Enhancing weak client-server user authentication with facial recognition. iOS client (Swift + Objective-C++) ↔ Python server. Face and landmarks detection courtesy of Vision.framework, recognition via OpenCV. Pictures are heavily preprocessed in order to reduce noise, remove the background and align them based on the detected position of the pupils. The computed face model is AES256 encrypted via the user's password, and it is decrypted only upon login. Passwords are hashed and salted, of course.
Project for the Data Security exam. It is just a POC of the technique hacked up in relatively little time, so don't expect it to be secure or up to any standard.
- Edit AppConfig.swift and cert_gen.sh with your local server name and cert info.
- Install OpenCV locally:
brew install opencv - Install required Python packages via
requirements.txt. If you're using virtualenv, you may need additional configuration to make OpenCV work. - Get or compile opencv2.framework for iOS (tested 3.4.1). It must come with the
contribmodule. - Put
opencv2.frameworkin theclient/Externaldir. - Run the server via start.sh.
- Run the client on your iOS device.
- Have fun.