An Efficient Optimized Mouse and Keystroke Dynamics Framework for Continuous Non-Intrusive User Authentication

Abstract

In recent years, user authentication based on mouse and keystroke dynamic is the most wanted topic to identify the external user and to secure information. Additionally based on the movement of the mouse and typing speed of keystroke the correct user was identified. But the problems of existing approaches are complicated data, data error, and malicious events. To overcome these threats, a novel cat recurrent neural model (CRNM) is proposed to identify the correct user and improve the accuracy rate. In this work, the CRNM approach is introduced to minimize the error rate, to detect unauthorized users by analyzing the user's mouse and keystroke dynamic. Consequently, the trained datasets verify the inputs and identify the correct user. Thus the proposed CRNM has been implemented in the python framework, to identify the correct user. Moreover, the proposed model is validated with other existing deep learning models in terms of accuracy, false acceptance rate (FAR), F-measure, recall, false negative rate (FNR), precision, and error rate.