Improvement of Authentication in Smartphones Using Embedded Pattern-Based Features

Last years, smart devices with touch-sensitive screens are being widely used to access different services. These devices impose the need for storing more personal and sensitive information about the users, such as images, contacts, and videos. To protect this information, deferent techniques are proposed in order to authenticate the users into these devices. In this paper, we propose a method for strengthening the pattern-based authentication, depending on behavioral features extracted from a touch-sensitive screen of smartphones. The extracted features are based on time, movement direction, size and pressure of the users touch on the screen. The proposed technique is based on a deep neural network which classifies the users into a legitimate user or an intruder by comparing the behavioral features with the stored templates. The proposed method is tested on data which are collected from (31) persons. The total number of collected attempts in the dataset is (7,750). The deep neural network has a good performance, compared to other techniques: the random forest classifier and the support vector machine classifier. The empirical authentication results show that the proposed method has an equal error rate of (3.38%).