Skip to content

ModHE: Modular Homomorphic Encryption Using Module Lattices: Potentials and Limitations

Authors: Mukherjee, Aikata, Mert, Lee, Kwon, Deryabin, Roy
Year: 2024
Venue: TCHES 2024

type: scheme collection: schemes status: auto

Links: ePrint 2023/895

Visual lesson: Module-LWE HE (ModHE) · Scheme table row: Module-LWE HE (ModHE)

Summary

Auto-extracted summary (unreviewed)

ModHE is a homomorphic encryption scheme built on Module-LWE, presented concretely as a module-lattice port of CKKS (with an RNS variant, ModRNS) in which a larger multiplicative depth is obtained by increasing the module rank rather than the polynomial degree, so a fixed-size hardware unit can be reused and parallelized. The authors synthesize an FPGA accelerator on a Xilinx Alveo U280 (8 RPAU units at 210 MHz) and estimate 18 ms for a degree-9 logistic-regression inference versus 1.8 s in SEAL on a CPU, while candidly reporting that ModRNS carries an (r+1)/2 ciphertext expansion, about 1.7x more cycles than a ring-based design at equal security, and reduced precision.

Abstract

The promising field of homomorphic encryption enables functions to be evaluated on encrypted data and produce results for the same computations done on plaintexts. It, therefore, comes as no surprise that many ventures at constructing homomorphic encryption schemes have come into the limelight in recent years. Most popular are those that rely on the hard lattice problem, called the Ring Learning with Errors problem (RLWE). One major limitation of these homomorphic encryption schemes is that in order to securely increase the maximum multiplicative depth, they need to increase the polynomial-size (degree of the polynomial ring) thereby also increasing the complexity of the design. We aim to bridge this gap by proposing a homomorphic encryption (HE) scheme based on the Module Learning with Errors problem (MLWE), ModHE that allows us to break the big computations into smaller ones. Given the popularity of module lattice-based post-quantum schemes, it is an evidently interesting research endeavor to also formulate module lattice-based homomorphic encryption schemes. While our proposed scheme is general, as a case study, we port the well-known RLWE-based CKKS scheme to the MLWE setting. The module version of the scheme completely stops the polynomial-size blowups when aiming for a greater circuit depth. Additionally, it presents greater opportunities for designing flexible, reusable, and parallelizable hardware architecture. A hardware implementation is provided to support our claims. We also acknowledge that as we try to decrease the complexity of computations, the amount of computations (such as relinearizations) increases. We hope that the potential and limitations of using such a hardware-friendly scheme will spark further research.
Source: pdf

