A Secure and Reliable Mobile Authentication Alternative Utilizing Hand Structure

As technology advances, methods to protect information on personal devices are shifting towards forms of biometric user authentication. When assessing these new modes of authentication, it is pertinent to consider their implementation of security, reliability, privacy, and convenience. Current facial and fingerprint identification technologies compromise privacy since they require the collection of personal data. On the other hand, non-biometric mobile authentication lacks security due to a preference for convenient and reliable techniques such as short passwords. Hand structure-based authentication is a solution that incorporates all of these criteria. With privacy and convenience inherent to this mode of authentication, the proposed experimental procedure leverages hand biometrics and the K-Nearest Neighbors, Support Vector Machines, and Bagged Decision Trees machine learning algorithms to examine the security-reliability trade-off. The results show that there are statistically significant properties of structure-borne signals that enable differentiation between individuals and hand positions. These properties, along with manipulation of algorithmic learning parameters, are used to enable the model to specialize either in security or reliability. While all of the developed models excelled in security, the Bagged Decision Trees model outperformed the others by rejecting false users at a peak accuracy of 99.9% and rejecting the wrong hand positions with a peak accuracy of 99.4%.