An Uncertainty-based Approach to Improve Accuracy of Biometric Metrics

Biometrics is used today for getting access to several services or facilities by using even mobile devices and systems for identification. Access to governmental offices and to high security buildings is more and more recurring to such biometric authentication solutions. Fingerprints, face authentication, iris or retinal vascular pattern recognition are some of the most common applications of this field. Literature is full of studies and research activities proposing further innovative or revised biometric metrics. However, accuracy and security are still two relevant issues affecting most of these applications. As a consequence, biometrics is still much too unreliable for everyday use and for accessing to high security services or places. The main issue, here faced by authors, concerns the accuracy of measurement and acquisition devices used for authentication. Uncertainty that affects measurements has a relevant effect on the authentication outcome. As a consequence, it could affect the final result so being cause of high false reject and false accept rates. In the paper, the authors propose an uncertainty-based approach to increase the accuracy of the current biometric metrics. The type A uncertainty evaluation offers an interesting solution to assess the measurement precision of the biometric system recognition. So uncertainty is used to provide quantitative information on the accuracy of acquired data. Experimentation has been carried out on biometric recognition based on retinal vascular patterns. Experimental results show an increase in sensitivity and accuracy of the biometric recognition metric so proving the relevance of the proposed approach.