华中科技大学学报（自然科学版）2025，Vol.53Issue(5)：52-57,6.DOI:10.13245/j.hust.250876
基于SVD和MLP的多分类能量信息泄漏评估方法
SVD and MLP-based multi-classification power information leakage assessment method
摘要
Abstract
Aiming at the problems that the existing power information leakage assessment methods based on deep learning are easy to hide and the training efficiency of neural network is low,a multi-classification power information leakage assessment method based on singular value decomposition(SVD)and multi-layer perceptron(MLP)was proposed.First,the raw power consumption data was processed based on SVD,and some representative features were selected to reduce the amount of data processing.Then,the projected data was divided into multiple groups according to the corresponding plaintexts,and the"multiple classification"MLP was used to assess the leakage.The verification was carried out based on simulated power,ASCAD power and Chipwhisperer power,respectively.Results showe that the proposed method requires 18.5%,17.1%and 13.6%less power traces to detect leakage,and 7.1%,18.8%and 21.1%less assessment time,respectively,than the existing deep learning-based assessment method.关键词
侧信道/能量信息泄漏/深度学习/奇异值分解/多层感知器Key words
side-channel/power information leakage/deep learning/singular value decomposition/multi-layer perceptron分类
信息技术与安全科学引用本文复制引用
郑震,严迎建,刘燕江..基于SVD和MLP的多分类能量信息泄漏评估方法[J].华中科技大学学报（自然科学版）,2025,53(5):52-57,6.基金项目
国家自然科学基金资助项目(62302519). （62302519）
