Sparse Reconstruction Based Watermarking for Secure Biometric Authentication

This paper presents a robust watermarking method to enhance the security of multimodal biometric authentication system. A compact representation of face is first generated by downsampling the raw image with a high ratio. Then, the face feature is employed as watermark and embedded into fingerprint image with a blind SS-QIM scheme. Under the framework of sparse reconstruction, the credibility of fingerprint data can be verified by checking the validity of extracted pattern in the authentication stage. Furthermore, if the extracted face watermark is valid, it can provide additional identity information for multimodal biometric authentication. Experimental results demonstrate that the face watermark can effectively verify the credibility of fingerprint images while not affecting their recognition performance. Plus, fusing valid face watermark with host fingerprint can also increase the reliability of authentication.