Metadata filtering for secure and high performance user identification

Sammanfattning: The aim of this thesis is to investigate the possibility of enabling biometric identification given a large number of enrolled users. This area is relatively unexplored compared to biometric authentication where a fingerprint template is compared against just a few number of templates. The selected approach is to use auxiliary information available during an identification. This information will be used in a filtering process, only retaining probable candidates for further consideration. The auxiliary information as used in this context is known as metadata and some examples includes location, name and age. The lack of existing suitable metadata data sets led to the development of a simulator able to simulate different metadata types. The goal of the simulator is to mimic realistic user behaviour as well as a growing database of enrolled users containing simulated metadata records. The simulator enabled rigorous testing of different filtering techniques in order to evaluate the feasibility of the different techniques. The results show that it is possible to perform a secure, reliable and high performance user identification using a combination of two or more metadata types. The security evaluation also indicates that it is possible to enhance the security of the system by using different strategies protecting against possible attacks. The conducted work resulted in a proof-of-concept implementation supporting further work within the newly explored area.