SEFIKS explains encrypted nearest neighbor search with LightPHE

The author explains how to perform encrypted nearest neighbor search using homomorphic encryption, dividing the problem into two distinct settings based on gallery ownership. In the first setting, where a server holds a plain gallery and a client has an encrypted query, the protocol uses Damgard-Geisler-Kroigaard encryption with interactive argmin and equality tests. In the second setting, where an encrypted gallery resides in an untrusted cloud, an additive cryptosystem boosted with the Catalano-Fiore transform enables direct distance evaluation. Practical multi-language code examples using LightPHE demonstrate implementation in Python, Go, Java, and TypeScript. Open article →