iBeta Kids Dataset - 45,600 Videos for Presentation Attacks Liveness Detection & Verifications

Overview
iBeta Kids Dataset is a commercial biometric video dataset produced by Unidata, purpose-built for training and evaluating facial recognition and liveness detection systems on children's biometric data. The dataset contains 45,600 video files featuring 60 children across multiple age groups, recorded in a controlled studio environment with varied lighting, angles, devices, and backgrounds. It is designed to help developers build biometric systems and recognition algorithms capable of detecting presentation attacks and ensuring accurate, safe child identification.
iBeta is a leading biometric testing laboratory accredited with ISO/IEC 17025 and certifying compliance with ISO/IEC 30107. Since 2018, iBeta Lab has certified over 200 companies — 17% of which were Unidata's clients.

Attack Types
The datasets contain four main categories of presentation attacks, covering the most critical spoofing scenarios for authentication systems and identity verification workflows:

1. Real Person — live subject recorded without any disguise; the baseline class for liveness detection training
2. 2D Mask — flat printed masks simulating facial images of children, used to test recognition systems against photo-based spoofs
3. 3D Mask — volumetric masks replicating face geometry, designed to challenge biometric systems that rely on depth or texture cues
4. Replay — pre-recorded video playback attacks targeting face detection and biometric verification pipelines

Subject Demographics
The datasets contain recordings of 60 children divided evenly across four age groups, with balanced gender representation across all groups:
- 7–8 years: 8 girls (13.33%) / 7 boys (11.67%)
- 9–10 years: 8 girls (13.33%) / 7 boys (11.67%)
- 11–12 years: 8 girls (13.33%) / 7 boys (11.67%)
- 13–14 years: 8 girls (13.33%) / 7 boys (11.67%)
Each subject is labeled with metadata including ID, attack type, background, device, lighting, angle, distance, and facial features — covering 15+ annotated attributes per recording. This granular annotation enables recognition algorithms to train on realistic, age-specific biometric data across the full spectrum of younger children's developmental stages.

Technical Specifications
- Format: MP4
- Video duration: 3–5 seconds per clip
- Total files: 45,600 | Files per set: 976
- Backgrounds: 4 unique studio settings
- Recording devices: iPhone, Samsung, Xiaomi Redmi
- Collection method: Recorded by the Unidata team in a rented studio under controlled conditions
The combination of multiple devices, varied angles and lighting, and 4 distinct backgrounds ensures that training data closely mirrors real-world deployment conditions for biometric recognition technology in authentication systems.

Use Cases
- Kids Safety & Education. Schools and childcare institutions can use this dataset to develop secure biometric identification systems for building access, attendance tracking, and identity verification. Facial recognition trained on children's biometric data from different age groups provides a reliable foundation for enhancing security without compromising privacy.
- Online Platforms & Age Verification. Digital platforms targeting younger audiences rely on accurate age verification to restrict access to inappropriate content. Databases containing labeled facial images across age groups enable the development of recognition systems capable of verifying a child's identity with high precision — protecting personal data and ensuring compliance with child safety regulations.
- Smart Home Technology. Smart home systems benefit from biometric recognition that distinguishes between family members and unauthorized individuals. Training data from this kids dataset supports the development of identification algorithms that can reliably recognize children at young age, enabling targeted access restrictions to sensitive areas.

Compliance & Security
The dataset contains real biometric data collected from child subjects in a controlled studio environment, with full consent and compliance with GDPR and all applicable data protection laws. Personal information and sensitive data are handled in accordance with legal and ethical standards. All datasets are stored on AWS cloud infrastructure certified to ISO 27001 and ISO 27701 standards, ensuring secure and privacy-preserving access for enterprise clients.

Summary
The iBeta Kids Dataset is a large-scale, specialized video collection purpose-built for training biometric recognition systems on children's facial data. With 45,600 annotated videos across four attack types — Real Person, 2D Mask, 3D Mask, and Replay — covering children ages 7–14, balanced across male and female participants, and recorded across multiple devices and lighting conditions, it is one of the most targeted kids biometric datasets available. Its 15+ annotated attributes per subject, combined with strict GDPR compliance and AWS-backed secure storage, make it an essential resource for organizations developing identity verification, anti-spoofing, and child safety solutions.