Benchmark K10 and Rejected hash
#1
Hi

I use cudaHascat and try to break "crypt" passwords with a dictionnary+rules file (Hardware: 3x Nvidia K10)

I think i have an issue with the speed (Very slow) compared to benchmarks. 112.7 kH/s is'nt very fast. When i use benchmark-mod=1 with "crypt", i have: 128.3 MH/s

An other strange thing is the number of rejected hash (almost 100% ?) How is it possible ?

Session.Name...: cudaHashcat
Status.........: Running
Rules.Type.....: File (cudaHashcat-1.01/crypt.rule)
Input.Mode.....: File (dico.lst)
Hash.Target....: File (crypt.hashes)
Hash.Type......: descrypt, DES(Unix), Traditional DES
Time.Started...: Tue Feb 11 18:59:32 2014 (6 secs)
Time.Estimated.: Fri Feb 21 20:28:58 2014 (10 days, 1 hour)
Speed.GPU.#1...: 18777 H/s
Speed.GPU.#2...: 18781 H/s
Speed.GPU.#3...: 18804 H/s
Speed.GPU.#4...: 18809 H/s
Speed.GPU.#5...: 18768 H/s
Speed.GPU.#6...: 18766 H/s
Speed.GPU.#*...: 112.7 kH/s
Recovered......: 0/1432 (0.00%) Digests, 0/1147 (0.00%) Salts
Progress.......: 60530238764562/158790171105922 (38.12%)
Rejected.......: 60513402828306/60530238764562 (99.97%)
HWMon.GPU.#1...: 99% Util, 50c Temp, -1% Fan
HWMon.GPU.#2...: 99% Util, 59c Temp, -1% Fan
HWMon.GPU.#3...: 99% Util, 55c Temp, -1% Fan
HWMon.GPU.#4...: 99% Util, 58c Temp, -1% Fan
HWMon.GPU.#5...: 99% Util, 56c Temp, -1% Fan
HWMon.GPU.#6...: 99% Util, 59c Temp, -1% Fan

And rules to generate "crypt.rule":

./mp64.bin -1 0 ":" > crypt.rule
./mp64.bin -1 0 "r" >> crypt.rule
./mp64.bin -1 0 "d" >> crypt.rule
./mp64.bin -1 0 "c" >> crypt.rule
./mp64.bin -1 0 "cd" >> crypt.rule
./mp64.bin -1 0 "cr" >> crypt.rule

./mp64.bin -1 0123456789 "^?1'8" >> crypt.rule
./mp64.bin -1 0123456789 "^?1^?1'8" >> crypt.rule
./mp64.bin -1 0123456789 "^?1^?1^?1'8" >> crypt.rule
./mp64.bin -1 0123456789 "^?1^?1^?1^?1'8" >> crypt.rule


Where is my mistake ?


Thx
#2
You have 1147 distinct salts, therefore the speed is divided by 1147 (since each word must be combined w/ those 1147 salts, roughlty speaking).

112.7 kH/s * 1147 = 128.464 MH/s

Descrypt passwords are limited to 8 characters. So the range is from 0 length to 8 length, everything above that length will be rejected.
Therefore you should chose your input files/rules more carefully (especially for descrypt).

Hope this helps
#3
Thx for the calcul, very helpful.

I'm aware about the descrypt specification but there is the same number of rejected password if i add '8 to my rules...

Edit: I find the answer by myself: "The problem is that a word greater than 8 character gets rejected before applying the rules so you cannot use rules to truncate the words to 8 characters"