Master of Engineering Science
Electrical and Computer Engineering
Dr. Miriam A.M. Capretz
Biometric based authentication is one of the most popular techniques adopted in large-scale identity matching systems due to its robustness in access control. In recent years, the number of enrolments has increased significantly posing serious issues towards the performance and scalability of these systems. In addition, the use of multiple modalities (such as face, iris and fingerprint) is further increasing the issues related to scalability. This research work focuses on the development of a new Hybrid Data Storage Framework (HDSF) that would improve scalability and performance of biometric authentication systems (BAS). In this framework, the scalability issue is addressed by integrating relational database and NoSQL data store, which combines the strengths of both. The proposed framework improves the performance of BAS in three areas (i) by proposing a new biographic match score based key filtering process, to identify any duplicate records in the storage (de-duplication search); (ii) by proposing a multi-modal biometric index based key filtering process for identification and de-duplication search operations; (iii) by adopting parallel biometric matching approach for identification, enrolment and verification processes. The efficacy of the proposed framework is compared with that of the traditional BAS and on several values of False Rejection Rate (FRR). Using our dataset and algorithms it is observed that when compared to traditional BAS, the HDSF is able to show an overall efficiency improvement of more than 54% for zero FRR and above 60% for FRR values between 1-3.5% during identification search operations.
Tiwari, Abhinav, "Hybrid Data Storage Framework for the Biometrics Domain" (2014). Electronic Thesis and Dissertation Repository. 1864.