Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 2 Apr 2008 08:46:33 -0600
From: RB <aoz.syn@...il.com>
To: john-users@...ts.openwall.com
Subject: Re: John + Boinc

> I need something like a wrapper around john. So that i can control
> john via Boinc.

Presumably so you can parallelize the process; interest in this goes
in spurts, so it must be that time again.  Short answer: nothing
particular to Boinc exists that I am aware of.

In greater detail: depending on what you want to do there are several
approaches to this each with varying success, but none that really
address it properly.  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.  Additionally,
the heavy use of assembly makes non-local (networked) parallelism much
harder.  Those two features (ordering & really fast ASM) are, in my
opinion, what set JtR apart from the rest of the crowd of password
crackers and makes it a professional tool, not a kiddie script.

Probably the most successful parallelization attempt is John
Anderson's maintenance of Ryan Lim's MPI patchset
(http://bindshell.net/tools/johntheripper), as it does a decent job of
spreading the workload across multiple nodes.  Unfortunately its
utility is reduced since it only really works for brute-force
(--incremental) mode.  Even so, with enough participating processors
and a fast enough backbone thrown at it, an MPI cluster can overcome
the lack of elegance with pure force and make a reasonably quick pass
at specific problem sets.

Other approaches try to split the keyspace (usually using some
--external module) among cluster participants, but are hugely
inefficient due to uneven key distribution.  You could probably script
something quick-and-dirty with bash for Boinc in this manner, but keep
in mind that it's likely not going to give you the speed or
scalability you're looking for.


I should work up a thesis or an RFC on an algorithm for predicting
when questions will be repeated on a given ML. ;)

-- 
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

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