Spoof Detection for Finger-Vein Recognition System Using NIR Camera.

Area of Science:

• Biometrics and Security
• Computer Vision
• Machine Learning

Background:

• Finger-vein recognition offers high performance but is vulnerable to presentation attacks (spoofing).
• Existing presentation attack detection (PAD) methods rely on handcrafted features, limiting their effectiveness.
• Deep learning has shown superior performance in various computer vision tasks.

Purpose of the Study:

• To propose a novel PAD method for near-infrared (NIR) camera-based finger-vein recognition.
• To enhance the detection capabilities beyond traditional handcrafted feature extractors.
• To improve the accuracy and robustness of finger-vein recognition systems against spoofing attacks.

Main Methods:

• Utilized a Convolutional Neural Network (CNN) for automatic feature extraction tailored for PAD.
• Applied Principal Component Analysis (PCA) for dimensionality reduction of extracted features.
• Employed a Support Vector Machine (SVM) for the final classification of real and fake finger-vein images.

Main Results:

• The proposed CNN-based method demonstrated superior performance in detecting presentation attacks.
• The combined CNN, PCA, and SVM approach outperformed previous handcrafted methods and standalone CNN methods.
• Experimental results confirmed the method's adequacy for robust finger-vein PAD.

Conclusions:

• The developed deep learning-based PAD method effectively enhances the security of finger-vein recognition systems.
• This approach offers a significant improvement over existing techniques for combating spoofing attacks.
• The study validates the potential of CNNs combined with PCA and SVM for advanced biometric security.