Please use this identifier to cite or link to this item:
Title: Machine-learning-based side-channel-attack on cryptographic devices
Authors: Liu, Huanjia
Keywords: Engineering::Electrical and electronic engineering
Issue Date: 2023
Publisher: Nanyang Technological University
Source: Liu, H. (2023). Machine-learning-based side-channel-attack on cryptographic devices. Master's thesis, Nanyang Technological University, Singapore.
Abstract: Correlation Optimization with Deep Learning Analysis (CO-DLA) has played a significant role in non-profiling Side-channel Attacks (SCA). Currently, the deep learning architecture applied to SCA is mainly Convolutional Neural Networks (CNN), which is used in CO-DLA to build a CNN based CO-DLA (CNN-CO- DLA). It has an outstanding performance in SCA, but the lack of preprocessing still limits its attack efficiency. In this dissertation, we propose an improved CO- DLA using preprocessing technique Wavelet Scattering Transform (WST), termed as WST-based CO-DLA (WST-CO-DLA). WST-CO-DLA utilizes the training-free WST as a preprocessing step to extract the time-frequency features for CO-DLA. With the help of WST, the number of traces required to attack the datasets with mainstream countermeasures is significantly reduced. When the WST extracts information from time domain, windows with various sizes are employed for feature extraction. The various sizes of the WST windows are capable of extracting features associated with different frequency components from different locations of the raw traces. This makes WST resistant to small deformation carried naturally by leakage trace, resulting in better attack efficacy in machine-learning-based non-profiling SCA. In our experiment, the proposed WST-CO-DLA requires 20%, 33%, and 20% lesser traces to reveal the secret key from the unprotected, random delay, and random masked dataset, respectively, as compared to the reported CNN-CO-DLA.
Schools: School of Electrical and Electronic Engineering 
Research Centres: Centre for Integrated Circuits and Systems 
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Theses

Files in This Item:
File Description SizeFormat 
  Restricted Access
2.76 MBAdobe PDFView/Open

Page view(s)

Updated on Jun 15, 2024


Updated on Jun 15, 2024

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.