Can brute force be useful ?

For example i pretty much find xanga network catpchas easy
But success rate is 4%

lets say if i manually set 200 questions, can brute force make proper rules out of this ?

I have 8 cores 4.5 ghz , 32 gb ram

with questions you mean 200 different captchas? Well I used 125 captchas here to get things done...but yes im sure it can be improved...just that brute force might not get you any further...maybe the new "Mask" function will be a help here. 

@sven

@sven any proper way there ?
actually i really dont want to spend manul time
if possible let computer work on multi threading

is that possible ?
i mean by 200 different captchas from xanga network type - so 200 different question images

and how do we use this mask function ?

also can you please add not solve catpchas option
save as folders

so i can focus on certain networks such as xanga

save unsolved/unknown captchas is already an option. Using Masks is something more complex and maybe not a good way to start. Ozz had some good tutorials written (also sticky thread).

@Sven

@Sven yes i tried that but it does not save as folders
save all in same place

i want to save them as folders. so type is determined but software were not able to solve
i want those to be saved as folders so i can particularly focus on them

such as all xanga catpchas saved under xanga folder

this is what should be in the "save (un)solved captcha to" field:

.\solved\%engine%\

.\unsolved\%engine%\

make sure you tick both options (solved & unsolved) as you want to get a good mixture of both depending how often they occured.

here are some samples to start with:

http://rghost.net/private/45003504/ce509a1874934cae7765706e3717b07a

@OZZ

@OZZ thanks. i have deleted that %engine%

@Sven

@Sven can you make brute force multi threaded ?

it's as much multi-threaded as possible. Some functions have to be called from main thread which slows it down a bit. Sorry, nothing much to do here unless i recode the whole image functions none vcl based....but that can take a long time.

@sven

@sven captcha breaker brute force is single threaded
uses only 1 thread

and it really does improve the performance

i used what @Ozz given and increased success rate from 4% to 6.77%


also what does processing number do ?

i talked about the process numbers in this tutorial:

https://forum.gsa-online.de/discussion/1378/sdk-for-dummies-how-to-improve-existing-captcha-definitions

to make it short, if "Process 01" didn't get any result for one captcha, than "Process 02" will try to solve it and so on (up to 10 processes overall).

@Ozz

@Ozz it is still confusing

so lets say i said find best catpcha for all images and used process 1
after processes finished saved it

then i will set process 2 and once again brute force all images

and when i save it, it will be saved as process 2 ?

so when that defined type catcpha came, it will first try process 1 and then if it fails it will try process 2 ?

>so when that defined type catcpha came, it will first try process 1 and then if it fails it will try process 2 ?

yes, that is how it works. if you have more than 1 process and click "Test" than it will give you the best order how CB should order the processes as well.

>then i will set process 2 and once again brute force all images

not all images. you just take the captcha CB didn't get an answer for (regardless if the answer was correct or not). just take every step i talked about in the tutorial. its very specific about that and if you do it the way i told than you should understand how it works imo.

>and when i save it, it will be saved as process 2 ?

if you defined it at "Process number = 2" then yes. if you overwrite "Process 1" than not.

@Ozz

@Ozz it seems like best way is
generating like 10 different set of images
such as each one is 100 images

brute force each set with best solution for all images and save each of them as 1 filter

with this way you can obtain 10 different filters

the main problem is brute force is not working multi threded so it is pretty slow

no, best way is to

- brute force all images in process 1 -> save and delete all images that gave an answer afterwards

- brute force remaining in process 2 -> save and delete all images that gave an answerd

- brute force ramaining images of process 2 in process 3 -> ....

yes, it takes time. you are able to run the SDK on every machine though, even if the trial period is over.

to speed up things you can only use a fraction of those captchas to get a working definition fast which will speed up brute forcing with more images.

so just take 100 and brute force a definition and take the rest of your captchas later so you have a solving rate of XX % right from the start for the complete captcha database.

@Ozz

@Ozz thanks it seems even better approac

i hope sven adds multi threading since this process can become multi thread pretty easy