Abstract

User authentication security is becoming more and more interesting and is an important area of research to address privacy concerns and security threats. Biometric has revolutionized the authentication technology and can be used to identify and authenticate users given their physiological and behavioral traits. Touchscreens have become the leading input medium, and hence touch-dynamics is becoming passive and unobtrusive form of biometric authentication. Though a numerous studies have been done in behavioral biometrics, touch dynamics behavioral biometric still needs to be explored. In our work, we proposed deep learning algorithms based on artificial neural network for user authentication and used touch data set. A customized Convolutional Neural Network (CNN) and Bi-Long Short Term Memory (LSTM) neural network is built for user authentication. By utilizing the Bioident database, our methodology achieved an overall accuracy of 98.7%. After being validated with more varied touch data, the findings of our proposed model demonstrate that it performs superior to current methods and can be employed in real-time authentication mechanisms.