Mask+Dict Limitations?
#1
I've looked around and searched quite a bit but I can't seem to find an answer to this conundrum.

Whenever I am launching an attack that runs dict+mask, I have no problems at all. Everything works fine.

A Mask+Dict attack, however, is running on only one gpu (I've tried setting --gpudevices and am on Windows, so I shouldn't need to register the cards. Plus, they work fine on other cracking modes.)

It is also running extremely slow in some cases (?d+dictionary = 28kk c/s vs dictionary+?d = 550m c/s)

Is this just a fundamental limitation of the hash cracking mode, or am I missing something?

Setup is 2x Radeon 6950, 1x 6870. Hash types are md5.
#2
you aren't giving the left side enough work to do, therefore it cannot reach full acceleration.
#3
(02-16-2013, 10:22 PM)epixoip Wrote: you aren't giving the left side enough work to do, therefore it cannot reach full acceleration.

The same issue happens even with a run of ?d?d?d?d?d?d, or something equally pressing.

I'm running all tests at loops 1024/accel 40, so Im pushing the card to 99% load, it's just.. running really slow with some combinations of mask+dict.

If there were issues with not having enough load, Dict+?d?d?d?d?d?d should have the same theoretical speed as ?d?d?d?d?d?d+Dict, but that's just not the case.
#4
It is.
-a6 and -a7 are totally different when it comes to GPU utilization. Read:
Quote:- if you use -a 0, -a 1 or -a 6, make sure the base dictionary (left) has at least 10 mio words (per gpu).
- if you use -a 3 or -a 7, make sure your left mask keyspace > 500k entries (per gpu). for -a 3, this is the keyspace of the first 4 chars of your mask.
- make sure your modifier (right) iterates at least through a keyspace of 1k. in case if -a 0 its the number of rules, in case of -a 1 and -a 7 its the number of words in the right dictionary, in case of -a 6 its the keyspace of the right mask.
#5
I'll have a look and see if that alleviates the problem.

Does this mean that the cracking will always underperform if I want to check masks that aren't requiring a lot of keyspace, or is there some method that I can use to get good performance out of a small keyspace when using it as a leftmask? I guess rules might work.
#6
True.

Rules often feed GPU well.
Big dict, many rules and you should be fine.
#7
It's not "underperforming," you're still getting the best performance you can get for that particular workload. It's just not able to achieve full acceleration because that particular workload happens to be too light.

Quote:Dict+?d?d?d?d?d?d should have the same theoretical speed as ?d?d?d?d?d?d+Dict

No, not true at all.
#8
(02-16-2013, 11:24 PM)M@LIK Wrote: It is.
-a6 and -a7 are totally different when it comes to GPU utilization. Read:
Quote:- if you use -a 0, -a 1 or -a 6, make sure the base dictionary (left) has at least 10 mio words (per gpu).
- if you use -a 3 or -a 7, make sure your left mask keyspace > 500k entries (per gpu). for -a 3, this is the keyspace of the first 4 chars of your mask.
- make sure your modifier (right) iterates at least through a keyspace of 1k. in case if -a 0 its the number of rules, in case of -a 1 and -a 7 its the number of words in the right dictionary, in case of -a 6 its the keyspace of the right mask.
I knew the general logic of this but never seen the actual suggested limits. Where did you get that from? I will definitely have to optimize some of my attacks knowing these.
#9
mastercracker Wrote: I knew the general logic of this but never seen the actual suggested limits. Where did you get that from? I will definitely have to optimize some of my attacks knowing these.

Guess what? It's da Boss himself.
https://hashcat.net/forum/thread-1527-po...ml#pid8590
#10
Ah ok. Should be added to the wiki somewhere. Probably in Patches, tips and tricks or maybe create a FAQ section.