Research Article Open Access

Effects of Visible and Near Infrared Polarized Lights on Spoofing Face Detection

Azim Zaliha Abd Aziz1
  • 1 Universiti Sultan Zainal Abidin, Malaysia

Abstract

Face spoofing countermeasure is vital to avoid an imposter from gaining access to security biometric systems by using face masks in various forms that mimic a valid user face. Recently, several studies have shown the ability of visible polarized light in distinguishing real and fake faces. In this paper, polarization imaging systems using visible and near infrared (NIR) are proposed to examine the effects on the polarization images as trial to distinguish between genuine face and spoof face presentation attacks: photo paper and iPad face display; based on the optical properties of the materials. The findings from the investigations suggest that in general, NIR light could not be used to distinguish between genuine face and photo paper under a polarization lighting condition. In contrast, the visible light provides significant difference of the Stokes images between the materials. Classification between real face and iPad display can easily be done by manipulating the polarization angle. A new feature fusion formula named as the SDOLP3F is introduced to differentiate between the real and the fake traits. The SDOLP3F results presented in this paper show the highest accuracy rate compared to the individual measures. The results illustrate the robustness of the proposed anti-spoofing algorithm based on a small sample.

Journal of Computer Science
Volume 15 No. 2, 2019, 288-301

DOI: https://doi.org/10.3844/jcssp.2019.288.301

Submitted On: 23 December 2018 Published On: 23 February 2019

How to Cite: Abd Aziz, A. Z. (2019). Effects of Visible and Near Infrared Polarized Lights on Spoofing Face Detection. Journal of Computer Science, 15(2), 288-301. https://doi.org/10.3844/jcssp.2019.288.301

  • 3,010 Views
  • 1,595 Downloads
  • 0 Citations

Download

Keywords

  • Anti-Spoofing
  • Face Biometric
  • Polarized Light
  • Polarization Imaging System
  • Spoofing Attacks