Pattern password authentication based on touching location

Abstract

Pattern passwords are one of the embedded authentication method of touchscreen devices, however it has some major drawbacks which briefly are identifiability and imitability. The password of the user is noticeable when entering the pattern due to shining circles. Therefore, what we put forward in this paper is a novel biometric implementation of a hidden system to pattern password authentication for increasing password security. As opposed to general research concept which extracts touch or keystroke durations, we focused on the touching coordinates calculated the distance of the line between the constant pattern node and the touched place as well as the angle. Using these inputs, we trained the neural network by Gauss-Newton and Levenberg-Marquardt algorithms and conducted the experiments with these trained classifiers.