Efficient Keyset Design for Neural Networks Using Homomorphic Encryption

With the advent of the Internet of Things (IoT), large volumes of sensitive data are produced from IoT devices, driving the adoption of Machine Learning as a Service (MLaaS) to overcome their limited computational resources. However, as privacy concerns in MLaaS grow, the demand for Privacy-Preservi...

Full description

Saved in:
Bibliographic Details
Main Authors: Youyeon Joo, Seungjin Ha, Hyunyoung Oh, Yunheung Paek
Format: Article
Language:English
Published: MDPI AG 2025-07-01
Series:Sensors
Subjects:
Online Access:https://www.mdpi.com/1424-8220/25/14/4320
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850071294608932864
author Youyeon Joo
Seungjin Ha
Hyunyoung Oh
Yunheung Paek
author_facet Youyeon Joo
Seungjin Ha
Hyunyoung Oh
Yunheung Paek
author_sort Youyeon Joo
collection DOAJ
description With the advent of the Internet of Things (IoT), large volumes of sensitive data are produced from IoT devices, driving the adoption of Machine Learning as a Service (MLaaS) to overcome their limited computational resources. However, as privacy concerns in MLaaS grow, the demand for Privacy-Preserving Machine Learning (PPML) has increased. Fully Homomorphic Encryption (FHE) offers a promising solution by enabling computations on encrypted data without exposing the raw data. However, FHE-based neural network inference suffers from substantial overhead due to expensive primitive operations, such as ciphertext rotation and bootstrapping. While previous research has primarily focused on optimizing the efficiency of these computations, our work takes a different approach by concentrating on the rotation keyset design, a pre-generated data structure prepared before execution. We systematically explore three key design spaces (KDS) that influence rotation keyset design and propose an optimized keyset that reduces both computational overhead and memory consumption. To demonstrate the effectiveness of our new KDS design, we present two case studies that achieve up to 11.29× memory reduction and 1.67–2.55× speedup, highlighting the benefits of our optimized keyset.
format Article
id doaj-art-01f34bd4de4446c1a8a88dcb854428e4
institution DOAJ
issn 1424-8220
language English
publishDate 2025-07-01
publisher MDPI AG
record_format Article
series Sensors
spelling doaj-art-01f34bd4de4446c1a8a88dcb854428e42025-08-20T02:47:21ZengMDPI AGSensors1424-82202025-07-012514432010.3390/s25144320Efficient Keyset Design for Neural Networks Using Homomorphic EncryptionYouyeon Joo0Seungjin Ha1Hyunyoung Oh2Yunheung Paek3Department of Electrical and Computer Engineering & ISRC, Seoul National University, Seoul 08826, Republic of KoreaDepartment of Electrical and Computer Engineering & ISRC, Seoul National University, Seoul 08826, Republic of KoreaDepartment of AI, Gachon University, Seongnam-si 13120, Republic of KoreaDepartment of Electrical and Computer Engineering & ISRC, Seoul National University, Seoul 08826, Republic of KoreaWith the advent of the Internet of Things (IoT), large volumes of sensitive data are produced from IoT devices, driving the adoption of Machine Learning as a Service (MLaaS) to overcome their limited computational resources. However, as privacy concerns in MLaaS grow, the demand for Privacy-Preserving Machine Learning (PPML) has increased. Fully Homomorphic Encryption (FHE) offers a promising solution by enabling computations on encrypted data without exposing the raw data. However, FHE-based neural network inference suffers from substantial overhead due to expensive primitive operations, such as ciphertext rotation and bootstrapping. While previous research has primarily focused on optimizing the efficiency of these computations, our work takes a different approach by concentrating on the rotation keyset design, a pre-generated data structure prepared before execution. We systematically explore three key design spaces (KDS) that influence rotation keyset design and propose an optimized keyset that reduces both computational overhead and memory consumption. To demonstrate the effectiveness of our new KDS design, we present two case studies that achieve up to 11.29× memory reduction and 1.67–2.55× speedup, highlighting the benefits of our optimized keyset.https://www.mdpi.com/1424-8220/25/14/4320machine learning as a serviceprivacy-preserving machine learningneural networksfully homomorphic encryptionCKKSrotation keyset
spellingShingle Youyeon Joo
Seungjin Ha
Hyunyoung Oh
Yunheung Paek
Efficient Keyset Design for Neural Networks Using Homomorphic Encryption
Sensors
machine learning as a service
privacy-preserving machine learning
neural networks
fully homomorphic encryption
CKKS
rotation keyset
title Efficient Keyset Design for Neural Networks Using Homomorphic Encryption
title_full Efficient Keyset Design for Neural Networks Using Homomorphic Encryption
title_fullStr Efficient Keyset Design for Neural Networks Using Homomorphic Encryption
title_full_unstemmed Efficient Keyset Design for Neural Networks Using Homomorphic Encryption
title_short Efficient Keyset Design for Neural Networks Using Homomorphic Encryption
title_sort efficient keyset design for neural networks using homomorphic encryption
topic machine learning as a service
privacy-preserving machine learning
neural networks
fully homomorphic encryption
CKKS
rotation keyset
url https://www.mdpi.com/1424-8220/25/14/4320
work_keys_str_mv AT youyeonjoo efficientkeysetdesignforneuralnetworksusinghomomorphicencryption
AT seungjinha efficientkeysetdesignforneuralnetworksusinghomomorphicencryption
AT hyunyoungoh efficientkeysetdesignforneuralnetworksusinghomomorphicencryption
AT yunheungpaek efficientkeysetdesignforneuralnetworksusinghomomorphicencryption