CAJ | 학술논문

传统小波分解去噪需要先验知识，缺乏自适应性，将多元经验模态分解间隔阈值（Multivariate Empirical Mode Decomposition，MEMD）滤波方法用于电磁泄漏信息预处理。首先对电磁泄漏信号进行对齐处理。然后，针对多元经验模态分解直接阈值滤波产生的不连续性问题，提出采用MEMD间隔阈值滤波方法进行不相关模态滤波，解决传统经验模态分解的模态混叠问题。最后，分别采用巴特沃斯低通滤波、小波阈值、MEMD-DT和MEMD-IT方法对密码芯片电磁泄漏数据进行去噪处理，通过基于最小二乘支持向量机分类识别，实验结果表明该自适应去噪方法的优于传统的电磁侧信道分析预处理方法。
전통소파분해거조수요선험지식，결핍자괄응성，장다원경험모태분해간격역치（Multivariate Empirical Mode Decomposition，MEMD）려파방법용우전자설루신식예처리。수선대전자설루신호진행대제처리。연후，침대다원경험모태분해직접역치려파산생적불련속성문제，제출채용MEMD간격역치려파방법진행불상관모태려파，해결전통경험모태분해적모태혼첩문제。최후，분별채용파특옥사저통려파、소파역치、MEMD-DT화MEMD-IT방법대밀마심편전자설루수거진행거조처리，통과기우최소이승지지향량궤분류식별，실험결과표명해자괄응거조방법적우우전통적전자측신도분석예처리방법。
To overcome the dependence of prior knowledge and non-adaptive of traditional filtering algorithm, a Multivariate Empirical Mode Decomposition interval threshold (MEMD-IT) de-noising which is a nonparametric signal de-noising approach is presented as a preprocessing stage for electromagnetic radiation signals. First, the near field electromagnetic leakage signals were captured while cipher device was executing RC4 algorithm. MEMD-IT is developed which can reduce the discontinuity induced by EMD-DT. Comparing with the other filters, such as Butterworth low-pass filter, wavelet threshold de-noising, and MEMD direct threshold (MEMD-DT) de-noising, the proposed de-noising method evaluated by the Hanming weight classification results based on LSSVM has a better performance.