A Survey on Unknown Presentation Attack Detection for Fingerprint

Abstract

Fingerprint recognition systems are widely deployed in various reallife applications as they have achieved high accuracy. The widely used applications include border control, automated teller machine (ATM), and attendance monitoring systems. However, these critical systems are prone to spoofing attacks (a.k.a presentation attacks (PA)). PA for fingerprint can be performed by presenting gummy fingers made from different materials such as silicone, gelatine, play-doh, ecoflex, 2D printed paper, 3D printed material, or latex. Biometrics Researchers have developed Presentation Attack Detection (PAD) methods as a countermeasure to PA. PAD is usually done by training a machine learning classifier for known attacks for a given dataset, and they achieve high accuracy in this task. However, generalizing to unknown attacks is an essential problem from applicability to real-world systems, mainly because attacks cannot be exhaustively listed in advance. In this survey paper, we present a comprehensive survey on existing PAD algorithms for fingerprint recognition systems, specifically from the standpoint of detecting unknown PAD. We categorize PAD algorithms, point out their advantages/disadvantages, and future directions for this area.

Biometrics: Trust, but Verify

“…We note that this paper is unique in that it aggregates and examines the main components of a trustworthy biometric recognition system into one manuscript. Indeed many surveys [13,[35][36][37][38][39][40][41][42][43] have been written in great detail on each one of these topics individually, however we posit that there is benefit in extracting the key points from each of these areas and summarizing them in one place such that researchers can very quickly and easily assess the current state of trustworthy biometric recognition systems. Furthermore, many of the existing survey papers on these individual topics have become outdated.…”

“…The continued success in spoofing modern day biometric recognition systems is not a consequence of a lack of research into developing presentation attack detection (PAD) systems. Indeed the past couple of decades have seen a plethora of research into developing PAD systems which can automatically detect and flag a spoof attack prior to performing authentication or identification [35,36,[98][99][100]. Typically these approaches are divided into hardware or software based approaches detecting face, fingerprint, and iris spoofs.…”

“…to capture features which differentiate bonafide acquisitions from PAs [101][102][103][104][105][106][107][108][109][110][111][112]. In contrast, software based solutions extract anatomical, physiological, textural, challenge response, or deep network based features to classify an input sample as live (bonafide) or presentation attack (spoof) [35,36,96,[98][99][100][113][114][115][116][117][118][119][120][121][122]. The culmination of these approaches can be seen in the high performances of the various algorithms submitted as part of the IARPA ODIN program 11 and also the public fingerprint and face liveness competitions (Table 2) [95,97].…”

On the Generalisation Capabilities of Fingerprint Presentation Attack Detection Methods in the Short Wave Infrared Domain

“…A detailed overview of hardware-based fingerprint PAD is presented in [19], hence we focus on unknown attacks [20] and generalisation approaches in this work. A summary of these works is given in Table 1.…”

Anomaly Detection with Convolutional Autoencoders for Fingerprint Presentation Attack Detection

“…Being a binary classification problem (bona fide vs. PA), common PAD approaches are trained on both classes and hence perform only as good as the chosen training set. In this scenario, unknown attacks [7] present only in the test set can significantly trouble the classifier, as it requires good generalisation properties that are hard to achieve. In order to avoid re-training the classifier each time a new PAI species is created, one-class classifiers can be used [8].…”