Abstract
AbstractBiometric system databases are vulnerable to many types of attacks. To address this issue, several biometric template protection systems have been proposed to protect biometric data against unauthorized use. Many of biometric protection systems require the biometric templates to be represented in a binary form. Therefore, extracting binary templates from real-valued biometric data is a key step in such biometric data protection systems. In addition, binary representation of biometric data can speed-up the matching process and reduce the storage capacity required to store the enrolled templates. The main challenge of existing biometric data binarization approaches is to retain the discrimination power of the original real-valued templates after binarization. In this paper, we propose a secure and efficient biometric data binarization scheme that employs multi-objective optimization using Nondominated Sorting Genetic Algorithm (NSGA-II). The goal of the proposed method is to find optimal quantization ...
Highlights
The growing need of biometrics in access control and verification applications makes security of biometric data a pressing and important issue
Several biometric template protection schemes, such fuzzy commitment 3 and BioEncoding 4 schemes, require the input biometric data to be in a binary form
The recognition accuracy is degraded. In view of this trade off between performance and security, we propose a new unsupervised binarization method based on an optimization strategy to search for the optimum quantization levels and encoding functions for each feature dimension to achieve balance between the security and the recognition accuracy for the biometric system
Summary
The growing need of biometrics in access control and verification applications makes security of biometric data a pressing and important issue. Biometric template protection schemes [1,2] have been developed to ensure biometrics privacy and security. The idea behind these schemes is to store an encoded version for the biometric template (the distinct traits extracted from biometric data) rather than the original one by applying a transformation function. As a consequence of this limitation, direct application of these schemes is restricted to binary-valued biometric data such as iris-codes 5. Real-valued templates are extracted from the raw biometrics through feature extraction stage. In order to produce the same binary string for a user in the verification stage, the binarization parameters (quantization and encoding) are stored as helper data in the biometric system
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
