Anti-Spoofing Replay Dataset - 42,743 Videos for Facial Recognition and Liveness Detection

Overview
The Anti-Spoofing Replay Dataset is a large-scale commercial biometric video dataset produced by Unidata, combining recordings from two complementary sources: phone-captured and PC-captured facial videos. Together they form a unified training resource of 42,743 video clips — 38,029 phone recordings organized into 20,018 paired sets, plus 4,714 PC-recorded clips — designed to develop and benchmark face antispoofing algorithms, facial recognition systems, and liveness detection technologies.

The dataset provides reliable training data for detecting replay attacks — one of the most prevalent spoofing techniques in real-world biometric authentication scenarios. By covering both mobile devices and desktop setups, it simulates the full spectrum of device types encountered in production security systems.

Dataset Composition
- Total videos: 42,743
- Phone recordings: 38,029 videos / 20,018 paired sets (2 videos per set)
- PC recordings: 4,714 videos / 4,714 individuals
- Data types: Video (MP4, MOV)
- Primary tasks: Face recognition, face detection, liveness detection, replay attack detection

Subject Demographics
Both male and female participants are represented across the full dataset. Each recording is accompanied by technical metadata — age, gender, and ethnicity — enabling demographic filtering and privacy-aware model training.
Gender breakdown: Male, Female.

Technical Characteristics
- Video formats: MP4, MOV
- Recording devices: Mobile phones (various models) and PC-based cameras
- Labeling: Technical metadata per subject — age, gender, ethnicity
- Collection methodology: Data was collected via crowdsourcing platforms, ensuring diverse recording environments and realistic replay attack conditions
- Compliance: GDPR and applicable data protection regulations; stored on AWS infrastructure certified to ISO 27001 and ISO 27701

Attack Types & Spoofing Scenarios
The dataset focuses on replay attacks — a category of presentation attacks where a recorded video of a real face is played back to deceive a recognition system. Two hardware contexts are covered:
- Phone-based replay attacks — spoofing attempts captured and replayed via mobile devices, reflecting the most common attack vector in mobile biometric authentication
- PC-based replay attacks — scenarios involving desktop or laptop cameras, relevant for web-based identity verification and enterprise access control
This dual-device coverage allows anti-spoofing models to generalize across different attack types and hardware profiles encountered in real-world deployments.

Use Cases
- Financial Services & FinTech. Banks and digital financial platforms use this dataset to strengthen biometric authentication pipelines, training detection systems to recognize replay attacks during remote onboarding, KYC procedures, and identity verification workflows. FinTech companies face up to 100,000 biometric spoofing attempts monthly — this data directly supports fraud prevention.
- Mobile Identity & Access Management. Mobile authentication providers refine face recognition and anti-spoofing algorithms under real-world conditions. The phone video portion covers spoofing techniques including replay attacks and printed photo attempts across diverse mobile devices.
- Enterprise Security Systems. Security technology providers improve liveness detection for access control and surveillance systems. Exposure to replay attacks and diverse spoofing techniques enables accurate detection of fake users.

Why This Dataset
1. 42,743 videos across two device categories in a single purchase
2. Metadata annotated per subject: age, gender, ethnicity — no PII exposed
3. Collected via crowdsourcing for real-world recording diversity
4. Compatible with standard anti-spoofing, face recognition, and liveness detection pipelines
5. Fully GDPR-compliant; AWS-hosted with ISO 27001/27701 security controls
6. Free sample available before purchase

Compliance & Security
All biometric data is collected from participants under legally permissible conditions and complies with GDPR and relevant data protection regulations. Storage is hosted on AWS cloud infrastructure certified to ISO 27001 and ISO 27701 standards.

Summary
The Anti-Spoofing Replay Dataset is a comprehensive, multi-device video collection purpose-built for training and evaluating facial recognition and liveness detection systems against replay attacks. With 42,743 total video clips — 38,029 from mobile phones in 20,018 paired sets, and 4,714 from PC cameras — it delivers the scale and device diversity needed to build robust anti-spoofing solutions. Rich metadata (age, gender, ethnicity), dual-format support (MP4/MOV), and full regulatory compliance make it a production-ready resource for biometric security teams, FinTech platforms, and computer vision researchers working on spoofing detection and authentication systems.