Case Study: Leading AI Security and Biometrics Company achieves improved anti-spoofing fraud detection with Shaip’s 25,000-video dataset

Anti-Spoofing Video Datasets Enhancing AI Security Against Fraudulent Attacks

Leading AI Security and Biometrics Company engaged Shaip to solve a critical data gap for anti-spoofing model training: the need for large-scale, high-quality, technically consistent video data with balanced ethnic representation and rich metadata. The customer required an off-the-shelf anti-spoofing video dataset that met strict specs (≥720p, ≥26 FPS), ensured one real and one replay attack per participant, and supported robust fraud-detection model development.

Shaip delivered a phased, validated dataset of 25,000 videos (12,500 real and 12,500 replay attacks) from 12,500 unique participants across five ethnicity groups, with 12 metadata attributes and four batch deliveries to ensure steady throughput and quality control. The Shaip dataset improved the client’s ability to detect replay attacks across devices, lighting and demographic variations, strengthened model robustness and scalability for future anti-spoofing enhancements, and was credited by the customer as instrumental in enhancing fraud detection in biometric authentication.