Wavelet Transform-principal component analysis in electromagnetic attack

Abstract

In the attack of the encryption algorithm ARCFOUR(RC4), new method of machine learning based on Wavelet Transform(WT) and Principal Component Analysis(PCA)is proposed. WT is used to reconstruct the new signals to extract information from the original signals effectively. Components in low and high frequency and reconstructed signals are made by WT. In order to analyze the impact of the signals on the success rate of prediction, four kinds of signals are adapted respectively for dimensionality reduction, which contains initial signals, reconstructed signals, and the signals in low and high frequency. By the classification of Support Vector Machine (SVM), the results show that the effect of the reconstructed signals is the best one. The reconstructed signals reduce the noise influence. In the range of 500 dimensions, the classification effect of the reconstructed signals is obviously better than others. As the dimension increases, the effect becomes small. The effect of signals in low frequency is more effective than that of the original in most of dimensions. The classification success rate is still high with fewer dimensions. In the four kinds of signals, the effect of the signals in high frequency is the worst. The results show that WT combined with PCA is a good method to handle with classification.