On the implementation of modified fuzzy vault for biometric encryption

Abstract

Abstract-Biometrics, such as irises and fingerprints, enable secure and non-repudiable authentication. Fuzzy vault is a scheme that can monolithically bind secret to biometric templates. Moreover, the modified fuzzy vault (MFV) leads to less entropy loss and requires less memory for storing the sketches. This paper proposes a novel low-complexity scheme to compute the monic polynomial for the sketch during the enrollment process of the MFV. An innovative interpolation method is also developed to reduce the computation complexity and latency of the verification process. Efficient hardware implementation architectures are developed in this paper for the proposed schemes and their complexities are analyzed in detail.

Reducing the latency of Lee-O'Sullivan interpolation through modified initialization

“…In the modified fuzzy vault scheme [2], a monic polynomial that passes each interpolation point is stored in the database. It was proposed in [9] to replace the Lagrange polynomial in (1) by this monic polynomial. Fig.…”

“…A basis conversion architecture can be found in [9] and is not re-drawn in this paper. For t = 2, it has nine memory blocks storing the coefficients of q l,i (x) (0 ≤ l, i < 3).…”

“…In the case of lowrate codes, some existing architectures no longer apply and new issues need to be addressed. The first implementation architecture for the modified fuzzy vault [2] was proposed in [9]. In this design, the Lee-O'Sullivan (LO) interpolation algorithm [10] is employed.…”

“…The initialization originally uses the Lagrange polynomial that passes each point. In the modified fuzzy vault scheme, a monic polynomial that passes each point is derived in the enrollment process, and can replace the Lagrange polynomial in the initial basis [9]. However, the conversion process has long latency using either of these initializations.…”

“…This paper proposes a novel initialization of the polynomial basis in the LO interpolation for low-rate codes. Using the original basis initialization in [10] and that in [9], the xdegrees of the polynomials can be raised significantly during the basis conversion process for low-rate codes. The raised xdegree leads to increase in not only the number of clock cycles in each iteration, but also the number of iterations required in the basis conversion.…”