+- hashcat Forum (https://hashcat.net/forum)
+-- Forum: Deprecated; Previous versions (https://hashcat.net/forum/forum-29.html)
+--- Forum: Old oclHashcat Support (https://hashcat.net/forum/forum-38.html)
+--- Thread: Openwall cracker ?! (/thread-3082.html)
Openwall cracker ?! - Mem5 - 01-27-2014
Hi,
Have you seen this twitt ?
https://twitter.com/HashSuite/status/427671690319892480
Quote:As far as we're aware, Hash Suite 3.0 is now world's fastest LM hash cracker on CPU and GPU: 5350M c/s on 7970GE (vs. oclHashcat's <1400M)
From openwall : http://hashsuite.openwall.net/performance
I will test it tomorrow but.. 5350M c/s VS 1400M c/s that's a huge difference !
Any hints ?
RE: Openwall cracker ?! - atom - 01-27-2014
Keep cool, there's no magic involved.
What this program is using is a technique called bitslice. DES based algorithms like LM can be improved alot by using bitslice compared to "traditional" techniques.
The original bitslice DES/LM source was posted on reddit a while back and at this time it was doing around 3000M c/s. I was able to improve it to around 6400M c/s which is actually faster than Hash Suite 3.0, you can read about here: http://www.reddit.com/r/crypto/comments/162ufx/research_project_opencl_bitslice_des_bruteforce/#c7sm2c3
Again, this technique works only for LM hashes, but for LM hashes you should use rainbow tables. If you really need to brute-force the entire 95^7 keyspace you can do that in less than a day on 1400M c/s since the algorithm is so designed that the keyspace is so small.
RE: Openwall cracker ?! - Sc00bz - 01-28-2014
The "full" keyspace is 69^7 so 1.5 hours vs 24 minutes.
Unless you add in foreign characters. These change depending on the system's code page. I tried to figure out the mapping and failed because I didn't know that it was depended on the system's code page. I got 105 characters with the English code page. Anyway it really only matters if you say fuck it and do the full full keyspace of 2^56.
RE: Openwall cracker ?! - Mem5 - 01-29-2014
(01-27-2014, 07:20 PM)atom Wrote: The original bitslice DES/LM source was posted on reddit a while back and at this time it was doing around 3000M c/s. I was able to improve it to around 6400M c/s which is actually faster than Hash Suite 3.0
Thanks for the clarification.
Why oclhashcat does not implement it ?
RE: Openwall cracker ?! - Mem5 - 01-30-2014
Considering oclhashcat last version and this command line :
Code:
-a 3 -m 3000 -n 160 --markov-disable -i --increment-min 1 --increment-max 7 -1 ?u?d?s 1000_hashes.txt ?1?1?1?1?1?1?1Speed is ~1200 Mh/s, duration ~1h30.
ATI 7970
1/ Any hints to speed it up ?
2/ is "?u?d?s" the same behaviour as "?a" when bruteforcing LM ?
Thank you.