Short-Term User Authentication Using Eyebrows Biometric For Smartphone Devices

With the unprecedented mobile technology revolution, biometrics have become a viable alternative to PINs and passwords for secured mobile access and transactions. Recent research has shown the potential of eyebrows as an independent biometric modality for recognition. Eyebrows leverage the benefit of being one-sixth of the facial region, and is therefore computationally efficient and offer fast throughput. The aim of this paper is to exploit eyebrow region for mobile user authentication, and to show the impact factor such as eyeglasses existence on eyebrow-based user authentication. To this aim, Histogram of Oriented Gradients (HOG), a local descriptor, and GIST, a global descriptor, extracted from left and right eyebrow regions are evaluated. The reported results on a publicly available VISOB dataset collected through mobile devices. show a minimum Equal Error Rate (EER) of 3.23% and Area under curve (AUC) of 0.9916 obtained from fusion of left and right eyebrow regions at score level using GIST descriptor. Further, comparative evaluation with existing methods is performed by evaluating the proposed method on FERET face dataset obtaining minimum Equal Error Rate (EER) of 1.08% and Area under curve (AUC) of 0.9990 on fusion of left and right eyebrow regions at score level without eyeglasses using HOG descriptor.