fheye

This project redefines personal and community security by enabling individuals to contribute to identifying criminals or locating missing persons—all while maintaining complete privacy. Users capture photos on their devices, where facial recognition technology detects faces and extracts unique embeddings, which are encrypted using Fully Homomorphic Encryption (FHE). These encrypted embeddings are uploaded to a blockchain network, ensuring that raw data never leaves the device. This privacy-preserving process enables computations on encrypted data, making it possible to compare embeddings securely without revealing sensitive information. Through decentralized infrastructure, the encrypted data can be matched against trusted reference images, such as those of wanted criminals or missing children, provided by authorized organizations. A match triggers a webhook that sends real-time, location-based alerts via Push Protocol to individuals near the detection site. This creates a system where communities can respond quickly to potential threats or emergencies without relying on centralized authorities. The use of blockchain ensures data immutability and transparency while preserving trust among participants. The project combines advanced technologies to create a scalable and impactful solution for modern security challenges. By leveraging FHE for privacy, blockchain for decentralization, and real-time notifications for responsiveness, it empowers users to play an active role in community safety. This system transforms everyday photos into a tool for public good, fostering a secure and collaborative environment where everyone contributes to making their surroundings safer and more informed.

Our project combines advanced encryption, decentralized networks, and real-time notifications to create a scalable and secure self-security system. We use Fhenix and Inco as the core Fully Homomorphic Encryption (FHE) layer, enabling computations on encrypted data. For computationally intensive tasks like generating facial embeddings from images, we rented GPU instances on Hyperbolic and configured them as FastAPI servers for efficient API-based processing. Since The Graph doesn’t natively support Fhenix, we deployed our own custom Graph Node. This allows us to index and query encrypted data on Fhenix contracts seamlessly. This setup ensures that decentralized data interactions remain efficient and secure. For real-time notifications, we integrated Push Protocol. When a match is detected between encrypted embeddings and reference images (e.g., missing persons or criminals), a webhook triggers location-based alerts to notify nearby users. A key innovation was transforming GPU machines into API servers, optimizing resource use for embedding generation. Additionally, deploying a custom Graph Node for Fhenix allowed us to overcome limitations and ensure compatibility. By blending these components, we created a decentralized, privacy-preserving system that empowers individuals to contribute to community security.