[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <12d3365b36d1d0efde89a2987306d397@smtp.hushmail.com>
Date: Tue, 16 Apr 2013 23:25:34 +0200
From: magnum <john.magnum@...hmail.com>
To: john-users@...ts.openwall.com
Subject: Re: Fuzzing with regular expressions
On 16 Apr, 2013, at 22:17 , Jan Starke <jan.starke@...ofbed.org> wrote:
> I just changed some things and was able to speed up rexgen by the
> factor of 5 (on my system) without using threads; additionally the
> ordering of the values is partly random. Maybe you want to give it a
> try...
I am delighted to report that under OSX (built with gcc/g++) r44 is 11.5 times faster than the last version I tried (which was r24 or so). Previous speed about 2.3MB/s (405K words/s) and now over 27 MB/s (4.6M words/s), using '[a-z]{0,5}'. This is still a bottleneck for very fast formats but, well, any way of producing candidates is and with the finer granularity of a regexp you might gain total time anyway.
> BTW, we've been able to crack a bunch of passwords during a pentest
> with rexgen and JtR, because we had an idea about how the passwords
> could look like and we could describe this using a simple regex :-)
Yes, for some patterns (with variable length parts like "abc[0-9]{1,3}def" there's just no way to do it (that easily) with any other tool I know of. Not to mention wilder regexps and back references!
Like I just wrote in another post I'd love to have this as a native mode in JtR but we can't use C++. OTOH, maybe we can add a HAVE_REXGEN in Makefile, stating that we have librexgen installed, and write a mode in C that just calls the lib.
BTW did you see my suggestion of supporting append/prepend to words read from stdin? That would be awesome.
magnum
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.
