Dual-Minimax Probability Machines for One-class Mobile Active Authentication

Active Authentication(AA) systems operating on mobile devices are expected to continuously authenticate the enrolled user based on enrolled sensor observations. Due to unavailability of training samples from negative classes, AA can be viewed as a one-class classification problem. In this work we introduce a Single-class Minimax Probability Machine(1-MPM) based solution called Dual Minimax Probability Machines(DMPM) for AA applications. In contrast to 1-MPM, proposed method has two notable differences. 1) We learn an additional hyper-plane to separate training data from the origin by taking into account maximum data covariance. 2) We consider the possibility of modeling the underline distribution of training data as a collection of sub-distributions. Intersection of negative half spaces defined by the two learned hyper-planes is considered to be the negative space during inference. We demonstrate the effectiveness of the proposed mechanism by performing evaluations on three publicly available face based AA datasets.