User-specific effects in Fingerprint Presentation Attacks Detection: Insights for future research

Abstract

A fingerprint presentation attacks detector (FPAD) is designed to obtain a certain performance regardless of the targeted user population. However, two recent works on facial traits showed that a PAD system can exploit very useful information from the targeted user population. In this paper, we explored the existence of that kind of information in fingerprints when textural features are adopted. We show by experiments that such features embed not only intrinsic differences of the given fingerprint replica with respect to a generic live fingerprint, but also contains characteristics present in other fingers of the same user, and characteristics extracted directly from spoofs of the targeted fingerprint itself. These interesting evidences could lead to novel developments in the design of future FPADs

Analysis of “User-Specific Effect” and Impact of Operator Skills on Fingerprint PAD Systems

“…Moreover, the above points were not addressed by the previous editions of LivDet, whose history and results over the years are reported in [3]. With regard to the "user-specific effect", we reported in [5] that some specific components or minute details are dependent on the user's skin characteristic. Ref.…”

“…In last years it was noticed that the presence of the same users both in the train set and in the test set, increased the PAD accuracy of the system. In fact, the existence of artifacts due to the human skin (person-specific) and to the particular curvature of ridges and valleys (finger-specific) can impact in PAD systems' performance [5,6] and can be exploit to improve the integrated system. In particular, the data acquired during the Green Bit 1000 400 400 400 1700 680 680 680 Orcanthus 1000 400 400 400 1700 680 658 680 Digital Persona 999 400 400 399 1700 679 670 679 Table 2: Number of samples for each scanner and each part of the dataset.…”

“…; • when a PAD must be integrated into a fingerprint verification system, the templates of genuine users are available, and there is no good reason to neglect them even during the PAD design. We referred to the additional information coming from the enrolled user as "user-specific effect" [5].…”

“…With regard to the "user-specific effect", we reported in [5] that some specific components or minute details are dependent on the user's skin characteristic. Ref.…”

“…Ref. [5] reports a statistical study on the relationships of textural features coming from the same fingerprints of the same user and even from his/her different fingerprints. The correlation we measured suggests that the consequent "bias" introduced into a PAD can be beneficial if the PAD must be integrated into a fingerprint verification system.…”