Captchonka is a pentesting tool to brute force captchas

Run help to see available options python captchonka -h or python captchonka --help.

You can also run it as ./captchonka on Unix systems.

All commands should be run within a module (mod) python captchonka --mod=violet (this will do nothing).

To see the list of available modules do python captchonka --list.

Before cracking you have to train your algorithm. To do so do:

• Run python captchonka --mod=violet --train="input/example.png"
• If process is successful then a list of characters will be saved in output/char/
• Now move manually each characted from output/char/ to core/mods/violet/char/LETTER

You can run training in verbose mode to see processed image at each step python captchonka --mod=violet --train="input/example.png" -v.

Detected characters may be automatically moved into their corresponding folders. For this:

• Name your image file so that it will contain the code in square brackets: example-[GPCKC].png. Square brackets can be placed anywhere in name
• Run training in auto-mode python captchonka --mod=violet --train "input/example-[GPCKC].png" -v -a

If code will be found in image name and its length will correspond to number of found characters then they will be moved into according folders.

Crack a captcha by running python captchonka --mod=violet --crack="input/example.png".

Run in verbose mode to see what happend under the hood and what are the guess probabilities python captchonka --mod=violet --crack="input/example.png" -v.

You can create your own module by creating a folder in core/mods/. This folder should have a file with the same name as folder name. In best case scenario you have to implement only cleanImage method. It should return a black-and-white image with no noise. Use core/mods/violet/violet.py as an example.

If overriding cleanImage is not enough you may look also override following methods:

• cleanImage - get a raw PIL Image and return one without noise
• getCharacters - extract a list of PIL Images ordered by their position in original image
• checkCharacterSizes - test if character size matches requirements
• getCodeFromString - get code from image title
• computeSimilarity - compare similarity between two 2D binarry lists, return a float between 0 and 100

If you have multiple PNG files in a folder (lets say input/test/) than you could run for f in input/test/*.png; do echo $f; python captchonka --mod=violet --crack="$f"; done

Alternatively you test efficency by cracking a batch of images. Run python captchonka --mod=violet --batch-crack="input/test/".

If you have multiple PNG files for training and each file has its code in name than you could run for f in input/violet/*.png; do python captchonka --mod=violet --train="$f" -a; done

Alternatively you can training a batch of images running python captchonka --mod=violet --batch-train="input/violet/" -a

• Red - error
• Green - success
• Blue - info
• Yellow - warning

• Python 2.7
• NumPy
• SciPy
• OpenCV

• Brew with Science brew tap homebrew/science
• Python brew install python --framework --universal
• Python pip (if it is not installed with Python) easy_install pip
• NumPy pip install numpy
• SciPy brew install homebrew/python/scipy
• OpenCV brew install opencv
• Termcolor pip install termcolor

• SciPy Advanced image processing

Based on CIntruder.

General Public License v3