In this paper we present a watermarking based multi-biometric fusion method that can embed fingerprint minutia information into host face images in the DCT (Discrete Cosine Transform) domain. This scheme has the advantage that in addition to prevent unauthorized biometric data manipulations, the biometric authentication can be performed efficiently using the fused biometric data without the need to extract the watermark. Orthogonal Locality Preserving Projections (OLPP) method is used in this approach to extract the most pertinent features which axe beneficial to identification of the watermarked face images. Preliminarily results using ORL and Yale face databases, and FVC2002 DB2 fingerprint database show the effectiveness of the proposed approach in achieving good authentication performance while preventing unauthorized manipulations of biometric data.
展开▼