Bioelectrical User Authentication

Abstract

There has been tremendous growth of mobile devices, which includes mobile phones, tablets etc. in recent years. The use of mobile phone is more prevalent due to their increasing functionality and capacity. Most of the mobile phones available now are smart phones and better processing capability hence their deployment for processing large volume of information. The information contained in these smart phones need to be protected against unauthorised persons from getting hold of personal data. To verify a legitimate user before accessing the phone information, the user authentication mechanism should be robust enough to meet present security challenge. The present approach for user authentication is cumbersome and fails to consider the human factor. The point of entry mechanism is intrusive which forces users to authenticate always irrespectively of the time interval. The use of biometric is identified as a more reliable method for implementing a transparent and non-intrusive user authentication. Transparent authentication using biometrics provides the opportunity for more convenient and secure authentication over secret-knowledge or token-based approaches. The ability to apply biometrics in a transparent manner improves the authentication security by providing a reliable way for smart phone user authentication. As such, research is required to investigate new modalities that would easily operate within the constraints of a continuous and transparent authentication system. This thesis explores the use of bioelectrical signals and contextual information for non-intrusive approach for authenticating a user of a mobile device. From fusion of bioelectrical signals and context awareness information, three algorithms where created to discriminate subjects with overall Equal Error Rate (EER of 3.4%, 2.04% and 0.27% respectively. Based vii | P a g e on the analysis from the multi-algorithm implementation, a novel architecture is proposed using a multi-algorithm biometric authentication system for authentication a user of a smart phone. The framework is designed to be continuous, transparent with the application of advanced intelligence to further improve the authentication result. With the proposed framework, it removes the inconvenience of password/passphrase etc. memorability, carrying of token or capturing a biometric sample in an intrusive manner. The framework is evaluated through simulation with the application of a voting scheme. The simulation of the voting scheme using majority voting improved to the performance of the combine algorithm (security level 2) to FRR of 22% and FAR of 0%, the Active algorithm (security level 2) to FRR of 14.33% and FAR of 0% while the Non-active algorithm (security level 3) to FRR of 10.33% and FAR of 0%.