Handcrafted Feature-Based Fingerprint Presentation Attack Detection

To distinguish between genuine and fraudulent fingerprints, this paper introduces a feature-based approach for effectively classifying fake and live fingerprints. The suggested approach uses texture extraction techniques to extract unique features from input fingerprint photos. This research article uses texture descriptors such as Gabor filters, Local Binary Patterns, and Phase Quantization. Each picture's complete feature vector is created by concatenating these features, which are then input into a Random Forest Classifier to do classification. Using a test dataset, the model's performance is assessed and it shows a 96% overall accuracy rate with balanced precision, recall, and F1 scores for the “Fake” and “Live” fingerprint classes. The Random Forest Classifier enhances the model's scalability, resilience, and capacity to manage high-dimensional data.