A Homomorphic Encryption Framework for Privacy-Preserving Spiking Neural Networks
Machine learning (ML) is widely used today, especially through deep neural networks (DNNs); however, increasing computational load and resource requirements have led to cloud-based solutions. To address this problem, a new generation of networks has emerged called spiking neural networks (SNNs), whi...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-10-01
|
| Series: | Information |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2078-2489/14/10/537 |
| _version_ | 1797573574778683392 |
|---|---|
| author | Farzad Nikfam Raffaele Casaburi Alberto Marchisio Maurizio Martina Muhammad Shafique |
| author_facet | Farzad Nikfam Raffaele Casaburi Alberto Marchisio Maurizio Martina Muhammad Shafique |
| author_sort | Farzad Nikfam |
| collection | DOAJ |
| description | Machine learning (ML) is widely used today, especially through deep neural networks (DNNs); however, increasing computational load and resource requirements have led to cloud-based solutions. To address this problem, a new generation of networks has emerged called spiking neural networks (SNNs), which mimic the behavior of the human brain to improve efficiency and reduce energy consumption. These networks often process large amounts of sensitive information, such as confidential data, and thus privacy issues arise. Homomorphic encryption (HE) offers a solution, allowing calculations to be performed on encrypted data without decrypting them. This research compares traditional DNNs and SNNs using the Brakerski/Fan-Vercauteren (BFV) encryption scheme. The LeNet-5 and AlexNet models, widely-used convolutional architectures, are used for both DNN and SNN models based on their respective architectures, and the networks are trained and compared using the FashionMNIST dataset. The results show that SNNs using HE achieve up to 40% higher accuracy than DNNs for low values of the plaintext modulus <i>t</i>, although their execution time is longer due to their time-coding nature with multiple time steps. |
| first_indexed | 2024-03-10T21:11:01Z |
| format | Article |
| id | doaj.art-a3e3270e0816452c9b07cb05ead6767a |
| institution | Directory Open Access Journal |
| issn | 2078-2489 |
| language | English |
| last_indexed | 2024-03-10T21:11:01Z |
| publishDate | 2023-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Information |
| spelling | doaj.art-a3e3270e0816452c9b07cb05ead6767a2023-11-19T16:47:52ZengMDPI AGInformation2078-24892023-10-01141053710.3390/info14100537A Homomorphic Encryption Framework for Privacy-Preserving Spiking Neural NetworksFarzad Nikfam0Raffaele Casaburi1Alberto Marchisio2Maurizio Martina3Muhammad Shafique4Department of Electrical, Electronics and Telecommunication Engineering, Politecnico di Torino, 10129 Torino, ItalyDepartment of Electrical, Electronics and Telecommunication Engineering, Politecnico di Torino, 10129 Torino, ItalyeBrain Lab, Division of Engineering, New York University, Abu Dhabi P.O. Box 129188, United Arab EmiratesDepartment of Electrical, Electronics and Telecommunication Engineering, Politecnico di Torino, 10129 Torino, ItalyeBrain Lab, Division of Engineering, New York University, Abu Dhabi P.O. Box 129188, United Arab EmiratesMachine learning (ML) is widely used today, especially through deep neural networks (DNNs); however, increasing computational load and resource requirements have led to cloud-based solutions. To address this problem, a new generation of networks has emerged called spiking neural networks (SNNs), which mimic the behavior of the human brain to improve efficiency and reduce energy consumption. These networks often process large amounts of sensitive information, such as confidential data, and thus privacy issues arise. Homomorphic encryption (HE) offers a solution, allowing calculations to be performed on encrypted data without decrypting them. This research compares traditional DNNs and SNNs using the Brakerski/Fan-Vercauteren (BFV) encryption scheme. The LeNet-5 and AlexNet models, widely-used convolutional architectures, are used for both DNN and SNN models based on their respective architectures, and the networks are trained and compared using the FashionMNIST dataset. The results show that SNNs using HE achieve up to 40% higher accuracy than DNNs for low values of the plaintext modulus <i>t</i>, although their execution time is longer due to their time-coding nature with multiple time steps.https://www.mdpi.com/2078-2489/14/10/537deep neural network (DNN)spiking neural network (SNN)homomorphic encryption (HE)Brakerski/Fan-Vercauteren (BFV)NorsePyfhel |
| spellingShingle | Farzad Nikfam Raffaele Casaburi Alberto Marchisio Maurizio Martina Muhammad Shafique A Homomorphic Encryption Framework for Privacy-Preserving Spiking Neural Networks Information deep neural network (DNN) spiking neural network (SNN) homomorphic encryption (HE) Brakerski/Fan-Vercauteren (BFV) Norse Pyfhel |
| title | A Homomorphic Encryption Framework for Privacy-Preserving Spiking Neural Networks |
| title_full | A Homomorphic Encryption Framework for Privacy-Preserving Spiking Neural Networks |
| title_fullStr | A Homomorphic Encryption Framework for Privacy-Preserving Spiking Neural Networks |
| title_full_unstemmed | A Homomorphic Encryption Framework for Privacy-Preserving Spiking Neural Networks |
| title_short | A Homomorphic Encryption Framework for Privacy-Preserving Spiking Neural Networks |
| title_sort | homomorphic encryption framework for privacy preserving spiking neural networks |
| topic | deep neural network (DNN) spiking neural network (SNN) homomorphic encryption (HE) Brakerski/Fan-Vercauteren (BFV) Norse Pyfhel |
| url | https://www.mdpi.com/2078-2489/14/10/537 |
| work_keys_str_mv | AT farzadnikfam ahomomorphicencryptionframeworkforprivacypreservingspikingneuralnetworks AT raffaelecasaburi ahomomorphicencryptionframeworkforprivacypreservingspikingneuralnetworks AT albertomarchisio ahomomorphicencryptionframeworkforprivacypreservingspikingneuralnetworks AT mauriziomartina ahomomorphicencryptionframeworkforprivacypreservingspikingneuralnetworks AT muhammadshafique ahomomorphicencryptionframeworkforprivacypreservingspikingneuralnetworks AT farzadnikfam homomorphicencryptionframeworkforprivacypreservingspikingneuralnetworks AT raffaelecasaburi homomorphicencryptionframeworkforprivacypreservingspikingneuralnetworks AT albertomarchisio homomorphicencryptionframeworkforprivacypreservingspikingneuralnetworks AT mauriziomartina homomorphicencryptionframeworkforprivacypreservingspikingneuralnetworks AT muhammadshafique homomorphicencryptionframeworkforprivacypreservingspikingneuralnetworks |