Robust Combination Method for Privacy Protection Using Fingerprint and Face Biometrics

Secure advance system for fingerprint privacy protection by combining different biometrics fingerprint and face into a new identity is proposed. In an enrollment, one fingerprint and face images are captured from same person. Then the minutiae positions and orientation from fingerprint and the reference points from both biometrics are extracted. LDN extracts directional information from face. To compute LDN features from face, face image is divided into some parts. LDN features allocation is taken out from face-parts. Then concatenate these features into feature vector, and use it as a face descriptor. Based on this extracted information and proposed coding strategies, combined template is generated and then stored in a database. In the verification, the system requires two queries; one fingerprint and one face from the same person. The two-step fingerprint matching algorithm is used for matching the fingerprint of same person against the generated combined minutiae template. For the face, chi-square dissimilarity measure is used for matching feature vectors of the person which are compared with all feature vectors of persons present in dataset. Fingerprint-face reconstruction approach is used to create combined fingerprint-face image from combined template. Hence, a virtual identity is nothing but the reconstructed image created from the two biometrics one fingerprint and one face and is used for matching purpose. FRR and FAR of the proposed system is low and is 1% each. Work proposed can create better identity when fingerprint-face images are randomly taken.