Multimodal Cryptosystem Based on Fuzzy Commitment

摘要

Biometric information security is a relatively new research interest in the field of biometric authentication. In view of the special relationship of human face and human ear in physiological position, the fusion of face and ear is a natural form of multimodal authentication system used for non-intrusive personal identification. This paper proposes a template protection method for multimodal biometric template. The first step is to apply feature level fusion after feature extraction using principal component analysis. The second step converts the face and ear fused real-valued template to binary template using non-invertible transformation. The third step encrypts the binary template by means of fuzzy commitment. The authentication experimental results on a face and ear multimodal dataset show that multimodal biometric template encryption performs better than the unimodal biometric encryption.