FIDESlib: A Fully-Fledged Open-Source FHE Library for Efficient CKKS on GPUs¶
Authors: Carlos Agulló-Domingo; Óscar Vera-López; Seyda Guzelhan; Lohit Daksha; Aymane El Jerari; Kaustubh Shivdikar; Rashmi Agrawal; David Kaeli; Ajay Joshi; José L. Abellán
Year: 2025
Venue: arXiv 2025
type: library collection: catalogue, sok-references status: auto
Links: arXiv:2507.04775
Summary¶
Auto-extracted summary (unreviewed)
FIDESlib is an open-source GPU library that implements all server-side CKKS operations, including bootstrapping, while remaining fully interoperable with OpenFHE for client-side key generation, encryption and decryption. The authors report that it is the first open-source GPU implementation with full CKKS functionality, that it beats the leading open-source GPU library Phantom in performance and scalability across several GPU platforms, and that its bootstrapping is at least 70x faster than the AVX-optimized OpenFHE CPU implementation.
Abstract¶
Word-wise Fully Homomorphic Encryption (FHE) schemes, such as CKKS, are gaining significant traction due to their ability to provide post-quantum-resistant, privacy-preserving approximate computing; an especially desirable feature in Machine-Learning-as-a-Service (MLaaS) cloud-computing paradigms. OpenFHE is a leading CPU-based FHE library with robust CKKS operations, but its server-side performance is not yet sufficient for practical cloud deployment. As GPU computing becomes more common in data centers, many FHE libraries are adding GPU support. However, integrating an efficient GPU backend into OpenFHE is challenging. While OpenFHE uses a Hardware Abstraction Layer (HAL), its flexible architecture sacrifices performance due to the abstraction layers required for multi-scheme and multi-backend compatibility. In this work, we introduce FIDESlib, the first open-source server-side CKKS GPU library that is fully interoperable with well-established client-side OpenFHE operations. Unlike other existing open-source GPU libraries, FIDESlib provides the first implementation featuring heavily optimized GPU kernels for all CKKS primitives, including bootstrapping. Our library also integrates robust benchmarking and testing, ensuring it remains adaptable to further optimization. Furthermore, its software architecture is designed to support extensions to a multi-GPU backend for enhanced acceleration. Our experiments across various GPU systems and the leading open-source CKKS library to date, Phantom, show that FIDESlib offers superior performance and scalability. For bootstrapping, FIDESlib achieves no less than 70x speedup over the AVX-optimized OpenFHE implementation.
Source: arxiv
Extracted details¶
| Field | Value | Source | Confidence |
|---|---|---|---|
| contribution | First open-source, feature-complete server-side CKKS GPU library including bootstrapping | llm | 0.85 |
| contribution | Full interoperability with client-side OpenFHE operations | llm | 0.85 |
| contribution | Heavily optimized GPU kernels for all CKKS primitives plus specialized routines (ScalarAdd, ScalarMult, HSquare) | llm | 0.85 |
| contribution | Integrated benchmarking and testing infrastructure | llm | 0.85 |
| contribution | Software architecture designed for future multi-GPU extension | llm | 0.85 |
| contribution | Evaluation across four GPU platforms against Phantom and OpenFHE | llm | 0.85 |
| schemes | CKKS | llm | 0.85 |
| libraries | OpenFHE | llm | 0.85 |
| libraries | Phantom | llm | 0.85 |
| libraries | CUDA | llm | 0.85 |
| tasks | privacy-preserving approximate computing | llm | 0.85 |
