Abstract:Homomorphic encryption (HE) allows computations to be directly carried out on ciphertexts and enables privacy-preserving cloud computing. The computations on the coefficients of the polynomials involved in HE are always followed by modular reduction, and the overall complexity of ciphertext multiplication can be reduced by utilizing the quotient. Our previous design considers the cases that the dividend is an integer multiple of the modulus and the modulus is in the format of $2^w-2^u\pm1$, where $u<w/2$. In this paper, the division is generalized for larger $u$ and dividend not an integer multiple of the modulus. An algorithm is proposed to compute the quotient and vigorous mathematical proofs are provided. Moreover, efficient hardware architecture is developed for implementing the proposed algorithm. Compared to alternative division approaches that utilize the inverse of the divisor, for $w=32$, the proposed design achieves at least 9% shorter latency and 79\% area reduction for 75% possible values of $u$.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the problem of low - complexity hardware implementation of integer division in Homomorphic Encryption (HE). Specifically, the paper proposes a general low - complexity integer division algorithm and its hardware architecture to address the following challenges: 1. **The case where the modulus has more non - zero bits**: - Previous designs were only applicable to the form of modulus $ q $ as $ 2^w - 2^u\pm1 $, and required $ u < w/2 $. This paper extends the design to the case where $ u\geq w/2 $. 2. **The case where the dividend is not an integer multiple of the modulus**: - Previous designs assumed that the dividend was an integer multiple of the modulus, which required an additional remainder calculation step. This paper proposes a method that does not require this assumption and directly deals with the case where the dividend is not an integer multiple of the modulus. 3. **Reducing hardware complexity and latency**: - The algorithm proposed in the paper is more efficient in hardware implementation than the existing division methods based on multiplicative inverses, reducing the area occupation and shortening the latency. #### Specific problem description - **Modulus reduction in homomorphic encryption**: Homomorphic encryption involves calculating the coefficients of polynomials, and these calculations usually require modulus reduction. The efficiency of modulus reduction directly affects the overall performance of homomorphic encryption. - **Limitations of existing methods**: - Existing methods such as [5] assume that the modulus $ q = 2^w - 2^u\pm1 $, and $ u < w/2 $, and the dividend is an integer multiple of the modulus. These assumptions limit their application range. - Other methods such as [9] and [10] use multiplicative inverses to implement division, resulting in high hardware implementation complexity and long latency. #### Solution - **General low - complexity integer division algorithm**: - An iterative algorithm is proposed to calculate the quotient through simple addition and shift operations, which is applicable to the case where $ u\geq w/2 $. - For the case where the dividend is not an integer multiple of the modulus, the influence of the remainder is considered by adjusting the quotient obtained during the iterative process. - **Efficient hardware architecture**: - An efficient hardware architecture is developed to implement the above algorithm, reducing the area occupation and shortening the latency. - Compared with existing methods, in the case of $ w = 32 $, for 75% of the possible $ u $ values, the latency of the new design is reduced by at least 9% and the area is reduced by at least 79%. ### Summary By proposing a general low - complexity integer division algorithm and its hardware architecture, this paper solves the efficiency problem of modulus reduction in homomorphic encryption, especially for the cases where the modulus has more non - zero bits and the dividend is not an integer multiple of the modulus. This not only expands the application range of existing designs but also significantly improves the efficiency of hardware implementation.

Low-Complexity Integer Divider Architecture for Homomorphic Encryption

Efficient integer division computation protocols based on partial homomorphic encryption

A lightweight protocol for privacy preserving division

A new lightweight additive homomorphic encryption algorithm

PaReNTT: Low-Latency Parallel Residue Number System and NTT-Based Long Polynomial Modular Multiplication for Homomorphic Encryption

Three-Input Ciphertext Multiplication for Homomorphic Encryption

Addition Circuit Optimization Using Carry-Lookahead and SIMD for Homomorphic Encryption

Coupling bit and modular arithmetic for efficient general-purpose fully homomorphic encryption

Area-Efficient Number Theoretic Transform Architecture for Homomorphic Encryption

Efficient Comparison and Addition for FHE with Weighted Computational Complexity Model

An Area-Efficient and Configurable Number Theoretic Transform Accelerator for Homomorphic Encryption

NTTU: an Area-Efficient Low-Power NTT-Uncoupled Architecture for NTT-Based Multiplication

Heterogeneous Reconfigurable Accelerator for Homomorphic Evaluation on Encrypted Data

New Algorithms and Architecture for a Scalable Low-cost Dual-fields Modular Multiplier/Divider

CiFlow: Dataflow Analysis and Optimization of Key Switching for Homomorphic Encryption

Efficient long division via Montgomery multiply

Fully Homomorphic Encryption Scheme Over Integers Based on DGHV Scheme

High-speed constant-time division module for Elliptic Curve Cryptography based on GF(2m)

A New Systolic Architecture for Modular Division

Fully Homomorphic Encryption over the Integers

A 28nm 68MOPS 0.18\mu\mathrm{j}/\text{op}$ Paillier Homomorphic Encryption Processor with Bit-Serial Sparse Ciphertext Computing