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

Abstract

Fingerprint Liveness detection, or presentation attacks detection (PAD), that is, the ability of detecting if a fingerprint submitted to an electronic capture device is authentic or made up of some artificial materials, boosted the attention of the scientific community and recently machine learning approaches based on deep networks opened novel scenarios. A significant step ahead was due thanks to the public availability of large sets of data; in particular, the ones released during the International Fingerprint Liveness Detection Competition (LivDet). Among others, the fifth edition carried on in 2017, challenged the participants in two more challenges which were not detailed in the official report. In this paper, we want to extend that report by focusing on them: the first one was aimed at exploring the case in which the PAD is integrated into a fingerprint verification systems, where templates of users are available too and the designer is not constrained to refer only to a generic users population for the PAD settings. The second one faces with the exploitation ability of attackers of the provided fakes, and how this ability impacts on the final performance. These two challenges together may set at which extent the fingerprint presentation attacks are an actual threat and how to exploit additional information to make the PAD more effective.Keywords Liveness Detection · User-specific effect · skilled operator.

LivDet 2021 Fingerprint Liveness Detection Competition - Into the unknown

“…In fact, the client task does not contain never-seen-before materials. On the other hand, in the material task, training and test sets share the same users, and the user-specific effect allows higher accuracy than in previous experiments and confirms what reported in [10]. Tab.…”

“…• Challenge 1, Liveness Detection in Action [9]: similarly to the previous edition, competitors were invited to submit a complete algorithm able to output both the probability of the image vitality (the so-called "liveness score") given the extracted set of features and an integrated match score ("integrated score") which includes the probability above with the probability of belonging to the claimed user. For this challenge, participants can decide whether to exploit the related "user-specific" information [10]). • Challenge 2, Fingerprint representation: In modern biometric systems, the feature vectors compactness and discriminability are fundamental to guarantee high performance in terms of accuracy and speed, especially for systems embedded in mobile devices.…”

LivDet2023 - Fingerprint Liveness Detection Competition: Advancing Generalization

“…• Challenge 1, Liveness Detection in Action [5]: Competitors were asked to submit a complete algorithm capable of producing both the " score", that is, the probability of being a bona fide sample and the "integrated score", which combines the previous score with the probability of belonging to the claimed user. Participants in this challenge can choose whether to use the related "user-specific" information [19]. • Challenge 2, Fingerprint representation: Compactness and discriminability of feature vectors are critical in modern authentication systems to ensure high performance in terms of accuracy and speed.…”

Review of the Fingerprint Liveness Detection (LivDet) competition series: from 2009 to 2021

“…The 2017 [20] edition of Livdet was focused on determining how much the data composition and collection techniques influence the PAD systems. Although the aim of the competition was clear from the beginning, the following aspects have been analyzed in a later work [23] and only marginally in the LivDet 2017 report for the sake of space: (1) the presence of operators with different skills in fabricating the replicas, one with a considerable degree of expertise (high skilled) and one composed of novice forgers (low skilled); (2) the presence of some users in both the training and testing parts of the three datasets. As a matter of fact, in previous editions, the train and test sets were completely separated as none of the users were present in both.…”