Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sat, 01 Dec 2007 12:56:05 +0100
From: Simon Marechal <simon@...quise.net>
To: john-users@...ts.openwall.com
Subject: Re: AES Bitslice and the PS3 MD5 cracking.

Solar Designer wrote:
> Well, a full MD5 compression function computation involves 64 rounds,
> with each round consisting of 7-8 operations.  Even if only half the
> rounds need to be computed (which might be possible when trying to match
> one specific hash value and trying candidate passwords in a hashing
> algorithm-dependent order) and each round is reduced to half the number
> of instructions (due to availability of suitable multi-op instructions -
> I have no idea if this is the case for the Cell - probably it is not),
> this still gives us over 100 instructions per "iteration".  Since the
> SPUs are single-issue, this is also the lower boundary for the number of
> clock cycles.

I did the same calculations and have the same conclusion. I wrote a POC
NT cracker that achieves 180M NT/s without any optimization (naive
implementation), so I guess that it should be possible to go faster than
200M MD5/s with a bit of work.

Using the PS3 devkit (7500$) would let a developper use the 7th core and
the GPU. It might be possible to use the GPU as elcomsoft claimed and
perhaps double that performance.

But the main conclusion of this presentation is totally true : you can
get your employer to buy you a PS3 :)

-- 
To unsubscribe, e-mail john-users-unsubscribe@...ts.openwall.com and reply
to the automated confirmation request that will be sent to you.

Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