Ready-to-license datasets covering 3D mask, makeup and replay presentation attacks — unannotated, with optional labeling and QA.
3D mannequin face dataset for liveness detection - 1,000 images. Listed under anti-spoofing and liveness detection datasets, and suitable for face anti-spoofing, liveness detection and presentation attack detection (PAD) model training.
Anti-spoofing advanced paper attack dataset for iBeta Level 2 - videos covering four different advanced paper attack types for liveness detection. Listed under anti-spoofing and liveness detection datasets, and suitable for liveness detection and presentation attack detection (PAD) model training, and iBeta Level 2 certification preparation.
iBeta and child face video dataset - 45,600 videos of adult and child participants of European ethnicity. Listed under anti-spoofing and liveness detection datasets, and suitable for face anti-spoofing, liveness detection and presentation attack detection (PAD) model training, including iBeta certification preparation.
iBeta Level 1 paper attacks captured with a dual-camera setup: an infrared (IR) camera plus an RGB webcam for PC and web-based onboarding. Each sample contains a synchronized IR channel and a standard webcam RGB view, covering bona fide recordings and four paper spoof variations: print and cutout, cylinder, on-actor, and 3D paper masks. Designed for ISO/IEC 30107-3 PAD Level 1 pre-evaluation in a PC or web onboarding flow - a web application accessed via a desktop webcam (IR + RGB).
iBeta Level 1 paper mask attack dataset - 22,000 videos for facial liveness detection and anti-spoofing. Listed under anti-spoofing and liveness detection datasets, and suitable for liveness detection and presentation attack detection (PAD) model training, and iBeta Level 1 certification preparation.
iBeta Level 1 - 3,000 videos of partial 2D paper mask attacks from 50 participants, recorded on a Galaxy A54 and an iPhone 14 Pro. Built for training and evaluating face anti-spoofing, liveness detection and presentation attack detection systems.
iBeta Level 1 video dataset - 35,800 videos covering participants of Asian, European and African ethnicity. Listed under anti-spoofing and liveness detection datasets, and suitable for face anti-spoofing, liveness detection and presentation attack detection (PAD) model training, and iBeta Level 1 certification preparation.
iBeta Level 2 - a cloth mask attack dataset for anti-spoofing and liveness detection, with more than 11,000 videos of elastic textile masks with printed faces, recorded on a Samsung S23 and an iPhone 14 Pro. It falls under the ISO 30107-3 Level 2 attack category and supports iBeta Level 2 certification preparation.
3,000 videos from 50 participants of print cardboard mask attacks with real accessories such as wigs, hats and glasses. Recorded on iPhone 12, iPhone 14 Pro and Samsung S23 devices. Built for training and evaluating face anti-spoofing, liveness detection and presentation attack detection systems.
A comprehensive biometric dataset for iBeta Level 2 liveness detection training and anti-spoofing research. The emphasis is on 3D attacks (masks) and on movements for active liveness (zoom-in and zoom-out, micro-movements), with high variability of devices and conditions and a high diversity of subjects.
iBeta Level 2 video dataset - 45,522 videos of participants of European ethnicity. Listed under anti-spoofing and liveness detection datasets, and suitable for face anti-spoofing, liveness detection and presentation attack detection (PAD) model training, and iBeta Level 2 certification preparation.
iBeta Level 3 high-fidelity rubber mask attack dataset - 1,000 liveness detection videos of high-fidelity mask attacks. Listed under anti-spoofing and liveness detection datasets, and suitable for liveness detection and presentation attack detection (PAD) model training, and iBeta Level 3 preparation.
A first-to-market anti-spoofing dataset for iBeta Level 3 certification testing, with more than 10,000 attack videos using 12 or more premium resin masks - the most advanced and realistic attack type in the PAD taxonomy.
An anti-spoofing dataset covering live sessions, replay, cut, print and 3D mask attacks for large-scale face anti-spoofing. It delivers a single, end-to-end resource for training and benchmarking facial liveness detection systems. By aggregating live sessions and eleven realistic presentation attack classes into one collection, it accelerates development toward iBeta Level 1/2 compliance and strengthens model robustness against the full spectrum of spoofing tactics.
Latex mask dataset for face anti-spoofing and liveness detection - 5,000 videos. Listed under anti-spoofing and liveness detection datasets, and suitable for liveness detection and presentation attack detection (PAD) model training.
A liveness detection dataset of video attacks using printed 2D masks. It focuses on cutout photo print attacks of the kind used by iBeta and NIST FATE to assess liveness detection algorithms, and is tailored to training AI models to identify variations of the cutout 2D print attack.
Liveness detection: replay attacks. More than 5,000 videos of display replay monitor attacks lasting 12 seconds or longer, alongside real photos. The attacks provide diversity of lighting, devices and screens.
A liveness detection dataset of replay attacks performed on mobile devices. It consists of 1,500 individuals who provided selfies, followed by 3,000 replay display attacks executed across 15 different mobile devices. The attacks are captured on a diverse range of low-, medium- and high-end mobile phones, providing extensive variation in screen types, lighting and environmental conditions.
Facial image dataset of participants wearing eyeglasses, face masks and sunglasses, captured indoors and outdoors with full 360-degree head rotation. Covers Chinese, White, Black, Indian, Southeast Asian and Middle Eastern subjects. Suited to occlusion-robust face recognition, face detection under partial cover, and head-pose estimation model training.
A paper mask attack dataset for training PAD and liveness detection models against low-cost 3D spoofing, with more than 2,000 videos recorded on iOS and Android devices. It falls under the ISO 30107-3 Level 1/2 attack category.
A specialized face anti-spoofing dataset featuring personalized resin mask attacks that target facial recognition systems and biometric authentication. This unique dataset uses custom-manufactured rubber face masks designed to replicate specific individuals' facial features with high precision.
A photo print attack dataset covering more than 3,000 individuals, for Presentation Attack Detection (PAD) Level 1. This biometric attack dataset uses printed photo attacks, which both iBeta and NIST FATE apply when assessing liveness detection algorithms, and is tailored to training AI models to identify photo print attacks on individuals. The print photo attacks include the zoom effects mandated by NIST FATE for improved AI training.
An anti-spoofing dataset with more than 12,500 attack videos featuring 18 hyper-realistic silicone masks. Designed for training liveness detection and presentation attack detection (PAD) models targeting iBeta Level 2 certification.
A dataset for PAD and liveness detection with three classes of presentation attack: cardboard masks with cut-out eyes, elastic textile mesh masks with full-color face printing, and full-size latex 3D masks. The mask types were selected on the basis of real-world attack scenarios.
This dataset is designed to enhance liveness detection models by simulating wrapped 3D attacks - a more advanced version of 3D print attacks, in which facial prints include 3D elements and additional attributes. It is particularly useful for iBeta Level 2 certification and anti-spoofing model training.
