Side channel analysis based on non-profiling singular spectrum analysis

Abstract

Side Channel Analysis is a powerful attack to recover the secret key by exploiting an extra source of information such as timing, power consumption and electromagnetic leakage. Pre-processing scheme is used to improve the attack performance in the restricted traces. One of the pre-processing schemes is Singular Spectrum Analysis, which makes the original trace can be split into main signal and noise signal traces. In order to reduce the cost of SSA, the adversary should choose the criterion to split actual power consumption into two groups according to degree of noise. In previous paper, the criterion was selected as Signal-to-Noise Ratio (SNR) which is one of profiling schemes. However, this scheme reaches on the limitation of the adversary assumption. That is, when the adversary should know the secret key information, SSA scheme can be used as pre-processing. In this paper, in order to utilize SSA scheme in realistic world, we bring the Normalized Inter-Class Variance (NICV) is previously suggested. Thus, by eliminating a powerful assumption, the SSA scheme allows to enable the reasonable pre-processing without the secret key information. Based on our suggestion, we can distinguish the main signal with only the NICV estimation from the original trace. As a result, we demonstrate that our proposal scheme yields significantly the similar results with another scheme based on SNR in MSP430 microcontroller unit.