A Heterogeneous Computing Framework for Accelerating Fully Homomorphic Encryption

In the digital age, privacy is increasingly important. The General Data Protection Regulation (GDPR) [22] lays out rules for how data is collected and protected, posing challenges for many organizations. While there are several privacy preserving techniques, fully homomorphic encryption (FHE) stands out as the most mathematically secure. FHE enables computations on encrypted data without the need for decryption. However, it introduces significant computational overhead compared to non-encrypted computations. In this paper, we introduce a heterogeneous computing framework designed to accelerate fully homomorphic encryption. This framework encompasses homomorphic evaluations on multiple platforms including CPU, GPU, and FPGA, paired with a tailored task scheduling algorithm. Each platform is equipped with comprehensive FHE functionalities and employs state-of-the-art implementations, allowing for standalone evaluation of FHE applications. The task scheduling algorithm strategically divides the computational tasks across the heterogeneous system, taking into account data transfer times to optimize application performance. Results show that the system reduces the latency effectively with additional computational platforms and provides more flexibility and accessibility of FHE for contemporary applications.