Feature Fusion Methods for Indexing and Retrieval of Biometric Data: Application to Face Recognition With Privacy Protection

Computationally efficient, accurate, and privacy-preserving data storage retrieval are among the key challenges faced by practical deployments of biometric identification systems worldwide. In this work, a method protected indexing is presented. By utilising feature-level fusion intelligently paired templates, multi-stage search structure created. During retrieval, list potential candidate identities successively pre-filtered, thereby reducing number template comparisons necessary for transaction. Protection probe as well stored reference templates created index carried out using homomorphic encryption. The proposed extensively evaluated in closed-set open-set scenarios on publicly available databases two state-of-the-art open-source face recognition systems. With respect to typical baseline algorithm an exhaustive search-based algorithm, enables reduction computational workload associated with transaction 90%, while simultaneously suffering no degradation performance. Furthermore, facilitating seamless integration protection encryption libraries, guarantees unlinkability, irreversibility, renewability data.