Skip to content

High-Precision Functional Bootstrapping for CKKS from Fourier Extension

Authors: Bian, Fu, Shen, Pan, Wang, Guan
Year: 2026
Venue: IACR ePrint 2026/367

type: scheme collection: catalogue, schemes status: auto

Links: ePrint 2026/367

Summary

Auto-extracted summary (unreviewed)

The paper introduces an amortized functional bootstrapping (FBS) framework for CKKS based on Fourier extension: by constructing bridge functions that make the periodically extended target function as smooth as possible, any C^k function on a bounded domain is approximated by a degree-n Fourier series with error O(n^(-k-2)), improving on the O(n^-1) bound of prior work. Implemented on OpenFHE, the method improves precision by 10-27 bits and reduces amortized FBS latency by 1.1-2x; plain bootstrapping reaches 31-bit precision at 0.61 ms amortized over 32,768 slots, and GeLU/Sigmoid/exp FBS reaches about 31-32 bits versus about 5 bits for the previous state of the art.

Abstract

We introduce a new (amortized) functional bootstrapping framework over the CKKS homomorphic encryption (HE) scheme based on Fourier extension. While approximating the modular reduction function in CKKS bootstrapping through Fourier series is a well-known technique, how such method can be efficiently generalized to functional bootstrapping is less understood. In this work, we show that, by constructing proper Fourier extensions, any function with a bounded domain in the smoothness class Cκ can be approximated by a degree-n Fourier series with errors of order O(n−κ−2) (except at the derivative singularity), improving on previous results on a global error bound of O(n−1) [AKP25]. To achieve such bound, we propose a new way of constructing Fourier extensions, such that the extended functions appear as smooth as possible in the sense of a Fourier approximation. By implementing our functional bootstrapping over OpenFHE, we demonstrate that we can improve the data precision by 10–27 bits and reduce the amortized FBS latency by 1.1×–2× over a variety of benchmarking functions.
Source: pdf

Extracted details

Field Value Source Confidence
contribution Fourier-extension-based functional bootstrapping framework for CKKS llm 0.90
contribution new construction of Fourier extensions (bridge functions) that maximizes smoothness, with O(n^(-k-2)) error bound llm 0.90
contribution OpenFHE implementation and comparison against state-of-the-art bootstrapping and FBS methods llm 0.90
contribution open-sourced implementation llm 0.90
schemes CKKS llm 0.90
libraries OpenFHE llm 0.90
tasks CKKS bootstrapping and functional bootstrapping llm 0.90
tasks homomorphic evaluation of GeLU, Sigmoid, exp, floor and S-box lookup tables llm 0.90
hardware Intel Core i9-14900KF CPU (3.20 GHz), 128 GB RAM, single-threaded llm 0.90
security level 128-bit llm 0.90
interaction non-interactive FHE llm 0.90
code url https://github.com/fyhssgss/FEFBS.1.2 (URL found in the full text) regex 0.50
code url https://github.com/fyhssgss/FEFBS llm 0.90
limitations precision of discrete/LUT functions is limited by input quantization (log2 of LUT size) llm 0.90
limitations stronger IND-CPA-D security (2^-128 failure rate) requires parameter adaptation (supplementary material) llm 0.90
limitations functions without matching boundary derivatives use delta=1/2, trading one bit of plaintext space for smoothness llm 0.90
llm relevance medium llm 0.90
key result amortized bootstrapping latency: 0.61 ms at 31.22-bit precision (20.14 s total) (32,768 slots, N=65536, OpenFHE v1.4.0, single thread) llm 0.90
key result bootstrapping precision and latency vs prior work: +11-21 bits precision, 1.4-2x lower amortized latency (vs BMTPH21, BKSS25, CHKS25) llm 0.90
key result GeLU FBS precision: 32.04 bits vs 5.46 bits (AKP25), 0.6724 ms amortized (32,768 slots) llm 0.90
key result Sigmoid FBS precision: 31.34 bits vs 5.09 bits (AKP25), 0.6221 ms amortized (32,768 slots) llm 0.90
key result overall FBS improvement: +10-27 bits precision, 1.1-2x lower amortized latency (across benchmark functions) llm 0.90
key result 32-bit FBS total latency: 22.1 s for 32,768 slots (0.67 ms amortized) (conclusion) llm 0.90

System card: Fourier-extension functional bootstrapping

Group 2. CKKS core and variants
Family CKKS functional bootstrapping
Arithmetic LUTs / smooth functions
Interaction Non-interactive
Model scale -
Status Active SOTA
Relation to the SoK Evaluates smooth functions during bootstrapping with better error bounds.
Note Improves on general CKKS functional bootstrapping for piecewise non-linear layers.
Systems page Fourier-extension functional bootstrapping

Keyword mentions

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

  • scheme: CKKS 64 [p.1], BFV 12 [p.1], TFHE 9 [p.1], BGV 4 [p.1]
  • library: OpenFHE 7 [p.1]
  • technique: Bootstrapping 84 [p.1], Lookup table 50 [p.1], Functional bootstrapping 20 [p.1], GELU 14 [p.4], Chebyshev 10 [p.2], Sigmoid 9 [p.4], RNS 7 [p.8], ReLU 6 [p.5], Polynomial approximation 5 [p.22], SIMD packing 5 [p.2], Quantization 5 [p.5], Minimax approximation 2 [p.9], Rotations 2 [p.2]
  • assumption: Lattice-based 3 [p.30], IND-CPA-D 2 [p.25]
  • topic: Homomorphic encryption 14 [p.1], Open source 3 [p.4], Benchmark 2 [p.1]

Files

Path (under papers/) Collection Pages Size SHA-256 Note
references/catalogue/2-131_high-precision-functional-bootstrapping-for-ckks-from-fourie.pdf catalogue 43 988.0 KB ee2384e0a814
references/schemes/s13-1_eprint-2026-367.pdf schemes 43 988.0 KB ee2384e0a814 byte-identical to references/catalogue/2-131_high-precision-functional-bootstrapping-for-ckks-from-fourie.pdf

Database key: high-precision-functional-bootstrapping-for-ckks-from-fourier-extension · created 2026-11-05 · updated 2026-11-05 · edit with python3 tools/paperdb.py set high-precision-functional-bootstrapping-for-ckks-from-fourier-extension COLUMN VALUE