Mobile User Identification through Authentication Using Keystroke Dynamics and Accelerometer Biometrics

Abstract

Biometrics is everything that can be measured in a human being. It has two types; behavioral and physiological. This paper discusses the use of keystroke dynamics, a form of behavioral biometrics that deals with the measure of how a person types, and the utilization of accelerometer biometrics as a form of behavioral biometric that measures how a person holds his mobile device. We collected biometric data from 30 volunteer participants by asking them to enter their 8-16-character password specimens 8 times using a customized tool in a mobile phone. The first 6 collection from each participant was set aside for the training set while the other 2 is for the test set. The data were then processed and extracted keystroke dynamic and accelerometer biometrics using a customized tool written in Java. Several well-known classifiers were trained using keystroke dynamic features alone, accelerometer biometrics alone, and the combination of both. Results show that Neural Network classifier using the combined features gave the most acceptable model. The model performance was further improved by removing some low ranking features defined by the Chi Square attribute evaluator and by removing some features that are highly correlated to other features.