Non-invertible analysis on Graph-based Hamming Embedding Transform for protecting fingerprint minutiae

Abstract

In this paper, we analyze the non-invertible property of Graph-based Hamming Embedding (GHE) for the purpose of fingerprint template protection. The GHE utilizes a set of geometrical invariant features Minutiae Vicinity Decomposition (MVD); GHE globally embeds the MVD neighborhood structure into a Hamming space via solving a graph-based optimization problem. As a result, a secure binary template is created. Pertaining to the non-invertibility, the resultant binary template is analyzed from two aspects: strong analysis and weak analysis. The result of analysis shows that the GHE template can be strongly protected from inversion.