Extracted details

Field Value Source Confidence
contribution ModHE: MLWE-based homomorphic encryption scheme, instantiated as a module version of CKKS llm 0.85
contribution RNS variant (ModRNS) of the module scheme llm 0.85
contribution rank-reduction property allowing the module rank to be managed dynamically llm 0.85
contribution flexible, reusable and parallelizable FPGA hardware architecture with implementation results llm 0.85
contribution analysis of potentials and limitations (ciphertext expansion, relinearization overhead, precision loss) llm 0.85
schemes CKKS llm 0.85
schemes ModHE (MLWE-based CKKS) llm 0.85
libraries SEAL llm 0.85
models logistic regression llm 0.85
tasks logistic regression inference llm 0.85
hardware Xilinx Alveo U280 FPGA llm 0.85
hardware Intel Core i5-10210U CPU llm 0.85
security level 128-bit llm 0.85
interaction non-interactive FHE llm 0.85
code url https://github.com/anishamukh/MODHE (URL found in the full text) regex 0.50
code url https://github.com/anishamukh/MODHE llm 0.85
limitations packing stays fixed (2^13 slots) as module rank grows, giving an (r+1)/2 ciphertext expansion rate llm 0.85
limitations (r+1)/2x more computation than ring-based RNS; roughly 1.7x more clock cycles for r=2 and about 160x timing for r=4 due to many ring-level relinearizations llm 0.85
limitations increased load on memory and number of computations (relinearizations) llm 0.85
limitations precision loss: 34-bit key-switch precision versus 40-bit, plus 3-bit loss from rank reduction llm 0.85
limitations performance figures are estimates based on HBM bandwidth specifications; HBM-processor interface not prototyped llm 0.85
limitations efficient bootstrapping for ModHE left as future work llm 0.85
llm relevance low llm 0.85
key result FPGA resource usage and frequency: 699,513 LUTs, 513,720 FFs, 3,200 DSPs, 1,034 BRAMs, 305 URAMs at 210 MHz (design with 8 RPAU units on Xilinx Alveo U280 (excluding HBM controller)) llm 0.85
key result logistic regression inference latency (estimated): 18 ms (3.8M clock cycles) (three-accelerator FPGA setting at 210 MHz, degree-9 Taylor approximation) llm 0.85
key result software baseline latency: 1.8 s (SEAL, N=2^15, Intel Core i5-10210U at 1.60 GHz) llm 0.85
key result cycle overhead vs ring-based accelerator: about 1.7x more clock cycles; 2.8x slower on one accelerator (ModRNS (r=2, N=2^14) vs [MAK+23] (N=2^15), same depth at 128-bit security) llm 0.85
key result ciphertext expansion and slowdown: 1.5x (r=2 compared with RingRNS) llm 0.85
key result key-switch precision: 34-bit (about 30-bit after rank reduction) vs 40-bit for RingRNS (ModRNS) llm 0.85

Scheme card: Module-LWE HE (ModHE)

Group 2. Exact word-wise FHE (BGV / BFV family)
Kind Scheme
Hardness assumption Module-LWE
Plaintext type Exact or approximate (BGV/BFV/CKKS-style)
SIMD packing Yes
Bootstrapping Yes
Post-quantum Yes
Status Niche
Fit for LLM inference Module lattices allow finer parameter choice than power-of-two rings.
Follow-ups Revisiting module-lattice HE (CIC 2025, 2025/1218).
Follow-up papers in the archive Revisiting Module Lattice-based Homomorphic Encryption and Application to Secure-MPC

Keyword mentions

Counts of vocabulary terms in the full text (first page of occurrence in brackets).

  • scheme: CKKS 31 [p.1], Leveled / somewhat HE 7 [p.2], BGV 5 [p.2], Hybrid HE / transciphering 5 [p.20], BFV 4 [p.2], ElGamal 2 [p.2], RSA 2 [p.2]
  • library: Microsoft SEAL 1 [p.27]
  • model: Logistic regression 3 [p.3]
  • hardware: FPGA 15 [p.2], ASIC 2 [p.2], GPU 1 [p.2]
  • technique: Relinearization 54 [p.1], RNS 36 [p.2], NTT 32 [p.4], Key switching 18 [p.6], Bootstrapping 13 [p.2], Rescaling 13 [p.8], Rotations 9 [p.8], Polynomial approximation 2 [p.27], Lookup table 2 [p.23]
  • assumption: Module-LWE 51 [p.1], Lattice-based 51 [p.1], RLWE 49 [p.1], LWE 23 [p.2], Post-quantum 5 [p.1], 128-bit security 3 [p.12]
  • task: Language modeling / generation 4 [p.9], Medical / healthcare 3 [p.1]
  • topic: Homomorphic encryption 72 [p.1], Benchmark 2 [p.26], Open source 2 [p.3]

Files

Path (under papers/) Collection Pages Size SHA-256 Note
references/schemes/s19-0_modhe-modular-homomorphic-encryption-using-module-lattices-p.pdf schemes 36 943.5 KB d80f787cf48f

Database key: modhe-modular-homomorphic-encryption-using-module-lattices-potentials · created 2026-11-05 · updated 2026-11-05 · edit with python3 tools/paperdb.py set modhe-modular-homomorphic-encryption-using-module-lattices-potentials COLUMN VALUE