Biometric Verification System
Eliminating demographic bias with an ethically sourced biometric dataset spanning 10+ countries.
The Challenge
A global identity verification platform faced heavy scrutiny because their facial recognition system showed significant performance disparities across different racial and age demographics. They urgently needed a highly diverse, ethically sourced dataset to fix this bias.
Our Solution
Dserve AI executed a massive, privacy-first collection campaign across 10+ countries. We collected face images, fingerprint scans, and iris photos from fully consenting participants spanning 30+ demographics and a wide range of age groups. To harden the system against fraud, we also generated thousands of anti-spoofing samples, including printed photos, 3D masks, and screen replays.
The Impact
"The new dataset allowed the client to retrain their model from the ground up. The resulting system achieved sub-0.1% false-accept rates while simultaneously reducing demographic bias gaps by 60%, setting a new ethical standard in the identity verification industry."
Ethical Sourcing & Liveness Pipeline
01. Informed Consent
Rigorous legal frameworks ensuring participants maintain data sovereignty and the right to delete.
02. Demographic Matrix
Targeted recruitment across 80 ethnicities to ensure perfectly balanced dataset distributions.
03. Spoof Generation
Creating deepfakes, 3D silicone masks, and high-res printouts to test the system's limits.
04. Adversarial QA
Red-teaming the dataset to find any remaining demographic bias before final delivery.