Research on Data Preprocessing in Side Channel Attack Based on Multi-Layer Perception

The data encryption process in the encryption chip will be leaked by means of timing operation time, probe collection operation power and electromagnetic signal interception, which makes side channel attack possible. The actual power consumption data usually contains a lot of noise, which reduces the success rate of side channel attack. This paper studies the preprocessing of FPGA power data, proposes two indexes -- MSNR (Maximum Signal Noise Ratio) and GE (Guessing Entropy), and the variance and mean curves of Hamming weight assist modelling to summarize three reliable preprocessing methods based on dynamic powers.