Benchmarking desktop and mobile handwriting across COTS devices: The e-BioSign biometric database

Area of Science:

• Biometrics
• Human-Computer Interaction
• Computer Science

Background:

• Dynamic signature and handwriting verification are crucial for secure authentication.
• Existing datasets often lack diversity in capture devices and writing tools, limiting real-world applicability.
• Sensor interoperability presents a significant challenge in biometric systems.

Purpose of the Study:

• To introduce the e-BioSign database, a novel resource for dynamic signature and handwriting analysis.
• To evaluate signature verification performance across diverse Commercial Off-The-Shelf (COTS) devices and writing instruments (pen vs. finger).
• To establish a benchmark for person verification in realistic, sensor-interoperable scenarios.

Main Methods:

• Acquired dynamic signature and handwriting data from 65 subjects using five COTS devices (Wacom, Samsung tablets).
• Collected data with both pen stylus and finger input on tablets to simulate mobile scenarios.
• Performed skilled forgeries and evaluated verification using feature- and time-function-based approaches under intra-device, inter-device, and mixed writing-tool conditions.

Main Results:

• Demonstrated the critical need for specialized methods to handle device interoperability challenges in signature verification.
• Highlighted the necessity for further investigation into the performance of finger-based signature acquisition.
• Established baseline performance metrics for signature verification across varied sensor and tool conditions.

Conclusions:

• The e-BioSign database provides a valuable public resource for advancing signature biometrics research.
• Findings underscore the importance of developing robust systems that account for device diversity and varied input methods.
• Future research can leverage this database to improve the understanding and accuracy of electronic signature verification.