iBeta Level 3 Dataset

The iBeta Level 3 Dataset covers the two mask families tested at iBeta Level 3 PAD certification: high-fidelity rubber masks and custom 3D resin masks. These are the most sophisticated presentation attack classes, professionally manufactured 3D masks that replicate specific target identities with photorealistic skin texture and three-dimensional face geometry. Rubber and resin masks produce fundamentally different visual artifacts, so a single dataset covering both is essential for production-grade L3 preparation

The dataset contains 10,000+ videos across 12 unique high-fidelity masks, combining both rubber and custom 3D resin mask families. Each mask is recorded with multiple attribute combinations (hairstyles, glasses, accessories) under an active liveness protocol (zoom-in/out, head tilts, left/right rotation, natural head movements). All videos are captured with front-facing (selfie) cameras on iOS and Android devices, with back camera angles available for select masks, across 40+ indoor office backgrounds and varied lighting conditions

iBeta Level 3 is the highest tier of biometric presentation attack detection testing, requiring liveness models to reliably detect high-fidelity 3D masks. Testing follows ISO/IEC 30107-3 at the strictest assurance level, where APCER, BPCER, and ACER thresholds are tighter than at Levels 1 and 2. L3 certification is typically required for deployments where commercial-grade 3D masks pose a real fraud risk, high-value identity verification, government identity systems, and banking authentication

The iBeta Level 3 Dataset provides combined coverage of both mask families in a single training resource (10,000+ videos across 12 masks). The specialized iBeta 3 High-Fidelity Rubber Mask Dataset focuses only on 7 rubber masks with 1,000+ videos. The 3D Resin Mask Dataset focuses only on 5 resin masks with 2,000+ videos. Use the combined dataset for comprehensive L3 training; use the specialized datasets for targeted single-material training or testing

All mask attacks were captured using front-facing (selfie) cameras on iOS and Android devices, with back camera angles available for select masks to provide additional viewpoint diversity. Each of the 12 masks is recorded across multiple attribute combinations (hairstyles, glasses, accessories) under an active liveness protocol including zoom-in, zoom-out, head tilts, left and right rotation, and natural head movements. Captures span 40+ indoor office backgrounds under natural and artificial lighting, including additional warm/cool directional lighting and low-light scenarios

The dataset is built for the most demanding iBeta PAD tier through three quality dimensions. First, professional mask construction: all 12 masks are manufactured to L3 certification realism standards, replicating specific target identities with photorealistic detail. Second, two-family coverage: training on both rubber and resin mask artifacts eliminates blind spots that single-material datasets leave open. Third, methodological depth: each mask is captured across multiple attribute combinations and lighting conditions for strong within-mask training diversity. Axon Labs datasets have contributed to 21% of companies passing iBeta certification in 2025

Yes. All data is collected with explicit written informed consent from every participant in compliance with GDPR Article 9, which specifically governs the processing of biometric data. The dataset is licensed for commercial use in AI model training, validation, and iBeta certification preparation. Comprehensive compliance documentation, including consent provenance, collection methodology, and legal basis review, is available upon request for regulatory or audit purposes

Yes. A sample version of this dataset is available on request, you can verify the quality of both rubber and resin mask attacks, active liveness sequences, attribute variation, and format compatibility with your training pipeline before committing to the full dataset. Submit a request through the form on this page for sample access, typically delivered within 1–2 business days