Accelerating Encryption and Decryption with Linear Feedback: A Comprehensive Study

In today's digital landscape, data security is of paramount importance. With the increasing reliance on cloud computing, it has become crucial to ensure that sensitive information remains protected from unauthorized access. One popular scheme for achieving this is fully homomorphic encryption (FHE), which enables computation on encrypted data while ensuring data privacy. A significant challenge in FHE is accelerating the encryption and decryption processes, particularly when dealing with high-degree polynomial multiplication.

To address this issue, researchers have proposed various strategies to optimize these operations. Among them, the BFV scheme has gained popularity due to its ability to utilize ring polynomials for secure computation. However, the encryption, decryption, and evaluation processes in the BFV scheme require high-degree polynomial multiplication, which can be computationally intensive.

In this paper, we present a comprehensive design and implementation of a hardware architecture that accelerates encryption and decryption in the BFV scheme using linear feedback. Our proposed solution is capable of analyzing each company individually or in groups at any given time, thanks to its scalable design for higher degrees of polynomial multiplication.

The results of our numerical analysis, performed using MATLAB software, demonstrate the impact of our proposed strategy on company resources used during disaster recovery. Furthermore, our accelerator achieves significant performance improvements over existing methods, including up to 9× acceleration in encryption and 3.5× acceleration in decryption compared to Microsoft SEAL on an Intel Core i7 processor with 8GB memory.

In addition to its impressive speed, our proposed Gaussian PRNG exhibits better correlation (2×) compared to rotation-only-based PRNGs. Experimental results also show that our accelerator outperforms other methods by up to 22× in terms of clock cycles for polynomial degrees of 1024.

To implement the convolution operation, we employ a systolic array to calculate polynomial convolution followed by simple delayed subtraction to calculate polynomial modulo reduction within our accelerator's core. Our design uses a linear feedback approach for calculating polynomial multiplication, which enables efficient computation and minimizes errors.

In conclusion, this paper presents a groundbreaking hardware architecture that accelerates encryption and decryption in the BFV scheme using linear feedback. Our proposed solution has far-reaching implications for securing cloud computing and ensuring data privacy in critical applications where security is paramount.

Linear feedback