Bio-Inspired Presentation Attack Detection for Face Biometrics.

Area of Science:

• Computer Science
• Biometrics
• Artificial Intelligence

Background:

• Face biometric systems are crucial for identity authentication in security settings like border control.
• The increasing use of portable devices with cameras and displays presents new vulnerabilities for face spoofing.
• Existing biometric systems face challenges with presentation attacks, requiring advanced detection methods.

Purpose of the Study:

• To explore vulnerabilities in face biometric systems, specifically presentation attacks.
• To introduce and evaluate a novel biologically-inspired presentation attack detection model called BIOPAD.
• To compare BIOPAD's performance against established models like HMAX and CNN.

Main Methods:

• Developed BIOPAD, a biologically-inspired model using Gabor features in a hierarchical structure.
• Processed visual information from faces and presentation attacks in visible and near-infrared spectral regions.
• Compared BIOPAD with HMAX and CNN on three presentation attack databases using SVM and k-NN classifiers.

Main Results:

• BIOPAD demonstrated superior performance over HMAX and CNN across all tested databases and classifiers.
• Achieved authentication rates with up to 99% accuracy in certain cases.
• Introduced a new visible and near-infrared presentation attack database for comparative analysis.

Conclusions:

• BIOPAD's unique approach of fusing spectral band information for face presentation attack detection is novel.
• Near-infrared visual information significantly enhances the ability to overcome presentation attacks.
• The BIOPAD model shows promising detection rates, improving the robustness of face biometric systems.