Perfect Timing! Using Spiking Neural Networks for Temporal Variations in Side Channel Analysis
Keywords
Loading...
Authors
Issue Date
2025-01-24
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
Techniques from artificial intelligence have been used in cyber security to analyse data leaked
through side channels. Examples of such side-channels include power consumption or electro
magnetic emissions. Currently, convolutional neural networks are the most promising network
architecture for the analysis of synchronised and desynchronised power traces. In this study,
we investigate the usability of a new generation of neural networks for side-channel analysis.
These networks originate from the field of neuromorphic computing. As spiking neural networks
operate in a temporal manner by nature, we hypothesise that such networks have the potential
to deal with desynchronised traces. We compared two convolutional neural network models
(ASCAD and VGG16) on their suitability for a conversion to a spiking implementation. Our
findings indicate that the ASCAD model is well suited for the conversion to a spiking neural net
work without performance loss. In contrast, the VGG16 model was less suitable for conversion,
but useful insights were gained. We found that spiking neural networks have the potential to be
applied in side-channel analysis, but more research is needed to investigate the full potential.
Keywords — side-channel analysis, neuromorphic, spiking neural networks, desynchroni
sation, temporal variations
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen