Deep insights on processing strata, features and detectors for fingerprint and iris liveness detection techniques.

Fingerprint and iris traits are used in sensitive applications and so, spoofing them can impose a serious security threat as well as financial damages. Spoofing is a process of breaking biometric security using artificial biometric traits. This spoofing can be avoided by detecting the liveness of the biometric traits. Hence, liveness detection techniques have become an active research area. However, liveness detection techniques are also prone to attack because of advanced spoofing materials. Hence, they are subjected to further development to face futuristic spoofing and compromising real biometric traits. To aid the development, this paper technically and informatically reviews the state-of-the-art liveness detection techniques in the last decade. Firstly, the paper reviews the processing strata, adopted features and detectors in the existing liveness detection techniques. Secondly, the paper presents the benchmark datasets, their characteristics, availability and accessibility, along with the potential spoofing materials that have been reported in the literature under study. Thirdly, the survey reports the performance of the techniques on the benchmark datasets. Eventually, this paper summarizes the findings, gaps and limitations to facilitate strengthening of liveness detection techniques. This paper further reports that the Fingerprint Liveness Detection (FLD) techniques such as Slim-ResCNN, JLW and Jung CNN have achieved a better accuracy of 94.30%, 98.61% and 97.99%, respectively on LivDet19 datasets. It has been observed that CNN-based architectures have outperformed in significant number of FLD datasets. In contrast, Support Vector Machine (SVM) with appropriate shallow and deep features has achieved equivalent performance against deep classifiers on detecting iris spoofs from Iris Liveness Detection (ILD) datasets. [ABSTRACT FROM AUTHOR]