Feasibility of Morphing-Attacks in Vascular Biometrics

Abstract

For the first time, the feasibility of creating morphed samples for attacking vascular biometrics is investigated, in particular finger vein recognition schemes are addressed. A conducted vulnerability analysis reveals that (i) the extent of vulnerability, (ii) the type of most vulnerable recognition scheme, and (iii) the preferred way to determine the best morph sample for a given target sample depends on the employed sensor. Digital morphs represent a significant threat as vulnerability in terms of IAPMR is often found to be > 0.8 or > 0.6 (in sensor dependent manner). Physical artefacts created from these morphs lead to clearly lower vulnerability (with IAPMR ≤ 0.25), however, this has to be attributed to the low quality of the artefacts (and is expected be increase for better artefact quality).