Popular mobile banking and e-commerce applications like Google Wallet, Apple Pay and Ali Pay have resulted in using personal devices like smartphones for secure access of services via biometric data captured from embedded sensor. Further, the improved optics on smartphones have been explored for biometric data capture in a contactless manner that can be used for various secure authentication applications. Specifically, the applications are exploring face, periocular, iris and finger photo characteristics for smartphone-based authentication using the embedded camera operating in the visible spectrum. This thesis is dedicated to explore iris, periocular and face data for authentication applications when captured from smartphones in the visible spectrum. We first present robust algorithms to use iris recognition for the data captured from the smartphones in an unconstrained manner. Further, we propose a new imaging setup to resolve the iris texture for images captured in the visible spectrum for subjects with heavily pigmented iris. As an alternative to mitigate the lower performance of iris recognition in the visible spectrum, we demonstrate the use of periocular characteristics for authentication. We then present a robust algorithm for feature extraction from periocular images for verification purpose. We also present a multi-biometric authentication system fully realized on the smartphone with good verification accuracy for secure access applications. In the end, we present a set of robust algorithms to detect the presentation attacks on the ocular biometric systems working on the smartphones. Additionally, the implementation of most of the proposed algorithms and the databases constructed during the course of this thesis are made available to promote reproducible research in biometrics.
Further information can be found at http://nislab.no/news2/kiranthesisdefense