Evaluation of local texture descriptors for eyebrow-based continuous mobile user authentication

Abstract

Mobile user authentication plays an important role in securing physical and logical access, especially to globally ubiquitous smart phones. Several studies have evaluated face, ocular and finger modalities for mobile user authentication. Human eyebrow is among the less explored traits for mobile biometric use cases where device front facing cameras can easily scan them. A handful of studies suggest the potential of human eyebrows for person authentication. Using Histogram of Oriented Gradients (HOG) based texture descriptors, we show equal error rates as low as 15.32% and areas under ROC curve as high as 0.92 on publicly available VISOB dataset when fusing left and right eyebrow units.