A methodical FHE-based cloud computing model

Abstract

Attacks such as Meldown and Spectre have shown that traditional cloud computing isolation mechanisms are not sufficient to guarantee the confidentiality of processed data. With Fully Homomorphic Encryption (FHE), data may be processed encrypted in the cloud, making any leaked information look random to an attacker. Furthermore, a client might also be interested in protecting the processing algorithm. While there has been research on ensuring the confidentiality of the processing algorithm, the resulting systems are impractical. Herein, we propose an automatic and methodical technique to approximate a wide range of functions homomorphically. As the approximations are all evaluated in the same manner, a homomorphic evaluator has no way to distinguish them. Since the derivation of the FHE circuit is decoupled from the function development process, users benefit from traditional programming and debugging tools. The proposed tools may exploit different kinds of number representations during the homomorphic evaluation of functions, namely stochastic number representations and fixed-point arithmetic, each with its own characteristics. Additionally, an implementation of the system is presented, its applicability is verified in practice for commonly used applications, including image processing and machine learning, and the two number representations are thoroughly compared.