The Swax Benchmark: Attacking Biometric Systems with Wax Figures

Abstract

A face spoofing attack occurs when an intruder attempts to impersonate someone who carries a gainful authentication clearance. It is a trending topic due to the increasing demand for biometric authentication on mobile devices, high-security areas, among others. This work introduces a new database named Sense Wax Attack dataset (SWAX), comprised of real human and wax figure images and videos that endorse the problem of face spoofing detection. The dataset consists of more than 1800 face images and 110 videos of 55 people/waxworks, arranged in training, validation and test sets with a large range in expression, illumination and pose variations. Experiments performed with baseline methods show that despite the progress in recent years, advanced spoofing methods are still vulnerable to high-quality violation attempts.

Directed Gaze Trajectories for Biometric Presentation Attack Detection

“…Most 3D PAD datasets are based on wearable hard-resin or silicone masks. Wax figure face databases were reported in [ 34 , 35 ] for super-realistic 3D presentation attack research. For the work presented in this paper, a database has been locally collected as none of the public databases include the type of challenge–response, artefacts and device formats, which are investigated in this paper.…”

Multi-Modal Face Anti-Spoofing Challenge

“…WFFD [27] introduced the first waxwork dataset, comprising 450 subjects and 2,200 images, significantly increasing fidelity. Similarly, SWAX [60] included images and videos of waxworks. Recent 3D FAS datasets have considered new attack types, ethnic diversity, and complex recording conditions, such as SiW-M [43] with 13 fine-grained attack types, CASIA-SURF CeFA [35] addressing ethnic bias with three ethnicities, and HiFiMask [39] encompassing six lighting conditions, seven recording devices, and six scenarios.…”

“…A review of current datasets, as shown in Table 1, reveals that most datasets contain either 2D or 3D PAs exclusively. Datasets such as [4, 9-11, 50, 52, 58, 68, 83] feature 2D print or display PAs, while [18,20,28,39,41,57,60,79] include 3D mask or model PAs. Some datasets, such as [31,43,84], encompass both 2D and 3D PAs; however, their diversity remains unsatisfactory due to deficiencies in other elements, such as the absence of high-fidelity 3D models.…”