Mobile Based Continuous Authentication Using Deep Features

Abstract

Continuous authentication is a promising approach to validate the user's identity during a work session, e.g., for mobile banking applications. Recently, it has been demonstrated that changes in the motion patterns of the user may help to note the unauthorised use of mobile devices. Several approaches have been proposed in this area but with relatively weak performance results. In this work, we propose an approach which uses a Siamese convolutional neural network to learn the signatures of the motion patterns from users and achieve a competitive verification accuracy up to 97.8%. We also find our algorithm is not very sensitive to sampling frequency and the length of the sequence.