On Sat, Oct 19, 2013 at 6:26 AM, Pawel Jakub Dawidek <pjd@freebsd.org> wrote:

On Mon, Oct 07, 2013 at 11:18:21PM +0100, Saso Kiselkov wrote:
> On 10/7/13 10:17 PM, Zooko Wilcox-OHearn wrote:
> > So, before I go on with my pitch for why you should consider BLAKE2,
> > first please clarify for me whether ZFS really needs a
> > collision-resistant hash function, or whether it needs only a MAC. I
> > had thought until now that ZFS doesn't need a collision-resistant hash
> > unless dedup is turned on, and that if dedup is turned on it needs a
> > collision-resistant hash.
>
> The reason is purely for dedup and pretty much nothing else. As such, we
> only need a hash with a good pseudo-random output distribution and
> collision resistance. We don't specifically need it to be super-secure.
> The salted hashing support I added was simply to silence the endless
> stream of wild hypotheticals on security.

Just FYI, Richard Yao just proposed providing VM images with existing
ZFS pool also for production use. This is great idea, but also a nice
proof why making assumptions on how exactly a general purpose software
will be used is bad. In this case your salted hashing is pointless as
everyone knows about the salt in those images. And we are back to square
one.

Saso, there are countless examples where so called hypothetical security
bugs turned out to be exploitable in practise. Which is especially true
for general purpose software that we have no control over how it is
being used.

As Zooko mentioned cryptoanalysis of the Edon-R algorithm stopped at the
point where we know it is not cryptographically secure and this is
unlikely we will see any further work in the subject, which in my eyes
is even worse, as we don't know how bad is it.

To sum up. Even if the OpenZFS community will agree to integrate Edon-R,
I'll strongly oppose having it in FreeBSD. In my opinion it is just
asking for trouble. I still like your change very much and would love to
see new, cryptographically secure hash algorithms for dedup in ZFS.
Currently there is no alternative, which is bad for security too.

--
Pawel Jakub Dawidek http://www.wheelsystems.com
FreeBSD committer http://www.FreeBSD.org
Am I Evil? Yes, I Am! http://mobter.com

So as I have asked previously, can you point to any working exploit of any of the commercially available deduplication stacks? None of them are cryptographically secure. All of them are used in extremely high value targets including top secret government installations. I would assume that almost a decade on there is working code out there if the exploits are anything beyond hypothetical. While I applaud your diligence, it seems beyond silly to me to hamstring people's options because you PERSONALLY wouldn't want to use that as an option. If I as a user understand and am willing to accept the risks associated in exchange for something approaching usable from a performance perspective, you're telling me I shouldn't get that option because you are worried about a theoretical exploit. That's an awfully slippery slope to start heading down. Where does that stop? Are you going to recommend BSD also stop supporting any system that boots off of PCBIOS which has been proven to be exploitable in the field? Are we going to remove md5sum from BSD because that's exploitable?

If you want to recommend they not have it as the default, great. But recommending it not be included at all because you don't feel comfortable with it seems like a GREAT disservice to the community. I thought this horse had already been beaten to death but I guess not.

--Tim