Optimizing Encrypted Computation in Cloud Environments: A Residue Number System-Based Approach

Abstract

Nowadays, cloud computing has become a default platform for encrypted computation of the data, and obscure and fast encrypted computation can be critical. Nowadays RSA and AES, among others, are devices providing different levels of security, but are relatively inefficient for advanced operations on encrypted data within very large-scale cloud environments. Since there are performances bottlenecks with the application of Fully Homomorphic Encryption (FHE), computation on ciphertexts is allowed but operating on ciphertexts can be slow. We explore how to integrate the Residue Number System (RNS) into encrypted computation frameworks, considering in particular encryptions in the cloud. The proposed framework acquires appreciable improvements on encryption/decryption latency, memory efficiency, and scalability, by making use of RNS properties, namely, carry free arithmetic, modular parallelism, and Chinese Remainder Theorem-based reconstruction. However, RNS based encryption is faster up to 10x faster with respect to FHE and takes much less memory than RSA. Results show this by means of theoretical exposition and experimental evaluations on synthetic and real-world data. This serves as a promising direction of the way to boost secure cloud computations in the case of latency sensitive and resource constrained environments.