| tasks | MLaaS cloud inference (motivation) | llm | 0.85 |
| hardware | NVIDIA RTX 4090 GPU | llm | 0.85 |
| hardware | four distinct GPU platforms (Table IV) | llm | 0.85 |
| threat model | client-server outsourced computation (client keeps keys, server computes on ciphertexts) | llm | 0.85 |
| interaction | non-interactive FHE | llm | 0.85 |
| code url | https://github.com/lightbulb128/troy-nova (URL found in the full text) | regex | 0.50 |
| code url | https://github.com/Desilo/liberate-fhe (URL found in the full text) | regex | 0.50 |
| code url | https://github.com/FHE-org/fhe-org.github.io/files/14896471/07-Kim.pdf (URL found in the full text) | regex | 0.50 |
| limitations | Multi-GPU backend is left as a future extension of the architecture | llm | 0.85 |
| limitations | GPUs lack efficient hardware support for modular arithmetic, which bounds CKKS kernel performance | llm | 0.85 |
| llm relevance | low | llm | 0.85 |
| key result | bootstrapping speed-up: at least 70x (FIDESlib on GPU vs AVX-optimized OpenFHE on CPU) | llm | 0.85 |
| key result | comparison with Phantom: superior performance and scalability (various GPU systems, leading open-source CKKS GPU library) | llm | 0.85 |
System card: FIDESlib¶
| Group | 7. FHE libraries, GPU back-ends and compilers |
| Family | CKKS with bootstrapping |
| Arithmetic | CUDA C++ |
| Interaction | GPU library |
| Model scale | Used by ELLMO, POLARIS |
| Status | Active SOTA |
| Relation to the SoK | OpenFHE-compatible GPU CKKS server library. |
| Note | Full GPU bootstrapping. Interoperates with OpenFHE client keys. |
| Code | https://github.com/CAPS-UMU/FIDESlib |
| Systems page | FIDESlib |
SoK reference [23]¶
C. Agull´o-Domingo, ´O. Vera-L´opez, S. Guzelhan, L. Daksha, A. El Jerari, K. Shivdikar, R. Agrawal, D. Kaeli, A. Joshi, and J. L. Abell´an, “Fideslib: A fully-fledged open-source fhe library for efficient ckks on gpus,” in 2025 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS). IEEE, 2025, pp. 1–3.
Listed on the reference page of the site.
Keyword mentions¶
Counts of vocabulary terms in the full text (first page of occurrence in brackets).
- scheme: CKKS 69 [p.1], BFV 2 [p.2], BGV 2 [p.2], RSA 2 [p.11]
- library: FIDESlib 75 [p.1], OpenFHE 54 [p.1], Phantom 14 [p.1], HEaaN (library) 3 [p.9], Liberate.FHE 3 [p.9], HEonGPU 2 [p.9], Cheddar 2 [p.9], Microsoft SEAL 1 [p.2], HElib 1 [p.2]
- model: Logistic regression 2 [p.9]
- hardware: GPU 81 [p.1], NVIDIA RTX 15 [p.3], Multi-GPU 7 [p.1], NVIDIA V100 6 [p.7], CUDA 3 [p.4], AVX-512 2 [p.7], FPGA 1 [p.1], ASIC 1 [p.1]
- technique: NTT 30 [p.2], Bootstrapping 29 [p.1], Rescaling 12 [p.2], Key switching 7 [p.2], Rotations 7 [p.2], RNS 6 [p.2]
- assumption: Lattice-based 5 [p.1], RLWE 2 [p.1], Post-quantum 2 [p.1]
- topic: Homomorphic encryption 46 [p.1], Open source 19 [p.1], Benchmark 12 [p.1]
Files¶
| Path (under papers/) | Collection | Pages | Size | SHA-256 | Note |
|---|---|---|---|---|---|
references/catalogue/7-098_fideslib-a-fully-fledged-open-source-fhe-library-for-efficie.pdf |
catalogue | 11 | 1.9 MB | 645de1f914d0 |
|
references/sok-references/023_fideslib-a-fully-fledged-open-source-fhe-library-for-efficie.pdf |
sok-references | 11 | 1.9 MB | 645de1f914d0 |
byte-identical to references/catalogue/7-098_fideslib-a-fully-fledged-open-source-fhe-library-for-efficie.pdf |
Database key: fideslib-a-fully-fledged-open-source-fhe-library-for-efficient-ckks-on · created 2026-11-05 · updated 2026-11-05 · edit with python3 tools/paperdb.py set fideslib-a-fully-fledged-open-source-fhe-library-for-efficient-ckks-on COLUMN VALUE