john-users - filter performances

[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Sun, 16 Oct 2011 23:11:15 +0200
From: Jérôme Loyet <ml@...bsd.com>
To: john-users@...ts.openwall.com
Subject: filter performances

Hi guys,

I'm new to jtr and I'm facing some performance problems (I think we
all do somehow)

I have a single traditional DES password to bruteforce. I know its
policy:  8 characters long (or more) and it uses at least one lower
case, one upper case, one numerical and one "other" char.

If I'm bruteforcing using the mode All (with a fixed 8 chars len) I
have 95^8 = 6634204312890625 possibilities

I want to reduce the number of tries as I know the policy. I have
95^4*26*26*33*10 = 18170005425000 possibilities

I'm using OpenMPI to parralize to 2 servers (16 cores each) and I can
compute around 40387K combinations per sec. Which means the following
ETA for trying all the passwords

6634204312890625 / 40387000 / 60 / 60 / 24 = 1901 days
18170005425000  / 40387000 / 60 / 60 / 24 = 5 days

The first case cannot be considered :)
But the second case is far more easy to considered

To filter the password not corresponding to the policy, I've tried to
apply a filter similar to the optimized one described in
http://www.openwall.com/lists/john-users/2009/10/28/11

But the performances are horrible. Here is a small benchmark (one
single john instance is running - aka one core, one server):

[root@...xxx run]# ./john -i:All8 --external=test --session=test pass.txt
Loaded 1 password hash (Traditional DES [128/128 BS SSE2-16])
guesses: 0  time: 0:00:00:16 0.00%  c/s: 5733  trying: cow_ie2M - cadry S9
guesses: 0  time: 0:00:01:26 0.00%  c/s: 36918  trying: ccosh-L3 - ccohB#40
guesses: 0  time: 0:00:03:22 0.00%  c/s: 59465  trying: 5991Br9$ - 5975Bic*
guesses: 0  time: 0:00:04:45 0.00%  c/s: 65719  trying: drculA$0 - ducht-L7
guesses: 0  time: 0:00:09:07 0.00%  c/s: 69311  trying: pduth1A# - pdul2#A4
guesses: 0  time: 0:00:20:35 0.00%  c/s: 85975  trying: 3dygut N - 3dygkb-F
guesses: 0  time: 0:00:26:28 0.00%  c/s: 86144  trying: 5myaf0$K - 5myon St


[root@...xxx run]# ./john -i:All8 --session=test pass.txt
Loaded 1 password hash (Traditional DES [128/128 BS SSE2-16])
... c/s is constant around 2408k
guesses: 0  time: 0:00:00:53 0.00%  c/s: 2408K  trying: 323akbab - 323agemc


Without filter I can compute 2408K and with the filter enabled the
computation rate is much lower. (And I don't understand why the c/s is
increasing and then stall).

Is there a way to achieve my goal with the same performances ?

thx a lot
++ fat

PS: I'm using john-1.7.8-jumbo-7