Power Consumption Attack Based on Improved Principal Component Analysis

Accompanied with the status quo and problems that the low efficiency in the traditional methods of principal component analysis (PCA) when we face the problems of correlated power attack with large amount of data, we presents an improved method to reduce the noise of power data by wavelet packet transform (WPT) and then reduce the dimension by traditional principal component analysis, based the conclusion we have arrived about the advantage of wavelet packet transform in signal processing. It is more productive than common methods in the data processing phase of the related power attack, especially on the occasion that we own high dimensional data with low signal to Noise Ratio (SNR). Just to show you where we can optimize, the middle position of SM4 encryption algorithm was selected to measure the power consumption, and compared with the results of traditional principal component analysis. The results show that not only is the number of curves has been significantly reduced, but the computational complexity has been decreased easily, by all means, the computational time is less than the original required time so that the attack efficiency is significantly improved. Aiming at the goal with a highly targeted way to reduce the amount of data which are needed to crack the key especially for course of power analysis, the proposal submitted by us have the certain advantages under this circumstance when we face the high latitude data with low SNR within the process of correlated power attack.