Non‐reference image quality assessment and natural scene statistics to counter biometric sensor spoofing

Abstract

Non-reference image quality measures (IQM) as well as their associated natural scene statistics (NSS) are used to distinguish real biometric data from fake data as used in presentation/sensor spoofing attacks. An experimental study shows that a support vector machine directly trained on NSS as used in blind/referenceless image spatial quality evaluator provides highly accurate classification of real versus fake iris, fingerprint, face, and fingervein data in generic manner. This contrasts to using the IQM directly, the accuracy of which turns out to be rather data set and parameter choice-dependent. While providing very low average classification error rate values for complete training data, generalisation to unseen attack types is difficult in open-set scenarios and obtained accuracy varies in almost unpredictable manner. This implies that for each given sensor/attack setup , the ability of the introduced methods to detect unseen attacks needs to be assessed separately.

A Contemporary Survey of Multimodal Presentation Attack Detection Techniques: Challenges and Opportunities

“…These are categorized into two main parts reference-based and non-reference-based. In the reference-based method, an undistorted [26] Face Presentation attack Self-constructed Proposed a conditional random fields (CRFs) method Shahin et al [4] Hand vein and Fingerprint Spoof attack Self-constructed Proposed a new prototype design for multimodal biometrics Rodrigues et al [5] Face, Fingerprint Presentation attack FVC2004DB1, FERET-b series Proposed two novel schemes, the extension of the likelihood ratio based fusion and fuzzy logic Liting et al [28] Face Presentation attack Self-constructed Presented a face live detection method which utilized physiological motion Bao et al [16] Face Photo-attack Self-constructed Proposed optical flow field method to recognize the liveness of face Chetty [29] Face and voice Spoof attack VidTIMIT, DaFeX Corpora Proposed a new fuzzy fusion technique for liveness detection Komogortsev et al [30] Eye Spoof attack Self-constructed Proposed Oculomotor Plant Characteristics (OPC) based liveness detection method Barrero et al [7] Face, Iris Indirect Attack BioSecure Presented and evaluated the first software-based attack Singh et al [31] Face Spoofing attack Self-constructed Proposed a liveness detection measure to recognize and find the liveness of face, based on challenge and response method Akhtar et al [14] Face, iris fingerprint, Spoofing attack ATVS-Flr, ATVS-FFp Proposed mobile biometric liveness detection (MoBio LivDet) method Wild et al [42] Face and Fingerprint Direct attack LivDet 2013, CASIA FASD Proposed 1-median filtering as a spoofing resistant Das et al [8] sclera and iris Direct attack Self-constructed Proposed a framework for software-based liveness detection Boulkenafet et al [24] Face Presentation attack MSU-MFSD,CASIA-FASD Proposed an approach based on color texture analysis for face spoofing detection Lee et al [37] Finger-vein Spoof attack Self-constructed Proposed a finger-vein biometric recognition system based on image quality assessment Parveen et al [23] Face Presentation attack UPM, CASIA FASD, NUAA A dynamic local ternary pattern has been proposed which utilizes Weber's law Sollinger et al [40] Iris, face fingerprint, and finger vein Presentation attack ATVS-Flr, ATVS-FFp, IDIAP Proposed non-reference image quality measures (IQM) to differentiate between fake and real data Kavitha et al [9] Face Fake face attack CASIA FASD, MSU MFSD Proposed multimodal biometric framework to detect face spoofing Sahidullah et al [17] Voice Replay attack Self-constructed Proposed a body conducted sensor and throat microphone for automatic speaker verification reference image is utilized to estimate the quality of the test image while in the non-reference-based method pre-trained statistical models are utilized to ...…”

“…Fingerprint images were analyzed in the spatial domain and frequency domain and final features were decided by the co-occurrence probabilities. In [40] non-reference image quality measures were proposed to distinguish between genuine and fake data. In this method, accurate classification of real versus fake iris, fingerprint, face, and finger vein data was achieved.…”

Deep convolutional neural networks for face and iris presentation attack detection: survey and case study

“…This is assuming that many image quality metrics are distorted by the display device or paper. Some researchers used individual quality-based features, while others used a combination of quality metrics to better detect spoofing and differentiate between bona-fide and attack iris [12,18,[53][54][55] or face [56][57][58] or both [59,60]. For example, Galbally et al [12] investigated 22 iris-specific quality features and used feature selection to choose the best combination of features that discriminate between live and fake iris images.…”

State of the Art in Vascular Biometrics

“…In [21], the authors successfully apply general-purpose non-reference image quality metrics to discriminate real finger vein images from fake ones. Subsequent work [242] additionally applies natural scene statistics and looks into the issue of cross-sensor and crosssubject finger vein presentation attack detection. However, it is often cumbersome to identify and/or design texture descriptors suited for a specific task in this context.…”