CPA performance enhancement based on spectrogram

In a side channel attack, misalignment is a major factor that decreases the attack effectiveness. In order to resolve this issue, correlation power frequency analysis (CPFA) was recently introduced in the frequency domain by Schimmel. This method changes signals from the time domain to the frequency domain to analyze the information using FFT and is able to analytically solve the decrease in the attack effectiveness due to the misalignment. However, for signals that change their frequency components randomly, the results of the analysis are not as good. Moreover, there is a critical point that loses information in the time domain. In order to solve this limitation, we have developed correlation power spectrogram analysis (CPSA), which has excellent performance in side channel analysis. This method converts the time domain information to time domain-frequency domain information using a spectrogram, and the changed information keeps the time information of regular resolution. This method shows excellent performance for the variation in frequency components, as well. In this study, AES power consumption signals were collected from ARM, IC CARD, and MSP430 chips that were developed in the SCARF system. Using these signals, the method shown in this paper yields better performance than CPA or CPFA.