Cryptohaze.com

Let me first start by saying that this piece of software is awesome!!!!!
Thanx Bitweasil

Now for my question:
I've already implemented a special hash algo, sha1($salt.$pass) where salt is 20 chars, in the 'old' cryptohaze source. Works great.
Now i'm trying to implement the same in the new multiforcer. There's an example ( the other way around ) in the MD5 pass/salt cuda source.
I don't fully understand the new cuda kernel code

I've read here in the forum https://www.cryptohaze.com/forum/viewtopic.php?f=6&t=1181&hilit=salt&start=10#p2861 that you are working on that.
Can you lead me in the right direction? Any hints?

Thanx in advance
/Renfield

The new kernel is using a radically different set of plaintext incrementors and I haven't quite worked out how to handle the $salt.$pass stuff yet, though I believe it will involve one of two methods.

The first would be to keep the plains in shared mem, and copy them in, modifying the incrementor code to work off the shared mem location.

The second would be to modify the incrementor code to work with the hash in the offset position, though this is a bit difficult with the possibility of varying salt lengths.

Would you be willing to send me the code for the old framework so I can add it to SVN?

Another possibility would be to shift the password N bytes over, insert the salt, run the hash, and shift the password back to increment it.

I can think of a few quick hacks that would work, though performance may be slightly less than ideal (though I'm not 100% sure what ideal is yet - shared mem is probably the way to go).

Please shoot me an email - if you're already modifying the code, I'd love to chat and talk about other features!

Thanx for your quick reply.

I'm away on holiday now, the source is @ home...

I've just worked with the Multiforcer code for less than a week now, and I must say it's well written and easy to follow/understand.

New algo:
It was really easy (you already did some algo's I could snatch code from).
I just copy/paste the most from

Code: Select all

CHHashTypeSaltedSSHA.cu

and changed a bit....
Fixed the CHashType* files, Derived from CHHashFileSalted32, added to CHashes, etc.

But if you want I can send you the code when I'm back next week.

Another possibility would be to shift the password N bytes over, insert the salt, run the hash, and shift the password back to increment it.

I think that is what I'm going to test. Don't understand the incrementor/shared mem code yet.

I have also setup a new VS2010 project for the 'new' multiforcer (the ones that CMake created had some problems and they are really bloated, absolute paths etc....)
I'm using my own curses library (TUI) that can create frame, child windows, menus, cmdline, idle-loop etc in the same manner as e.g. MFC

I've also done some changes so that the display is cleared when exiting and the elapsed time and number of hashes/cracked is diplayed.
Have some more suggestions/changes I'm going to test...
Don't know yet if I'm going to continue with the 'old' one or change entirely to the 'new' multiforcer ?!!?

Do you continue working with the 'old' one?

Renfield wrote:I've just worked with the Multiforcer code for less than a week now, and I must say it's well written and easy to follow/understand.

Thank you. The goal of it is a framework that people can extend. It sounds like you are able to do this.

Renfield wrote:It was really easy (you already did some algo's I could snatch code from).
I just copy/paste the most from

Code: Select all

CHHashTypeSaltedSSHA.cu

and changed a bit....
Fixed the CHashType* files, Derived from CHHashFileSalted32, added to CHashes, etc.

Awesome. That's how it's supposed to work.

Renfield wrote:But if you want I can send you the code when I'm back next week.

Please do. I would like to integrate it into the existing code base. This is open source code, after all!

Renfield wrote:I think that is what I'm going to test. Don't understand the incrementor/shared mem code yet.

The old stuff is fairly straightforward, though the new framework is messier. It's not bad once you understand how I'm doing it.

I'm using my own curses library (TUI) that can create frame, child windows, menus, cmdline, idle-loop etc in the same manner as e.g. MFC

Please share. :p

I've also done some changes so that the display is cleared when exiting and the elapsed time and number of hashes/cracked is diplayed.
Have some more suggestions/changes I'm going to test...

Please share. :p

Don't know yet if I'm going to continue with the 'old' one or change entirely to the 'new' multiforcer ?!!?

Do you continue working with the 'old' one?

I would appreciate you changing entirely to the "new" code if possible.

The new framework is where all of my development effort is going, for several reasons:
- It supports CUDA, OpenCL, and CPU all at once instead of just CUDA
- The network code is updated to work properly across all hash types, not just unsalted hash types
- The OpenCL side of things is a good bit faster due to runtime metaprogramming
- The framework is better. It is written with 4+ years of experience in this realm, and is more general purpose.

So, if possible, please contribute to the new framework. The CUDA stuff is about the same - it should be straightforward to port things to the new CUDA framework, and I am happy to help come up with a good way of doing this.

If you have the time, I would love someone willing to bang out kernels for different algorithms - it's largely "cut and paste" as you know