Publication:
GPU Acceleration of Sparse Fully Homomorphic Encrypted DNNs

Simple item page
dc.contributor.authorD'Agata, Lara
dc.contributor.authorAgulló-Domingo, Carlos
dc.contributor.authorVera-López, Óscar
dc.contributor.authorShivdikar, Kaustubh
dc.contributor.authorYudha, Ardhi W. B.
dc.contributor.authorYaman, Ferhat
dc.contributor.authorKaeli, David
dc.contributor.authorAbellán Miguel, José Luis
dc.contributor.authorColbert, Ian
dc.contributor.authorCano, José
dc.contributor.departmentIngeniería y Tecnología de Computadores
dc.contributor.otherFacultades de la UMU::Facultad de Informática
dc.date.accessioned2026-04-17T07:53:31Z
dc.date.available2026-04-17T07:53:31Z
dc.date.copyright© 2026 Copyright held by the owner/author(s).
dc.date.issued2026-04-13
dc.description.abstractFully homomorphic encryption (FHE) has recently attracted significant attention as both a cryptographic primitive and a systems challenge. Given the latest advances in accelerated computing, FHE presents a promising opportunity for progress, with applications ranging from machine learning to information security. We target the most computationally intensive operation in deep neural networks from a hardware perspective, matrix multiplication (matmul), and adapt it for execution on AMD GPUs. We propose a new optimized method that improves the runtime and complexity of ciphertext matmul by using FIDESlib, a recent open-source FHE library designed specifically for GPUs. By exploiting sparsity in both operands, our sparse matmul implementation outperforms its CPU counterpart by up to 3.0\times and reduces the time complexity from cubic to semi-linear, demonstrating an improvement over existing FHE matmul implementations.
dc.formatapplication/pdf
dc.format.extent8
dc.identifier.citationarXiv:2604.11659v1
dc.identifier.doihttps://doi.org/10.48550/arXiv.2604.11659
dc.identifier.urihttp://hdl.handle.net/10201/226522
dc.languageeng
dc.relationThis work was supported in part by Advanced Micro Devices, Inc., and the AMD AI & HPC Cluster Program. It was also funded through the grant CNS2023-144241 from MICIU/ AEI/10.13039/501100011033 and by the European Union NextGenerationEU/PRTR. Additionally, this research was conducted within the context of the grant RYC2021-031966-I funded by MICIU/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/PRTR.
dc.relation.ispartofThe 6th Workshop on Machine Learning and Systems (EuroMLSys ’26), April 27–30, 2026, Edinburgh, Scotland Uk. ACM, New York, NY, USA, 8 pages.
dc.relation.publisherversionhttps://arxiv.org/abs/2604.11659v1
dc.rightsAttribution-NonCommercial-ShareAlike 4.0 International*
dc.rights.accessRightsinfo:eu-repo/semantics/openAccess
dc.rights.urihttp://creativecommons.org/licenses/by-nc-sa/4.0/*
dc.subjectDNN acceleration
dc.subjectSecure computation
dc.subjectSparse matrix multiplication
dc.subjectFully homomorphic encryption
dc.subject.odsNo relacionado con ningún objetivo de desarrollo sostenible
dc.titleGPU Acceleration of Sparse Fully Homomorphic Encrypted DNNs
dc.typeinfo:eu-repo/semantics/article
dc.type.versioninfo:eu-repo/semantics/acceptedVersion
dspace.entity.typePublicationes
relation.isAuthorOfPublication0408fa9f-b61e-48cb-a08f-ae8999f885ee
relation.isAuthorOfPublication.latestForDiscovery0408fa9f-b61e-48cb-a08f-ae8999f885ee
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
FHE___EuroMLSys_2026.pdf
Size:
1.24 MB
Format:
Adobe Portable Document Format
Description:
Download
License bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
license.txt
Size:
1.37 KB
Format:
Item-specific license agreed upon to submission
Description:
Download
Collections
Artículos