This is a MATLAB code that cracks the EZ-Gimpy CAPTCHA. The cracking process consists of 3 steps:

1. Removing the background from the CAPTCHA image. This is done by converting the CAPTCHA image into a binary image

2. Vertically segment the image to extract the letters of the CAPTCHA word

3. We use a character recognition algorithm that recognizes the extracted characters

To run the cracker with the dataset provided in the folder "Dataset" just type

[correct, wrong, failed] = hack_captchas;

The correct attribute holds the number of correctly cracked CAPTCAHA, wrong contains the number of the wrongly cracked CAPTCHAs and failed contains the number of the CAPTCHA images that failed in the cracking process due to exceptions

Dataset (folder): contains the CAPTCHA images. The dataset is taken from http://www.cs.sfu.ca/~mori/research/gimpy/ez/

Training Set (folder): contains the characters which are used as training set

chars.mat: represents a hash the key is the character number and the value is the character

p.mat: represents the features vector

t.mat: a matrix that represents the excepted output

hack_captchas: checks the "Dataset" folder and loops through all the CAPTCHA images in it to extract the word in it

extract_word.m: takes a CAPTCHA image and returns the word in it

hierarchicalCentroid.m: the features extracting algorithm used to extract the features from the characters

training_data.m: traing the characters images in the "Training Set" folder