A non-invertible Randomized Graph-based Hamming Embedding for generating cancelable fingerprint template

Abstract

Biometric technology is likely to provide a new level of security to various applications. Yet if the stored biometric template is compromised, invasion of user privacy is inevitable. Since biometric is irreplaceable and irrevocable, such an invasion implies a permanent loss of identity. In this paper, a fingerprint template protection technique is proposed to secure the fingerprint minutiae. Remarkably, by incorporating Randomized Graph-based Hamming Embedding (RGHE), the generated binary template can be strongly protected against inversion. The proposed method adopts a minutiae descriptor, dubbed as minutiae vicinity decomposition (MVD) to derive a set of randomized geometrical invariant features together with random projection. The discrimination of randomized MVD is then enhanced by User-specific Minutia Vicinities Collection scheme and embedded into a Hamming space by means of Graph-based Hamming Embedding. The resultant binary template enjoys four merits: (1) strong concealment of the minutia vicinity, thus effectively protects the location and orientation of minutiae. (2) Well preservation of the discriminability of MVD in the Hamming space with respect to the Euclidean space without accuracy performance degradation. (3) Template is revocable due to user-specific random projection. (4) Speedy matching attributed to bit-wise operations. Promising experimental results on FVC2002 database vindicate the feasibility of the proposed technique.