Template Protection for Multimodal Biometric System.
ABSTRACT In order to meet the application requirements regarding universality, permanence and performance, many multimodal biometric approaches have been proposed in recent decades. In these approaches, fused templates are stored unprotected as biometric reference in a token or database. Once the database is compromised, the attacker can potentially learn details that are sensitive to the data subject and obtain sufficient information to impersonate the subject. This paper proposes a template protection approach for multimodal biometric systems, which stores a hash value instead of the fused template. The adaptive non-uniform quantization (ANUQ) method is used to eliminate the gap between the fuzzy character of biometric features and the sensitivity of the hash function. Experiments show that our approach increases the security of the system while the level of biometric performance kept.