iBeta Level 1 Dataset - 35,800 Videos for Presentation Attacks Liveness Detection & Verifications

Overview
The iBeta Level 1 Dataset is a commercial biometric video dataset produced by Unidata, designed to train and evaluate face liveness detection systems against presentation attacks and to support compliance with the iBeta Level 1 PAD (Presentation Attack Detection) certification. The dataset is compliant with the ISO/IEC 30107-3 standard — the international benchmark for biometric testing and attack detection quality assurance.

iBeta is a leading independent biometric testing laboratory accredited with ISO/IEC 17025 and certifying compliance with ISO/IEC 30107. Since 2018, iBeta has certified over 200 companies — 17% of which were Unidata's clients. By achieving the iBeta Level 1 certification, technology companies demonstrate their commitment to robust biometric security and reliable spoofing detection.

Attack Types
The dataset covers 8 attack types recorded in controlled studio conditions, simulating real-world presentation attacks against face recognition and authentication systems:

1. Real Person — genuine live video of an actor
2. 2D Mask — printed photo cut out along the facial contour
3. Wrapped 2D Mask — printed photo attached to a cylindrical surface
4. 2D Mask with Eyeholes — printed photo with cut-out holes for eyes
5. 3D Mask — multi-panel cardboard construction forming a volumetric portrait
6. Smartphone Replay — a person's photo displayed on a phone screen
7. PC Replay — a person's photo displayed on a computer monitor
8. Tablet Replay — a person's photo demonstrated on a tablet screen

Each attack type was recorded across multiple devices and backgrounds to maximize scenario diversity for biometric evaluations and PAD testing.

Subject Demographics
The dataset features 50 actors across a range of demographic profiles, ensuring realistic diversity for biometric systems training and independent testing.
Gender distribution:
- Male and female participants are both represented

Ethnicity breakdown:
- European — 60%
- Asian — 20%
- African — 20%

Age distribution:
- 18–25: ~32% (11,456 videos)
- 26–32: ~28% (10,024 videos)
- 33–39: ~18% (6,444 videos)
- 40–46: ~14% (5,012 videos)
- 47+: ~8% (2,864 videos)

Actor features include: bald, beard/mustache, makeup, scar, piercing, and no features — covering a wide range of real-world appearance variability critical for anti-spoofing models.

Technical Specifications

1. Total videos: 35,800
2. Format: MP4, MOV
3. Resolution: min 720×1280 — max 2160×3840 (up to 4K)
4. Video duration: ~5 seconds per clip
5. Number of backgrounds: 10 unique environments
6. Recording devices: Samsung Galaxy A53, OPPO A18, iPhone 14, Apple iPad 10.2, and others
7. Data collection: Studio environment by the Unidata team
8. Labeling: Technical metadata per video — age, gender, ethnicity

Use Cases
- Financial Services. Banks and fintech companies use this dataset to strengthen biometric authentication solutions and detect spoofing attempts before they compromise digital identity systems. It supports certification processes and compliance with accredited biometric security standards, helping prevent identity fraud in mobile banking and payment platforms.
- Biometrics & Technology Industry. Technology companies and research teams rely on this dataset to validate recognition systems, improve testing capabilities for attack detection, and benchmark liveness detection models. It enables refining authentication solutions across the global biometrics industry, supporting both product development and independent quality assurance.
- Certification & Compliance. Accredited laboratories rely on the dataset to benchmark biometric evaluations against Level 1 requirements and confirm compliance with international standards such as ISO/IEC 30107.

Compliance & Security
All data consists of real-world recordings collected from actors under controlled conditions. The dataset complies with GDPR and applicable data protection regulations. Storage is hosted on AWS cloud infrastructure certified to ISO 27001 and ISO 27701 standards.

Summary
The iBeta Level 1 Dataset is a large-scale, high-quality video collection purpose-built for training and evaluating face liveness detection systems at the iBeta Level 1 certification standard. With 35,800 videos across 8 attack types, 50 actors, diverse ethnicity and age coverage, up to 4K resolution, 10 backgrounds, and metadata-rich labeling, it delivers the breadth and realism needed for robust biometric security research and enterprise PAD certification workflows. Over 30 companies have passed iBeta certification using Unidata datasets.