File(s) under permanent embargo
Towards Privacy-Preserving Neural Architecture Search
conference contribution
posted on 2023-02-07, 00:40 authored by F Wang, LY Zhang, Lei PanLei Pan, S Hu, Robin Ram Mohan DossRobin Ram Mohan DossMachine learning promotes the continuous development of signal processing in various fields, including network traffic monitoring, EEG classification, face identification, and many more. However, massive user data collected for training deep learning models raises privacy concerns and increases the difficulty of manually adjusting the network structure. To address these issues, we propose a privacy-preserving neural architecture search (PP-NAS) framework based on secure multi-party computation to protect users' data and the model's parameters/hyper-parameters. PP-NAS outsources the NAS task to two non-colluding cloud servers for making full advantage of mixed protocols design. Complement to the existing PP machine learning frameworks, we redesign the secure ReLU and Max-pooling garbled circuits for significantly better efficiency (3 436 times speed-up). We develop a new alternative to approximate the Softmax function over secret shares, which bypasses the limitation of approximating exponential operations in Softmax while improving accuracy. Extensive analyses and experiments demonstrate PP-NAS's superiority in security, efficiency, and accuracy.
History
Volume
2022-JunePagination
1-6Publisher DOI
Start date
2022-06-30End date
2022-07-03ISSN
1530-1346ISBN-13
9781665497923Title of proceedings
Proceedings - IEEE Symposium on Computers and CommunicationsEvent
2022 IEEE Symposium on Computers and Communications (ISCC)Publisher
IEEEUsage metrics
Categories
No categories selectedKeywords
Licence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC