On the Effectiveness of Impedance-Based Fingerprint Presentation Attack Detection

Area of Science:

• Biometrics
• Cybersecurity
• Sensor Technology

Background:

• Biometric recognition is a reliable authentication method, but vulnerable to presentation attacks (PAI).
• Fingerprint sensors are susceptible to attacks where fake fingerprints are presented to capture devices.
• Existing systems lack robust defenses against sophisticated presentation attack instruments.

Purpose of the Study:

• To develop a novel presentation attack detection (PAD) method for fingerprint recognition.
• To enhance the security of thin-film transistor (TFT) based fingerprint capture devices.
• To differentiate between genuine fingerprints and artificial presentation attack instruments (PAI) using impedance measurements.

Main Methods:

• Modified a thin-film transistor (TFT) fingerprint capture device to acquire finger impedance data.
• Trained a presentation attack detection (PAD) algorithm using impedance values from bona fide and artificial fingers.
• Evaluated the PAD algorithm on a dataset of 42 different presentation attack instrument (PAI) species.

Main Results:

• The developed PAD algorithm achieved high accuracy in detecting most presentation attacks.
• Achieved an Actual Presentation Attack Detection Error Rate (APCER) of 2.89% in a user-friendly scenario.
• Achieved a Bona Fide Presentation Classification Error Rate (BPCER) of 0.2% in a user-friendly scenario.
• Identified a weakness in detecting certain unknown presentation attack instrument (PAI) species.

Conclusions:

• Fingerprint impedance measurement is a viable method for presentation attack detection (PAD).
• The modified TFT sensor offers improved security against common presentation attacks (PAI).
• Further research is needed to address detection vulnerabilities against novel or unknown presentation attack instruments (PAI).