Visible-Light Camera Sensor-Based Presentation Attack Detection for Face Recognition by Combining Spatial and...

Area of Science:

• Computer Science
• Biometrics
• Artificial Intelligence

Background:

• Face-based biometric systems are widely used but vulnerable to spoofing attacks using artificial facial images.
• Existing face presentation attack detection methods primarily rely on spatial information, limiting their effectiveness.

Purpose of the Study:

• To develop a robust face presentation attack detection system utilizing both spatial and temporal information.
• To improve the security of face recognition systems against sophisticated spoofing techniques.

Main Methods:

• A novel approach combining deep features from a stacked convolutional neural network (CNN)-recurrent neural network (RNN) with handcrafted features.
• Utilizing visible-light camera sensor data to capture both spatial and temporal facial dynamics.
• Experiments conducted on two public datasets to validate the proposed method.

Main Results:

• Temporal information alone is sufficient for detecting face image-based attacks.
• Handcrafted image features significantly enhance the detection performance of deep features.
• The proposed method demonstrates superior performance compared to existing state-of-the-art techniques.

Conclusions:

• The integration of spatial and temporal information provides a more effective solution for face presentation attack detection.
• The proposed CNN-RNN and handcrafted feature fusion method offers enhanced security for biometric recognition systems.
• Future work can explore further optimization and real-world deployment of this advanced detection system.