Datasets:
video video | label class label 2
classes |
|---|---|
0Samsung_S23 | |
0Samsung_S23 | |
0Samsung_S23 | |
0Samsung_S23 | |
0Samsung_S23 | |
0Samsung_S23 | |
1iPhone14Pro | |
1iPhone14Pro | |
1iPhone14Pro | |
1iPhone14Pro | |
1iPhone14Pro | |
1iPhone14Pro | |
1iPhone14Pro | |
0Samsung_S23 | |
0Samsung_S23 | |
0Samsung_S23 | |
0Samsung_S23 | |
0Samsung_S23 | |
1iPhone14Pro | |
1iPhone14Pro | |
1iPhone14Pro | |
1iPhone14Pro | |
1iPhone14Pro |
Textile 3D Face Mask Attack Dataset
Cloth mask attack dataset for anti-spoofing and liveness detection — 11,000+ videos of elastic textile masks with printed faces on Samsung S23 and iPhone 14 Pro. ISO 30107-3 Level 2 attack category, iBeta Level 2 certification prep
Overview
Cloth mask attack dataset for training anti-spoofing and liveness detection models — 11,000+ videos of elastic textile face masks (balaclava-type) with printed facial images. Captured on Samsung S23 and iPhone 14 Pro in indoor environments with varied lighting
Textile masks conform to the wearer's head through fabric elasticity, partially mimicking facial geometry — a low-cost alternative to silicone and latex 3D masks that tests the same ISO 30107-3 Level 2 attack category
Technical Specifications
| Parameter | Value |
|---|---|
| Videos | 11,000+ |
| Unique subjects | 30+ IDs |
| Devices | Samsung S23, iPhone 14 Pro |
| Video length | ~7 sec |
| Active liveness cues | Zoom in/out, left/right rotation, head tilt |
| Attributes | Varied hairstyles, glasses, wigs, beards |
| Environment | Indoor, multiple lighting conditions |
Spoof Success Demonstration
This screenshot demonstrates that our cloth mask attack successfully bypasses a state-of-the-art face liveness detection model from KBY-AI — a widely used commercial solution known for its high performance in biometric verification systems. The model classified the spoofed face as Real with a high confidence score (0.89), confirming that even top-tier PAD systems can be vulnerable to this type of attack

Attack Classification
Falls under 3D artifact / mask attacks in the ISO 30107-3 taxonomy. Textile masks are tested in iBeta Level 2 certification alongside silicone and latex masks — this dataset covers the budget-accessible end of that attack spectrum
Training Scenarios
- Multi-class PAD: combine with silicone mask, latex mask, and replay attack data for full attack coverage
- iBeta Level 2 prep: pair with live session data to simulate certification test conditions
- Active liveness testing: zoom and rotation cues included
Related Datasets
- Silicone Mask Attack Dataset — high-fidelity 3D mask attacks
- Latex Mask Dataset — flexible 3D mask attacks
- Wrapped 3D Paper Mask Dataset — paper-based volumetric attacks
- Face Anti-Spoofing Dataset — 100,000+ videos, 11 attack types
Full version of dataset is availible for commercial usage - leave a request on our website Axonlabs to purchase the dataset 💰
- Downloads last month
- 53