statgen giving same result regardless of list
#1
I'm using two different password lists to generate hcstat files: rockyou and the linkedin passwords. now, linked in has the keyword "link" more times than any other word, but regardless of that, i get the same stats as rockyou. Also, I noticed the process seems almost too fast to actually go through every line. if i do a length of 4 i always get "mari" first on both lists, telling me that either i'm doing something wrong or something isn't working right. any advice?

Code:
hcstatgen.bin linked.hcstat < linkedin.txt

Code:
sp64.bin --pw-min=4 --pw-max=4 --threshold 5 linked.hcstat | more

I've tested with different thresholds but still get the same result. i grep my lists to confirm, but it shows link is a more common keyword
#2
it's not based on keywords, it's based on per-position statistical analysis.

the threshold is also way too low. i don't know what thresholds you've tried, but the current threshold and length you provided will only output 5^4 words using the top 5 characters in each position.
#3
by the way, the best way to test is to try to crack the same list with both hcstat files, and see which one is more effective. if you find that the rockyou-based hcstat is more effective against the linkedin list than the linkedin-based hcstat, then there is a problem.
#4
I'm not trying to push in here and epixoip is considerably smarter than I am but I think the way statgen works is going to be changed.

I "think" at the moment Markov has some hard coded bias and I guess the change to it will be that this bias is removed and it will work exactly according to the input list.
#5
yes, the latest beta is quite different
#6
(03-19-2013, 02:27 PM)epixoip Wrote: yes, the latest beta is quite different


Thanks epixoip, do you have any better idea how it will work ? I am hoping and assuming it will be purly based on the sample list given.
#7
i'll defer to atom for this one
#8
(03-19-2013, 04:36 PM)epixoip Wrote: i'll defer to atom for this one

Smile OK thanks.
#9
I am not sure yet how to implement it in a way a GPU could benefit from it. but since you are only interessted in WPA (i finally learned that Smile) you can still use stdin pipe. So when the new standalone markovprocessor comes out you will see a big difference depending on the input.
#10
(03-21-2013, 02:25 PM)atom Wrote: I am not sure yet how to implement it in a way a GPU could benefit from it. but since you are only interessted in WPA (i finally learned that Smile)

Yay ! Big Grin

Unfortunately just as you have learned that, I am starting to become interested in other hashes now ! ha ha ! However I think WPA will always be my favorite Smile

(03-21-2013, 02:25 PM)atom Wrote: you can still use stdin pipe. So when the new standalone markovprocessor comes out you will see a big difference depending on the input.

I cannot wait for this Smile I am really looking forward to it. I hope it will only take into account the input list.

Will we always have to pipe ? Will this never be built into oclhashcat ? You know my ongoing mission to reduce code and command lines so please just humor me Smile