A Zero-Knowledge Backend Architecture for Preventing Data Leakage Using Hashing and Encryption

Backend services commonly receive, store, search, and analyze sensitive information in plaintext, creating exposure risks in databases, logs, caches, and application memory. This paper presents a zero-knowledge backend architecture that combines client-side data protection, protected search tokens, access-policy checks, and selected arithmetic over ciphertext representations. A lightweight Node.js prototype uses built-in modules only and demonstrates protected user-record storage, prefix-token search, and limited encrypted arithmetic. In the supplied experiment, username search took 0.077063 ms and encrypted aggregation took 0.035841 ms; statistics decryption took 20.696818 ms. These are prototype measurements, not general performance guarantees. The design reduces direct plaintext storage but is not a complete zero-knowledge system or production-ready cryptographic implementation. In particular, deriving a key directly from a predictable user identifier is insecure for production use, and protected indexes leak search structure. The contribution is an integrated architecture and a clearly scoped prototype evaluation for processing selected data without first decrypting stored values on the backend.