Plenoptic Face Presentation Attack Detection

Area of Science:

• Computer Science
• Biometrics
• Optical Imaging

Background:

• Current face recognition systems are vulnerable to presentation attacks, limiting their biometric applications.
• Presentation attacks, such as spoofing with photos or masks, pose a significant security risk.
• Existing detection methods often struggle with diverse and sophisticated attack types.

Purpose of the Study:

• To develop a passive presentation attack detection (PAD) method for face recognition systems.
• To leverage complete plenoptic imaging for robust PAD.
• To evaluate the proposed method against various presentation attack types.

Main Methods:

• Utilized a complete plenoptic imaging system to capture the full light field information.
• Derived the complete plenoptic function of light rays using a single detector.
• Constructed a multi-dimensional face database including 50 subjects and seven types of presentation attacks.

Main Results:

• The proposed passive PAD method demonstrated superior performance compared to state-of-the-art techniques.
• Effectively detected all seven types of presentation attacks included in the study.
• Achieved higher accuracy and reliability in distinguishing genuine faces from spoofed ones.

Conclusions:

• Complete plenoptic imaging offers a promising approach for advanced passive presentation attack detection.
• The developed method significantly enhances the security and reliability of face recognition systems.
• This technology has the potential to broaden the secure application of biometrics in real-world scenarios.