Secure and Efficient Biometric Systems

Hierarchical Bloom Filter Framework for Security, Space-efficiency, and Rapid Query Handling in Biometric Systems

Published in EEE 10th International Conference on Biometrics Theory, Applications and Systems (BTAS)

Overview: Incorporating biometric technology into IoT could enable and assist a variety of futuristic applications such as cardless border crossing, secure access control in smart buildings, and patient tracking/anti-fraud in hospitals. However, existing approaches suffer from large storage requirements and latency in query handling. Additionally, there are privacy risks due to security breaches of raw biometric templates. In this paper, we propose a hierarchical Bloom filter (HBF) based identification system for large-scale biometric systems that reduces storage requirements while providing template security and rapid handling of queries. We address the challenge of incorporating a hash-based bloom filter with noisy biometric data by introducing a mathematical framework that is adaptive to characteristics of any biometric database. Our proposed architecture is implemented using a face database containing 30,000 facial templates and achieves 92.05% reduction in storage size with 99.82 reduction in average query time without sacrificing accuracy

sss

Fig: High level diagram of the proposed method: (a) Enrolling templates into HBF database and (b) Query handling

As shown in the figure, the HBF operates in two stages. In the enrollment stage, the proposed system begins functioning by receiving input face images from users, preprocessing them, generating binary templates, and finally enrolling the users into the HBF with their corresponding binary templates. In the query stage, the query binary template is generated for the query person (following the same procedure at enrollment stage), binary template is provided to the HBF, and finally the HBF determines the no. of total BFs authenticating the query template. If a certain number of BFs can authenticate it, the query person is considered as "authenticated".