Detecting Morphing Attacks through Face Geometry Features.

Area of Science:

• Computer Science
• Biometrics
• Digital Forensics

Background:

• Face-morphing technology creates composite images challenging identity verification systems.
• Automated detection of morphed passport photos is crucial for border security applications.
• Existing methods often analyze single images, limiting their effectiveness.

Purpose of the Study:

• To develop a differential approach for detecting morphed passport photos.
• To leverage facial landmark analysis for identifying morphing artifacts.
• To assess the effectiveness of this method in a real-world border control scenario.

Main Methods:

• Analyzing the geometric inconsistencies of facial landmarks between passport and probe images.
• Utilizing automated algorithms to detect landmark locations.
• Employing supervised classifiers to identify morphed image pairs.

Main Results:

• Facial landmark locations provide discriminative information for detecting morphed passport photos.
• The differential analysis approach shows promise in identifying manipulated images.
• Classifier sensitivity to training and testing data compositions was explored.

Conclusions:

• Facial landmark analysis in a differential scenario is an effective method for detecting morphed passport photos.
• This technique can significantly improve the security of machine-assisted border control.
• Further research can explore different feature transformations for enhanced detection.