Regular Expression Interpreter

[Hash-IT]

atom,

Can I make yet another feature request please, that ocl-hashcat-plus has a regular expression interpreter that runs using the GPU ?

I am amazed at just how powerful regular expressions are and think it would be a fantastic tool for ocl-hashcat-plus to have. You may have noticed the "Limiting the consecutive occurrence" topic is very popular. If we had this interpreter then users could make their own rules up without having to keep asking you to add them.

We could also achieve what my "Limiting the consecutive occurrence" topic set out to do as users could contribute regular expression code like M@LIK has done.

This feature could be a simple command from the user that ocl-hashcat-plus reads from a text file (supplied by the user) which contains all the regular expressions they want to apply to the input of ocl-hashcat-plus, whether reading from a password list or accepting direct input from maskprocessor.

This filtering would be great for slow algorithms such as WPA.

Thank you.

[undeath]

RegEx is much too slow.

[M@LIK]

Like undeath said, this needs a whole new level of programming.
To Re-program each and every single feature\code in regex into GPU-supporting language.

However, we can improve our own hashcat's rule engine to support some of regex codes, but that's in the hands of atom's.

[Hash-IT]

OK thanks guys, damn I thought this was going to be great !

[gat3way]

I believe implementing a regexp parser is possible on GPUs. Implementing some finite state automata to parse regular expressions is not something I believe would be very GPU-friendly though - because there would be lots of branching and memory accesses involved. Yet there was some CUDA-based project extending snort to parse network packets on GPUs. They reported some speed improvements over CPU overall, but the difference was not that huge.

[ntk]

there is a way to increase speed performance of regExp by factore of 10.without GPU.

[Hash-IT]

there is a way to increase speed performance of regExp by factore of 10.without GPU.

Is this possible on windows ?

[ntk]

sorry it is known only in GNUlinux

[Hash-IT]

sorry it is known only in GNUlinux

OK thanks for the info.

Have you set this up on your linux computer ? Could you filter the lists we are talking about on the other thread ? I can send you a script for generating the A-Z. If you can filter 10x faster that would be great !

[ntk]

I like you very much, but NO thanks I am afraid I could not doing all the list from A-Z for you. hihihi Space is the problem Hash-IT. 26 X 54GB is a lot. I am struggling already with clearing space for three of your list.

Unless you manage to doing magic by lifting them to some common space in the ether for us all, and by doing that also free my HD, otherwise I would not create then delete then create. I know it is fast enough with modified grep for this filter purpose, after testing 2nd list it is proven

I did post a comparison example and reason of change on your other thread too.

you can generate list for either n, t or k. then compare with mine 54GB (for each alphabet), created in 106min, if running 2 grep instances or 88min if one instance only. My PC is AMD 2.8GHz.