BioKey-ZK: Zero-Knowledge Biometric Key Generation and Fine-Grained Identity Verification Using Standard RGB Cameras

Project

PFE-05 — Security, Computer Vision, Applied Cryptography, Cybersecurity

Enable face-based login without storing biometric data on a server. The key is derived on the device and proven using zero-knowledge techniques.

Objectives

• Develop twin-grade facial recognition using ArcFace or ViT on fine facial regions.
• Implement key derivation with a fuzzy extractor and a device-held second factor.
• Create a zk-SNARK zero-knowledge proof so that neither the image nor the key leaves the device.
• Conduct security evaluations, including reconstruction and poisoning attack tests.

Key challenge

A face alone lacks sufficient entropy, so it must be bound to a device secret.

Target

Achieve at least 92% accuracy on identical twins, with FAR/FRR comparisons against an ArcFace baseline.

Tech stack

PyTorch, SnarkJS, OpenCV, ArcFace, ViT, PyCryptodome, FastAPI, Circom, React Native.

Profile

Student in Computer Science, cybersecurity, or AI.

Format and location

• 1 intern
• Full-time, on-site internship
• 4 to 6 months
• Knowledge Place, Monastir, Tunisia
• Market: Global, with applications in finance and health

General requirements

• Willingness to learn and take on the challenge
• Relevant skills and interests
• Adequate written and spoken English
• Ability to work alongside people from diverse backgrounds and disciplines
• Availability for long-term work after the internship