>> > >> > exploit
>>>>> > >> > lots of AWS machines and install backdoors in their 
>>>>> > >> > hypervisors/dom0.
>>>>> > >> > The fact that they will patch the bug two weeks later might be
>>>>> > >> > irrelevant then.
>>>>> > >> > 
>>>>> > >> > After all, how are you going to check whether your physical server 
>>>>> > >> > has
>>>>> > >> > been compromised? Most people don't use any form of trusted boot, 
>>>>> > >> > but
>>>>> > >> > even if they did, it's not a silver bullet as we have demonstrated 
>>>>> > >> > a few
>>>>> > >> > times in a row. And if you don't have trusted boot, as most 
>>>>> > >> > people, you
>>>>> > >> > have very little chances to detect a custom-made backdoor. Even if 
>>>>> > >> > you
>>>>> > >> > are allowed to reboot the machine and boot "good known binaries", 
>>>>> > >> > which
>>>>> > >> > often you cannot do, are you going to manually audit all the 
>>>>> > >> > firmware,
>>>>> > >> > ACPI tables, etc? Not to mention about the integrity of the actual 
>>>>> > >> > VMs,
>>>>> > >> > that might have also got compromised (and checking for integrity 
>>>>> > >> > of a
>>>>> > >> > running OS, such as Linux or Windows, is just undoable).
>>>>> > >> > 
>>>>> > >> > Having that said, 2 weeks might be a bit short to prepare such an
>>>>> > >> > advanced attack. In this respect, it would be probably beneficial 
>>>>> > >> > to
>>>>> > >> > keep the embargo period as short as possible (that still allows
>>>>> > >> > important players to patch before others). 1 week perhaps?
>>> > > I agree on the short embargo period and I am not against having a list.
>>> > > 
>>> > > However I don't think that the list should be limited to the "important
>>> > > players". How do we define an "important player"?
>>> > > If we decide that important players are the big ones, suddenly big
>>> > > players become the only ones that can be entrusted with sentive
>>> > > informations.
>>> > > 
>>> > > Any distros can join linux-distros, no matter the size.
>> > 
>> > I didn't say the list should be limited to the important players -- I
>> > said it likely makes no harm, from the security standpoint, if we
>> > allowed select "important" _service providers_ to patch before others
>> > (because, again, the world doesn't see what Xen binaries are running on
>> > their servers, and so this doesn't leak out info about the bug, at least
>> > it shouldn't in most cases).
>> > 
>> > I think that the list should essentially contain only some core Xen
>> > devels and all the software vendors that _are known_ to build and
>> > distribute products on top of Xen (such as qubes-os.org ;) Perhaps the
>> > "important" service providers could be brought in on a case-by-case basis.
> I understand now. This is very similar to having two list, one for
> software vendors and another one for service providers.
> 
> While all the security issues would be discussed in the software vendors
> list, only the most critical ones would be sent to the service providers
> list?
> 
Now, when you mentioned it, I thought that perhaps it would be a better
idea to have one list only for core Xen developers, and the other list
for Xen users, i.e. software vendors such as qubes-os.org, an perhaps
also for some providers.

Perhaps the users from the 2nd list could be informed only when the
specific vulnerability affects their product. As an example, say there
is a bug in qemu reported to the primary list. Now, a query can be send
to all the members of the 2nd list whether any of them are interested in
getting info about the bug, something like:

"There's been a bug found in the qemu used by Xen X.X.X, if you believe
this might affect your product/infrastructure security, reply to this
message with YES, and you will get more info".

The catch is, of course, that if it could be later proven that some
vendors answered YES, although their products were not affected (e.g. by
design), those vendors should be somehow banned from the list. E.g. if I
replied YES to the above query, you should know I went on the dark side
of the force, as Qubes OS keeps qemu outside of the TCB, and so we
shouldn't be caring about such things :)

This is somehow problematic with Xen providers, as opposed to software
vendors, as it's might not be clear how Xen has been deployed by a
particular provider, and thus whether e.g. a bug in pygrub affects them
or not...

> But the key question remain: would we allow a small service provider to
> join the service providers list? I think that we should.

It would probably be very easy for the proverbial Chinese to setup a
small service provider and join the list just to get prompt info about
bugs they can exploit in order to 0wn the proverbial AWS...

joanna.

Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.