Analysis of authentication system based on keystroke dynamics

Abstract

Currently, almost any kind of data is stored on the Internet, which makes security issues of the private data very crucial. The main objective of this paper is to analyze multimodal authentication security systems that are combined with keystroke dynamics (KSD). The keystroke dynamics are one of the behavioral biometrics which describes the typing rhythm of the individual. The proposed work aimed to design a multimodal authentication system using a machine learning/deep learning algorithm for user classification based on traditional and touchscreen features of keystroke dynamics. The results have shown that XGBoost has a higher overall classification accuracy (90.91%) in comparison with the previous works (85.90%) on the same dataset. Confusion Matrix and Receiver Operating Characteristics (ROC) curves were derived for further performance evaluation of the system. Area Under the Curve (AUC) values for each of the classes and for the whole classifier were also reported.