Page - 65 - in Joint Austrian Computer Vision and Robotics Workshop 2020

PresentationAttacksandDetection inFinger-andHand-VeinRecognition LucaDebiasi,ChristofKauba,HeinzHofbauer,BernhardPrommegger,AndreasUhl DepartmentofComputerSciences,UniversityofSalzburg {ldebiasi,ckauba,hhofbaue,bprommeg,uhl}@cs.sbg.ac.at Abstract. Biometric recognition systems, especially vascular pattern based ones, are becoming more popular. However, these systems are still susceptible tosocalledpresentationattacks,wherea forgedrep- resentation of the original biometric is presented to thesystemtryingtomimic theoriginalbiometricand fool the system. We propose a presentation attack approach for finger- and hand-vein recognition sys- tems using paper prints as well as wax and silicone artefacts. We further develop a suitable presenta- tionattackdetection (PAD)schemebasedonnatural scene statistics and acquire a corresponding hand veinpresentationattackdataset. EvaluatingthePAD schemeonthedatasetconfirmeditssuccessinthede- tection of the forgedsamples. 1. Introduction Inourmodernworld there isanevergrowingneed forpersonalauthentication. Biometricauthentication systems are one way to overcome the typical prob- lems of classical authentication methods, e.g. dis- closed or forgotten passwords, lost or stolen keys and forged signatures. Biometric authentication sys- tems are based on so called biometric traits, which areuniquebehaviouralorphysiological characterist- ics of a person. These are inherently linked to a per- sonandcannotget lost,beforgottenorbestolen. The most prominent examples of biometric traits include fingerprints, face and iris. Recently, vascular pattern based biometrics (usually denoted as vein recogni- tionbasedsystems)gainmoreattentionaswell,with finger- and hand-vein based systems being the most widely used ones [27]. Vein based systems exhibit some advantages over other biometric systems, e.g. fingerprint and face recognition ones. They rely on the structure of the vascular pattern formed by the blood vessels inside the human body tissue, i.e. it is aninternalbiometric trait. Thispatternonlybecomes visible in near-infrared (NIR) light, as the haemo- globin in the blood absorbs NIR light, rendering the blood vessels (veins) visible as dark lines in the cap- tured images. Vein based systems are more resistant toforgeryandtheyareneithersusceptibletoabrasion nor skin surfaceconditions [11]. Despite the advantages mentioned above, biomet- ric recognition systems are far from being perfect. Almost all of the currently employed systems are susceptible tospoofingorpresentationattacks (PAs). APAisdefinedaspresentation to thebiometricdata capture subsystem with the goal of interfering with the operation of the biometric system according to the ISO/IEC 30107-1 standard [4]. This corresponds to the creation of a forged representation mimick- ing the original biometric trait (also called a spoof- ing artefact) that is used to spoof/fool the biometric system. PAsareposingasevereprobleminpractical applications as a genuine user may be impersonated. By launching a successful PA, an adversary is able to gain illegitimate access to the system. In con- trast to passwords and tokens, a biometric trait can neither be replaced nor revoked. Hence, if a system isprone toPAs, it canno longerbeconsideredas se- cure. Fortunately, there are counter-measures which aimtodetectPAsbyequipping thebiometric system witheitheradditionalhardwareor softwareperform- ingpresentationattackdetection (PAD). In this work we focus on PAs and PAD for finger- as well as hand-vein recognition systems. We pro- pose several approaches to create spoofing artefacts using different materials replicating the vein pattern of genuine subjects. Furthermore, corresponding PA datasets are acquired and a PAD approach, tested on handveins, ispresented. The restof thepaper isorganisedas follows: Sec- tion 2 gives an overview on PAs and PAD schemes for finger- and hand-vein recognition. In Section 3 the generation of the spoofing artefacts is explained. 65