抄録
Electromagnetic emissions leak confidential data of cryptographic devices. By exploiting such emissions, electromagnetic analysis (EMA) is performed with EM probes to extract secret information from these devices. Owing to the locality of EM emissions, namely, secret information may leak from multiple locations around cryptographic devices, it is difficult to determine the exact location before conducting an EMA. In this paper, signal variance of EM emissions during encryption is proposed to identify the information leakage of unprotected and protected cryptographic modules. We prove that signal variance is an equivalent metric to Difference of Means (DoM). Thus, by computing the higher signal variances based on near-field scan, the data-dependent EM emissions are disclosed, namely, the leakage locations are found. In addition, a small and low-cost probe is made to verify the proposed EMA on application-specific integrated circuit (ASIC) implementations. The experiment on AES PPRM1 implementation indicates that misjudgments of leakage are reduced, and the accuracy is improved 48.6% compared with existing methods. Moreover, the experiment of EMA against AES WDDL implementation shows that signal variance is also effective in exposing the leakage locations in the presence of countermeasures. The performance of EMA is enhanced.
