Cryptohaze.com

Hi,

Could you implement the support of LM hashes ?
A LM hash can be divided in 2 parts, and is only uppercase.. the charset is a bit less than other algorithm.
Do you plan to do it ?
Thanks.

I know of LM

I'm actually looking at grabbing the FRT LM implementation to plug in - there may be faster ways of doing it, but at least for now, doing that would give me LM support. It's on my list, just working on a lot of stuff at once lately

Bang, Zoom, Right in the kisser!

Code: Select all

+------------------------------------------------------------------------------+
|'p' to pause              Cryptohaze Multiforcer 1.00             'q' to quit |
|                               Mode: Standalone                               |
|Hash type     :        LM |     Passwords Found       |                       |
|Current PW len:         4 |                           |  0: GPU: 3.11M/s      |
|Total hashes  :         1 |  123                      |                       |
|Cracked hashes:         5 |  BBB                      |                       |
|Total time    : 00:00:08  |  AAA                      |                       |
|WUs: 0/1 (0.0%)           |  CCC                      |                       |
|                          |  ABC                      |                       |
|--------------------------|  ┌──────────────────┐     |                       |
| Thread 0 mem loaded      |  │      PAUSED      │     |                       |
| Creating 1 threads       |  │  Press any key   │     |                       |
| Starting pw len 4        |  └──────────────────┘     |                       |
| Threads joined           |                           |                       |
| Td 0: out of WU.         |                           |                       |
| Thread 0 mem loaded      |                           |                       |
| Creating 1 threads       |                           |                       |
| Starting pw len 3        |                           |                       |
| Threads joined           |                           |                       |
| Td 0: out of WU.         |                           |                       |
| Thread 0 mem loaded                                    TOTAL: 3.11M/s        |
|                                                                              |
+------------------------------------------------------------------------------+


Ignore the glacial rate... this is on an Ion2 Netbook. I get like 38M NTLM/sec on this.

There are also very definitely some optimizations I can make. This is a pure "First pass, get it working" implementation.

... like, "I'm going out drinking with friends now, so I have the big sbox lookup tables in constant memory instead of shared." That sorta silly stuff.

I really need a remote build system... I'll try to push out a beta build soon.

I also need to write a file class that splits the hash into two bits and merges them together - I'm not doing this right now.

But, later this week, it should be ready to rock, and I'll try to push up a new copy of the multiforcer to play with.

Oh, and maybe some LM support love in the rainbow tables...

Code: Select all

+------------------------------------------------------------------------------+
|'p' to pause              Cryptohaze Multiforcer 1.00             'q' to quit |
|                               Mode: Standalone                               |
|Hash type     :        LM |     Passwords Found       |                       |
|Current PW len:         5 |                           |  0: GPU: 4.46M/s      |
|Total hashes  :         1 |  BBBB                     |                       |
|Cracked hashes:        11 |  AAAA                     |                       |
|Total time    : 00:00:13  |  DDDD                     |                       |
|WUs: 0/1 (0.0%)           |  ABCD                     |                       |
|                          |  CCCC                     |                       |
|--------------------------|  ┌──────────────────┐     |                       |
| Thread 0 mem loaded      |  │      PAUSED      │     |                       |
| Creating 1 threads       |  │  Press any key   │     |                       |
| Starting pw len 5        |  └──────────────────┘     |                       |
| Threads joined           |  CCC                      |                       |
| Td 0: out of WU.         |  ABC                      |                       |
| Thread 0 mem loaded      |                           |                       |
| Creating 1 threads       |                           |                       |
| Starting pw len 4        |                           |                       |
| Threads joined           |                           |                       |
| Td 0: out of WU.         |                           |                       |
| Thread 0 mem loaded                                    TOTAL: 4.46M/s        |
|                                                                              |
+------------------------------------------------------------------------------+


THERE we go. That's better numbers.

Great!
Coupled with network distribution, this is big!

I estimate ~400M +-3% for two GTX 480.
Wicked if it comes true!
Cant wait to find out.

Rolf wrote:Coupled with network distribution, this is big!

Network distribution definitely works. I tested that yesterday.

Rolf wrote:I estimate ~400M +-3% for two GTX 480.
Wicked if it comes true!
Cant wait to find out.

That sounds about right for a Fermi. Sadly, all of mine are offline right now (power glitch, haven't been home to reboot it yet), or I'd have better numbers.

I'm finishing up the LM hash file class today (automatically handles splitting things up into segments, probably will *not* do automatic case correction for now but I can add that in later in a full SAM dump class).

Hopefully I can actually get a release out this weekend with all the various improvements. I'm hoping to have a more robust workunit class finished by then as well, so network support will actually be solid and not lose a bit of work if clients disconnect.

Output format proposal:

Code: Select all

9BB567983FE10AE9AAD3B435B51404EE:P4S$
070CD9CE1D0A27FCBD3873D1639C4383:xxxxxxxAZE
B757BF5C0D87772FAAD3B435B51404EE:1234
B64FC1365AA65D45AAD3B435B51404EE:BAD
91BBCA250E3F338BAAD3B435B51404EE:BWAH
8C6F5D02DEB21501AAD3B435B51404EE:ABC
8F704E846BF64093AAD3B435B51404EE:GOOD!
E52CAC67419A9A224A3B108F3FA6CB6D:xxxxxxxD
E52CAC67419A9A2222C34254E51BFF62:xxxxxxxD!


Right now, I'm only taking LM hashes - not NTLM. I'll add that support later (LM + NTLM, case correction, etc).

For now, my plan is to display the found bits, and if I've found a non-null second bit and not found the first bit, I'll just put lowercase 'x' characters.

Sane?

Yep!

Rolf wrote:I estimate ~400M +-3% for two GTX 480.

Bad news.

Your estimate is off by a good bit.

//EDIT: Newer numbers with different thread/block counts (768/360/-m 500)

I've got 328M on a GTX480, and 221M on a stock clocked 470.

*does maths*

A 480, based on performance ratios (which are quite accurate for this stuff), is 1.285x as fast as a 470.

That puts a 480 right around 283M/s (or ~565M/s for two).

In other words, my box can do *ALL* US-charset LM in under 3 hours.

I could kiss you right now!