Secure fuzzy fingerprint vault against correlation attack

Abstract

Abstract:The fuzzy vault scheme has emerged as a promising solution to the user privacy and the fingerprint template security problems. Recently, however, this scheme is shown to be susceptible to a correlation attack that finds the real minutiae using multiple vaults enrolled for different applications. To protect the fuzzy fingerprint vault from the correlation attack, we propose an approach to add chaffs in a more structured way such that distinguishing the fingerprint minutiae and the chaff points obtained from two applications is computationally hard. Experimental results show that the proposed approach achieves much secure performance than adding chaffs randomly without a significant degradation of the verification accuracy.

A dissection of fingerprint fuzzy vault schemes

“…Most fingerprint fuzzy vault implementations adopt this representation of minutiae as stand-alone points, but they differ according to what attributes they use. Examples of implementations that employ: (i) only (x, y) coordinates of each minutia in Cartesian or polar format, include: [3,4,15,16,25,27,29,30]; (ii) {(x, y), θ} of each minutia, include: [5,13,[20][21][22]; (iii) {(x, y), θ, t} of each minutia, include: [2,11,12,17,18,23,26].…”

“…This approach was adopted by [5,20,22], and [2,17,18,26] incorporated the type attribute as well; however, most of these implementations still employed around 200 chaff points. Two similar and interesting propositions to enable the addition of a larger number of chaff points were outlined in [12] and [11]. In [12], the authors use {x, y, z, θ, t} for each minutia point and thus each chaff point, where z = θ/α (α was set to 18).…”

“…They were able to generate 1,000 chaff points using this representation. In [11], a 3D lattice is first generated, where the x-and y-dimensions of the lattice correspond to the xand y-dimensions of the fingerprint image, and the z-dimension corresponds to the minutiae orientations, θ, quantized into three sections of 120° each. The genuine minutiae points are placed into the lattice first and any remaining empty cells are filled by randomly generated chaff points.…”

“…Most publications in this area are related to the geometric hash table proposed in [2], where each point in the fuzzy vault is expressed relative to every other point (minutiae and chaff points). This technique, or variants of it, have been adopted in [11,12,18,26], mostly by the authors of [2]. Other translation and rotation invariant features proposed for alignmentfree fingerprint fuzzy vaults are generally based on nearestneighbour minutiae structures, and examples include: [7,10,14].…”

Smartcard-based Secret Distribution for Secure Fingerprint Verification

“…Furthermore, the polynomial can be reconstructed only when the information stored in both the smartcard and the server is available. In order to enhance the security level of the fuzzy fingerprint vault against the correlation attack, we apply the structured chaff insertion approach [18] additionally. That is, by inserting chaff points into a fingerprint template structurally in order to make multiple systems look similar, an attacker cannot get any hint even from two vaults.…”