Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 9 Apr 2008 00:41:07 +0400
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: John + Boinc

RB and Simon have already provided some good responses (thanks!) but
I'll add a few comments as well:

On Wed, Apr 02, 2008 at 09:27:44AM +0200, Markus Friedel wrote:
> I need something like a wrapper around john. So that i can control john
> via Boinc.

I am not aware of any existing efforts specific to JtR & BOINC, but
there have been many other efforts to introduce distributed processing
into JtR.  I have a collection of 7 implementations here:

	ftp://ftp.openwall.com/pub/projects/john/contrib/parallel/

On Wed, Apr 02, 2008 at 08:46:33AM -0600, RB wrote:
> The crux of the matter is that parallelizing a
> workload like JtR intelligently is not a small problem, mostly due to
> the intelligent candidate password ordering it does.

Yet some of the implementations actually got it right.

> Additionally, the heavy use of assembly makes non-local (networked)
> parallelism much harder.

I don't see a problem here.

> Probably the most successful parallelization attempt is John
> Anderson's maintenance of Ryan Lim's MPI patchset

I agree.  Most importantly, this one is actually maintained - it is not
one of those academic projects that get abandoned in a month.

On Thu, Apr 03, 2008 at 06:15:15PM +0200, Markus Friedel wrote:
> i have about 80 AMD Dual-Core Opteron 180 / 2.4 GHz with each 4 gig RAM

Your RAM won't be needed.

> also i have 400 Passwords form LDAP.

You haven't mentioned the hash type.  Is it salted or not?  How many
different salts?

> My first thought was to deploy one passwd to each PC and let work them
> for some hours.

That would be inefficient because there's a lot of common work in
processing a candidate password against different hashes - especially
against saltless ones.  For saltless hashes, you're better off keeping
them all on just one machine rather than spreading them over all of your
machines like that.

> The seconde thought was to deploy the same passwd to all PCs and split
> the range of the wordlist they have to use.

This is a lot better, however wordlist-based cracking is usually fast
enough to be performed on just one machine.

What you may do is use "incremental" mode and distribute the work across
a few of your machines manually, with different MinLen/MaxLen settings.
For example, for 4 CPU cores (2 of your machines), you could use lengths
0 through 5 on one CPU core, then 6, 7, and 8 on the remaining three,
respectively.  Yes, 78 out of your 80 machines will just stay idle.  To
make use of them all, you really do need one of those distributed
processing hacks of JtR.

> So i don't know if it is the right approach to use my ressources to get
> some of the passwords cracked.
> The purpose of this project is to find weak passwords and shows it to me
> which are the ones.

I suggest that you start by running a single instance of JtR on just one
machine.  Chances are that you will get a lot of passwords cracked.
Then you'll decide what to do next.

Good luck!

Alexander

P.S. Markus, you posted your first message on this topic as a "reply" to
an unrelated message.  This results in incorrect threading in some
web-based archives of this mailing list.  Can you and others please be
more careful about this?  Whenever you post something on an entirely new
topic, post the message anew, not as a "reply".  Simply changing the
Subject may not be enough to break the thread.

-- 
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