Fingerphoto morphing attack generation using texture descriptors based landmarks.

Area of Science:

• Computer Science
• Biometrics
• Cybersecurity

Background:

• Smartphone biometrics, particularly fingerphotos, are increasingly popular for authentication due to usability and scalability.
• Existing fingerphoto verification systems are susceptible to sophisticated direct and indirect attacks.
• Morphing attacks, which blend features of multiple individuals, pose a significant threat to biometric security.

Purpose of the Study:

• To propose novel algorithms for generating realistic fingerphoto morphing attacks using smartphones.
• To evaluate the vulnerability of current fingerprint verification systems to these generated morphing attacks.
• To investigate the effectiveness of proposed detection algorithms for fingerphoto morphing attacks.

Main Methods:

• Developed three image-level algorithms to generate high-quality fingerphoto morphing images with minimal distortions.
• Conducted experiments on two distinct smartphone-captured datasets under varied environmental conditions.
• Implemented and tested detection algorithms leveraging both handcrafted and deep features for morphing attack identification.

Main Results:

• The proposed morphing algorithms successfully generated high-quality attacks that compromised commercial fingerprint verification systems.
• Fingerphoto verification systems demonstrated significant vulnerability to the generated morphing attacks.
• The developed morphing attack detection methods exhibited a high error rate, indicating challenges in accurate detection.

Conclusions:

• Smartphone fingerphoto biometrics are vulnerable to novel morphing attacks.
• Current detection methods struggle to accurately identify these sophisticated morphing attacks.
• Further research is needed to enhance the robustness of fingerphoto biometric security against morphing attacks.