CAPTCHA - Completely Automated Public Turing Test To Tell Computers and Humans Apart[1] The project is a MATLAB tool to decipher CAPTCHA images for 10 different classes of CAPTCHA using Machine Learning.
The objectives behind the project are manifold; firstly, the application will help us uncover a few security issues with the text CAPTCHA, which would stress for the need to develop more complicated CAPTCHA and thus enhance the strength of one of the parameters on which web security is based. The idea was chosen in order to show the concerns in today’s world where the security methods developed to preserve confidentiality in online systems, of which image based CAPTCHA represent of the latest developments, are not only insecure but are prone to attacks by hackers with high success rates. Secondly, it will also help to provide an aid to the millions of visually impaired people around the world that are using the web. Lastly, breaking a visual text based CAPTCHA would definitely mean that a computer performing the way humans think, and thus could help greatly in furthering the study in Computer Vision and Artificial Intelligence.
- The repository contains complete code including database files used for training the neural networks.
- Summary of set up
- Configuration
- Dependencies
- Database configuration
- How to run tests
- Deployment instructions
- Writing tests
- Code review
- Other guidelines