Centralized Multi-Key Fully Homomorphic Encryption from CKKS

Fully Homomorphic Encryption (FHE) enables computation over encrypted data without revealing its content. However, most existing FHE schemes are designed for singlekey settings, limiting their use in collaborative applications involving multiple data owners. Multi-Key FHE (MKFHE) extends support to such scenarios but often incurs ciphertext expansion proportional to the number of participants and relies on complex, multi-round decryption. We propose a Centralized Multi-key Fully Homomorphic Encryption (CMKFHE) scheme that enables efficient and scalable multiuser computation. Each child node holds a distinct FHE key derived from public parameters generated by the parent, and functions as a standard FHE system. Ciphertexts produced by different children also support homomorphic evaluation, with the result decryptable only by a parent node possessing a master key. Computations within a child domain remain decryptable by that child, while sibling nodes cannot access each other's ciphertexts. We present a construction based on the CKKS framework that supports approximate arithmetic and ciphertext packing while maintaining constant ciphertext size, making it suitable for applications such as federated learning and privacy-preserving analytics.