Using selected-plaintext sets for efficient evaluation of EM information leakage from cryptographic devices

In this paper, we present a method for selecting a specific set of acquired EM traces with the aim to improve the efficiency of EM analysis. The proposed method improves the correlation factor by using a set of selected plaintexts. We demonstrate the efficiency of our method through correlation electromagnetic analysis (CEMA) experiments conducted using SASEBO. The results clearly show the impact of selecting a subset of plaintexts on the outcome of CEMA. With this selection method, the proposed technique can be used for evaluating the security of cryptographic devices within a short period of time by using a small number of waveforms, even in cases where the noise is large, for example, outside a cryptographic device.