1 Abstract Minutia vicinity representation was recently proposed by Yang & Busch to generate a protected fingerprint template scheme [14], the resultant protected template enjoys good accuracy and free from alignment. However, Yang and Busche’s scheme is highly likely reversible [18]. This paper proposed a new minutiae representation technique known as Minutia Vicinity Decomposition (MVD) whereby each minutia vicinity is decomposed into four minutia triplets. A set of geometrical invariant features can be extracted from the minutia triplet to construct a fingerprint template. The invariant features with random offsets salting mechanism enhance the reversibility, revocability as well as performance accuracy of the resultant protected fingerprint template. Promising experimental results on FVC2002 DB2 justify the feasibility of our proposed technique. 1. Introduction The security and privacy of the biometric template are gaining more attention in consequence of the emergent of numerous biometric applications. One of the major concerns in biometric template protection is the revelation of the user’s privacy due to the strong binding between biometric template and the user’s identity. For instance, personal particulars could be tracked through biometric templates from one application to another by cross-matching between biometric databases. This is further complicated by the fact that biometrics cannot be reproduced or replaced when compromised. Due to these concerns, a biometric system with protected template is required immediate attention. Ideally, template protection schemes must fulfill the following requirements [1, 2]: (a) Diversity. A protected template must not allow cross matching across different applications, thereby ensuring user’s privacy. (b) Revocability. A new template can be reissued once the old protected template is compromised. (c) Non-invertibility. It must be impossible or computationally hard to obtain the original biometric template from the protected instance and helper data. (d) Performance. Satisfactory protected system performance accuracy in terms of False Rejection Rate (FRR) or False Acceptance Rate (FAR). Among various biometric template protection schemes, fingerprint minutiae protection is of great interest due to its wide acceptance and usage in commercial fingerprint recognition systems. In literature, many minutiae-based fingerprint template protection schemes have been proposed. The reported schemes can be broadly divided into two categories, namely alignment based or alignment-free based approach. For the alignment-based approach, a registration point (core or delta) is required to align the fingerprint image before further processing. A well-known instance was reported by Ratha et al. [3] wherein the fingerprint minutiae data is transformed by a sequence of three non-invertible transforms functions: Cartesian, Polar and surface folding. Although the three transformation functions were claimed to be non-invertible due to the many-to-one mapping property, a scheme proposed by Feng et al. [4] reveals that Ratha’s surface folding transforms is possibly degenerated when the transformed template and parameters are known to the attacker. “Fuzzy vault”, another popular instance, was proposed by Juels and Sudan [5], wherein secret can be regenerated only if both enrolled and query minutia data are substantially overlapped. Many improved versions of fuzzy vault have been proposed, such as [6, 7, 8]. Ang et al. [9] proposed a key-dependent transformation method wherein a line through the core point is first specified and the minutiae above the line are reflected symmetrically below the line. Different template can be obtained by changing the line orientation. Nagar et al. [10] presented a richer set of fingerprint features which consist of minutiae features, ridge orientation features and ridge wavelength features. This feature set is finally binarized into bit-string to represent user template. In contrast to the alignment approach, no registration point is needed in the alignment-free approach. For example, Farooq et al. [13] presented an instance of a binary fingerprint representation. Their idea is based on the fact that fingerprints can be represented by a set of triangles derived from sets of three minutiae. Seven invariant features, the length of three sides, the three angles between the sides and minutiae orientations and the height of the triangles are deliberately extracted and further quantized Fingerprint Template Protection with Minutia Vicinity Decomposition Jin Zhe 1 Andrew Teoh Beng Jin 2 1 Faculty of Information Science & Technology, 2 School of Electrical & Electronic Engineering, Multimedia University, Malaysia Yonsei University, South Korea jin.zhe@mmu.edu.my bjteoh@yonsei.ac.kr 978-1-4577-1359-0/11/$26.00 ©2011 IEEE