Multi Biometric Authentication using Score Level Fusion and Non-Invertible Template Formation Technique

Abstract

Single Level Fusion – Score Level or Feature Level is the one in which the Multimodal Biometric Recognition generally can subsists. Feature Extraction, Fusion Formation and Template are the fundamental steps of the Multimodal Biometric Recognition. The matching performance and the accuracy of fusion cannot be increased by bringing single level of fusion technique because of the variation in the matchers. As the fusion process is carried out, the secured template formation is another task in the present system. The hackers should not recognize the stored template as they can easily change the features of the original data. To overcome this, Non-Invertibility Template Formation is carried out in the multimodal biometric recognition. The proposed technique consists of new fusion technique and secured template formation technique such as Matched Performance-Based (MPB) fusion technique and Non-Invertibility Template formation technique.  In order to increase accuracy, both the two levels Feature Level and Score Level are utilized in fusion process of the Matched Performance-Based (MPB). The score obtained through the Feature – Level Fusion is utilized to carry out the performance increment with the modality rate.   In this article, the frequently used biometric traits-Fingerprint and Iris based multibiometric system which utilizes the Feature Level and Score Level is proposed. A novel normalization technique called the Overlap Extrema-Variation-Based Anchored Minmax (OEVBAMM) is also proposed to bring about the fusion.  The feature set with unknown relationship is gained from Iris and Fingerprint and are neither compatible nor homogenous.  Non-invertible transform is utilized to secure the fused pixel information at feature level. The fused image’s textural information belongs to Fingerprint and Iris is employed to build a unique feature vector.  Unlike the passwords and tokens, the compromised biometric templates are neither be revoked nor be reissued. Hence the principal focus of our work lie on Biometric Template Security.  The possibility of getting variation in the acquired biometric traits of a user makes ensuring security very challengeable in maintaining the recognition performance. The proposed work includes the biometric fusion with the Non-Invertible Transform for template security purpose.