A fingerprint spoof detection system that uses Local Binary Pattern Histogram (LBPH) features (proposed in A comparative study of texture measures with classification based on featured distributions). It uses Support Vector Machines (SVM) as a classifier.
Python 3
scikit-image
scikit-learn
NumPy
OpenCV
Run the code using: python main.py
These days, anyone can easily fabricate the fingerprint of anyone with the help of latex, gelatin, etc. claim his/her indentity. To avoid this, this fingerprint spoof detection system has been made. Given an input image of a fingerprint, it classifies it as spoof or real. The accuracy using a simple linear SVM is up to more than 80% (and can be improved much using other complex models). This system can be used as an effective counter-measure against spoof attacks.
The model was trained & tested on the LiveDet2011 Dataset . The dataset has not been uploaded due to copyright issues (The tree structure of the dataset is here in this code however for convenience)