iBeta Level 3 Dataset
10,000+ videos from 12+ mask attacks across two mask types,
tailored for iBeta Level 3 certification
Check samples on Kaggle
iBeta Level 3 dataset summary
|
Parameter
|
Value
|
|---|---|
|
Volume
|
10,000+ videos with 12 unique high-fidelity 3D masks
|
|
Coverage
|
High-fidelity rubber and custom 3D resin mask attacks for iBeta Level 3 PAD
|
|
Demographics
|
12 unique mask configurations across varied identities and ethnicities
|
|
Devices
|
iOS and Android phones (multiple device models)
|
|
Conditions
|
Indoor capture across two locations and 80+ backgrounds, with natural, artificial, directional, and low-light conditions
|
Introduction
The iBeta Level 3 Dataset is the most comprehensive collection of advanced 3D mask attacks built for iBeta Level 3 certification – the highest tier of biometric presentation attack detection testing. It combines two distinct attack categories: high-fidelity rubber masks and custom 3D resin masks, totaling 12 unique mask configurations across 10,000+ videos. Each capture session is performed on diverse iOS and Android devices under controlled conditions to mirror real-world spoofing scenarios. iBeta Level 3 certification tests biometric systems against ISO/IEC 30107-3 – the international standard for biometric presentation attack detection at the highest assurance level, where commercial-grade 3D masks must be reliably distinguished from real faces
Dataset Features
- Two mask types in one dataset: The dataset combines two distinct mask families: high-fidelity rubber masks and custom 3D resin masks, which produce fundamentally different visual artifacts. Training on both eliminates the blind spots that single-type mask datasets leave open
- Dual-recording methodology: Each resin mask attack video is paired with a real face video of the same individual whose features the mask was modeled after, enabling direct real-vs-spoof comparison and stronger contrastive training within a single identity
- Per-mask attribute variation: Each of the 12+ mask identities is recorded across ~50 attribute combinations: different hairstyles, glasses, and accessories, to maximize within-mask visual diversity and prevent models from overfitting to surface-level cues
- Active liveness protocol: Every video follows a standardized active liveness sequence: zoom-in and zoom-out phases, left and right rotation, and natural head movements, mirroring the active checks issued by production verification systems
- Multi-camera capture: All masks are recorded with front-facing (selfie) cameras, with back camera angles available for select masks to provide additional viewpoint diversity for models that need to handle non-selfie capture surfaces
- iBeta Level 3 difficulty alignment: Capture protocols and PAD scenarios are aligned to iBeta Level 3 certification, where APCER, BPCER, and ACER thresholds are tighter than at Levels 1 and 2, making this the most difficult certification tier to prepare for
- Continuously expanding: The dataset is actively maintained and growing as new high-fidelity mask types and capture scenarios are added, customers receive ongoing access to additions as the iBeta Level 3 standard evolves
What's Inside
High-Fidelity Rubber Masks
Advanced 3D rubber masks with realistic skin-like texture and flexibility. These masks allow natural facial movement, making them particularly challenging for active liveness checks. 7 unique mask identities with diverse attributes
3D Resin Masks
Custom-manufactured resin masks designed to replicate specific individuals’ facial features with high geometric precision. Rigid structure with fine detail — a different challenge profile from rubber masks. 5+ unique mask identities with dual-recording (real face + mask)
Real-World Validation: Open Liveness Model Stress Test
To demonstrate the practical value of this dataset, we tested its samples against Doubango’s open-source face liveness model – a publicly available liveness detection SDK used as a reference implementation by anti-spoofing researchers and developers
Key result: Doubango’s liveness classifier rated the high-fidelity mask attack below as 86.85% genuine – failing to detect the spoof at the liveness layer. The attack was only flagged by Doubango’s secondary injection-detection module, not by liveness scoring itself
Use cases and applications
iBeta Level 3 Certification Compliance:
- Helps to train the models for iBeta level 3 certification tests
- Allows pre-certification testing to assess system performance before submission
Inhouse Liveness Detection Models:
- Used for training and validation of anti-spoofing models
- Enables testing of existing algorithms and identification of their vulnerabilities against spoofing attacks
How companies achieved iBeta with us
Datasets like this contributed to 21% of companies that passed iBeta certification in 2025 – all Axon Labs clients
From Failed Attempt to Success: UK Fintech Passed iBeta 2
UK fintech invested in datasets from another vendor for iBeta 2 and failed. The gap analysis revealed critical missing attack vectors. They needed a solution fast
Double Success: E-KYC Startup Achieves iBeta 1 & 2
A US-based e-KYC startup needed to prove their facial recognition technology met the highest industry security standards. Their ambitious goal: pass iBeta 1 and 2 Certification
Technology company from Vietnam: iBeta Level 2 success
A Vietnam-based AI/Big Data firm coached by Axon Labs passed iBeta PAD Level 2 on the first try with 0% successful spoofs; the solution claims 99.9% face-recognition accuracy
Learn the iBeta 3 context
iBeta Level 3 introduces stricter requirements: mandatory active liveness (zoom in/out), realistic high-fidelity mask attacks, and tighter APCER/BPCER/ACER limits. For implementation details and a practical checklist, see our: iBeta Level 3 guide
File format and accessibility
- Format: Videos are optimized for compatibility with mainstream ML frameworks
- Resolution and frame rate: Videos are high-resolution with frame rates calibrated for capturing quick and realistic mask placements, ensuring precise data for model training
Potential customisation options:
- Filming attack videos with targeted movements (zoom in/zoom out, specific gestures)
- Filming attack videos on target devices (webcams, specific phone models)
- Using your SDK for custom attack scenarios spoofing your ML model
- RGB and USB camera support for diverse research and testing needs
- Multi-angle recordings (close-up, far, left, right) available for select masks
Legal & Compliance
We prioritize data privacy, ethical AI development, and regulatory compliance. Our Silicone Mask Attack Dataset is collected and processed in full accordance with global data protection standards including GDPR, ensuring legality, security, and responsible AI practices
Sample dataset
A sample version of this dataset is available on Kaggle. Leave a request for additional samples in the form below
Have a question?
We collect data from our internal team. All information is further verified by our specialists
Once your enquiry has been sent, we will contact you to discuss the details and complete the necessary paperwork. The timing of receiving the dataset depends on the specific request and additional requirements
Our unique selling point is to provide legally clean datasets to our customers. We obtain the consent from all the participants to use their data for AI model development. We are able to provide comprensive reporting on the licensing, data collection and privacy compliance of our datasets. Although there seems to be a diverse response to how to control AI development and deployment, we are able to service global customers seeking to launch global AI products.
The dataset follows iBeta testing protocols and includes diverse attack scenarios that mirror real-world spoofing attempts. It covers both passive and active liveness testing requirements with proper demographic representation and standardized capture conditions essential for certification preparation
The price depends on your specific requirements. Please submit a request to receive a free consultation
Contact us
Tell us about yourself, and get access to free samples of the dataset
Didn't find what you were looking for?
Our collection includes many datasets for various requests
iBeta Level 1 Dataset
– 35,000+ videos
– 85+ participants
– zoom in and
zoom out
iBeta Level 2 Dataset
– 25 000+ videos
– 3D masks
– iBeta Level 2
iBeta Level 3 Dataset
– 10,000+ videos
– 12 Unique Masks
– iBeta Level 3
Display Replay Dataset for Liveness Detection
– 9,000+ videos
– 6,500+ participants
– Balanced mix of genders and ethnicities
