Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Mon, 6 Aug 2012 18:21:04 +0400
From: Aleksey Cherepanov <aleksey.4erepanov@...il.com>
To: john-users@...ts.openwall.com
Subject: Aleksey's writeup for Crack Me If You Can 2012

I participated in cmiyc 2011 and was a coordinator in phdays hash
runner 2012. This time I was coordinator but it was very smooth: no
problems with formats (except sunmd5 that was rather a nice
challenge), no users to subscribe during contest, no big problems at
all.

Team worked well. I said what to do only once (though maybe I had to
say it more times).

I delegated Alexander Cherepanov to do all server scripting tasks. So during the
contest I was able to do other things, some cracking, track talks and
situation.

I could mention some problems but they are minor. So do not feel too
guilty. ;-)


Preparations

We had training contests a week before the contest. At the same time
we setup ircd. I think these two points helped us.

I subscribed all users before the contest. Though Jim aka Spank was
subscribed right before the contest because his subscription request
was lost but request from other email worked well.

I changed motd and faq files. Though I think I had to promote them
more for those who participate first time. Also I had to check that
all members uploaded at least one .pot because Spank did not upload
.pot files so it seems to be my fault.


Problems and actions

After contest start I faced different small problems.

There were no scripts around 7z and gpg. So the first thing to do was
to write wrapper around 7z. I spent two hours going in wrong direction
but when I realized it I wrote script to crack 7z in half of hour.
Though I did not cracked any 7z. I should admit that I have low basic
cracking skills because I crack only during contests.

I wrote gpg wrapper too but it did not handle false positives. I did
not finished gpg wrapper but we solved enough challenges. While it
searched for passwords I looked for dmg cracking but after all I did
not find any free software tool to open/mount/crack it. Among others I
tried 'mount' on it but it stuck eating 100% cpu, ignoring kill and
preventing soft reboot. Also I did not understand how to crack .asc
files (gnupg* challenges).

Patterns were much smarter than before. Though there were
"traditional" patterns, especially 'password' pattern. I wrote a
script to mangle words just like I want but that script proved itself
to be rather useless because it uses too much memory. Nevertheless I
built 'password' pattern and finished it against fast hashes. Frank
exhausted it for mscash2 on gpu. samu finished it for md5*. For slow
hashes I prepared reduced version but it was not effective enough to
be finished. I split my script into small that do not consume so much
memory. So I got 'washington' pattern mentioned by Frank. It was huge
but not so efficient so it was not done against slow hashes.

I built 'grandson' pattern but it turned out to be a part Swiss cities
pattern. There were some problems with: many lists, many rules that
were not really right (though they were fruitful in any case).  Solar
Designer asked rofl0r to reduce list of cities to only found in
cracks.

So here I should describe workflow found during the contest. It works
because patterns for fast and slow hashes were the same. So we found
something and start wide range attack or just do some extensive search
that gives some results on fast hashes. Then we reduce pattern to be
exact. Here we could realize that there are more than one pattern
(i.e. 'password' + 1 any char in any position, 'password' + leet,
'password' + leet + any char + number). So we get many patterns that
are smaller and we run them against slower hashes (for instance
mscash2 on gpu is slow but enough fast for quite big patterns). Then
we try to reduce pattern more and run against most slow hashes.

Though even reduced 'password' pattern was too big for one man to run.
I asked Alexander Cherepanov to run it on 48 cores but it'd took 1.5
days so we postponed it. I started to think how to distribute attack
but got nothing good. So my "hardcore pattern search" turned out to be
not so hardcore. On the other hand Jim and Simon did distribution
manually through files on server.

I noticed problem with wikipedia list: there were different lists and
nobody knew did we tried any against all hashtypes or not.

After all I'd say we need a table format-attack so we could keep track
of exhausted attacks and patterns.

Both distribution and tracking were goals for my MJohn project so I
regret that I did not make it in time.

I used my old one-liner to ease pattern search. After contest I found
a bug in it: it does not search groups of words with one different
letter - it searches groups of words with the same beginning. It is
not so cool. I played with it a bit: one letter difference is too
strict, two letters are much better.

Trying 'password' pattern I got a problem with file names: in first
case filename did not contain format name (md5a), in other case I had
to use number of file in names of session. I stopped on the second
option.

Elijah found 'pride and predjustice' pattern. While I looked on it
Solar Designer asked me to script phrase ripper. I scripted it in 5
minutes and it worked. I was afraid that straightforward method
produced too many candidates but it did not. Also I was not sure that
[\w,] pcre range to catch words is good so then I tried to include
more signs in words (\S range, all except white spaces) and to make
pack of 1-5 words. 5 words did not yield anything, 1-3 words yielded
something but then I refreshed my uncracked and about nothing were
left - only one word with colon but I guess it would be found in any
case with rules against cracked or basewords.

In the end I run mangled cracked against fast hash types. Mangling
words I considered to store wordlists on disk because I did mangling
in many steps with branches and did not want to wait again redoing
previous steps for new branches. So I spent all disk space in my home
dir partition. It was nasty. I had to think about it before the
contest.

Mangling revealed that kikugalanet+6digits pattern has also variant
with 8 digits. I found it after the contest and I guess it was not
tried during the contest. Though it does not seem to be so effective.

After the contest I tried to find Lord of the Rings on the Internet to
download but failed because it is under copyright. It makes situation
'p&p vs LotR' not so comic.

I think I was able to improve some things making better layout of
files.

Just noticed Solar Designer had problem with submission. I think
KoreLogic could provide a way to do test submissions.

After the contest Alexander Cherepanov pointed out to me that I had
wrong time on my computer (about 45 seconds). Right time setup should
be one of points during preparation.

This contest had a lot of very different tasks that made it very
interesting.


Team work

We had a good time. There were jokes. We supported and helped each
other. Everyone did small things and together we got great result. I
think our relationships were improved and it is important and will
give us more results in the future.

We got many good things during the contest (including fun, experience,
bugs, code for sxc, sunmd5) hence so close victory is not important.
We are on the way to become the best forever.

Thanks!

-- 
Regards,
Aleksey Cherepanov

Powered by blists - more mailing lists

Your e-mail address:

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