Authors: Wang, Qizheng
Wang, Lianhai
Xu, Shujiang
Zhang, Shuhui
Shao, Wei
Mihaljević, Miodrag J. 
Affiliations: Computer Science 
Mathematical Institute of the Serbian Academy of Sciences and Arts 
Title: Single-Layer Trainable Neural Network for Secure Inference
Journal: IEEE Internet of Things Journal
Issue Date: 2024
Rank: ~M21a
ISSN: 2327-4662
DOI: 10.1109/JIOT.2024.3480195
Abstract: 
Secure neural network inference provides privacy guarantees for both the client and the server, and is an integral approach in Machine Learning as a Service setting (MLaaS). However, the multi-layer structure in the neural network introduces frequent activation function calculations, which causes large overhead. Most of the prior secure inference systems focused on designing cryptographic protocols to improve computational efficiency, but high computing and communication overhead are still bottlenecks in practicality. In this work, we refocus on the potential of shallow neural networks and propose a model with only one trainable layer to reduce the required computation. Our main contributions are in three-fold: (i) introduce training-free weights and formally prove their contribution in the model expressivity; (ii) design the Self Enhanced Module that is more suitable for shallow models as a alternative for the activation function; (iii) propose a linear layer with multi-scale and normalization property, named Nested&Norm Conv. We conduct extensive experiments on visual datasets and the results demonstrate the proposed single-layer trainable model holds promise as a viable platform for secure inference in practical applications.
Publisher: IEEE

Show full item record

Page view(s)

59
checked on Dec 7, 2024

Google ScholarTM

Check

Altmetric

Altmetric


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.