Keystroke-Based Biometric Authentication in Mobile Applications

Abstract

Keystroke dynamics is the study of how people can be distinguished based on their typing rhythms. It is considered as a real alternative to several authentication mechanisms. In order to improve authentication security in mobile applications, especially mobile chat applications , we propose in this work an approach of user keystroke-based authentication which is based on two important mechanisms, firstly the authentication is done on the basis of a supervised classification model trained on the dynamic keystroke database (android platform) the prediction accuracy has reached high values. When the attacker used the same behavior as the legitimate user we are strengthening security by another authentication mechanism based on three parameters during the opening of a communication session between two users (the number of messages per session, connection time and inter-click time). The proposed system has shown satisfactory results by combining user behaviours and connection parameters during an open session between two users to reinforce the authentication aspect.