研究成果

Physical attacks on electronic devices that involve cryptographic functions, such as IC cards and credit cards, pose a significant threat to security. Side-channel and fault attacks are typical threats, and in recent years, deep learning-based analysis methods have also been proposed to extract encryption keys effectively. Threshold implementation is one of the countermeasures to thwart side-channel attacks, which conceals physical leakage by splitting the processed data into multiple values termed shares. The protected circuit must be evaluated both theoretically and practically. However, while evaluation methodologies for side-channel analysis have been well established, those for fault analysis have not yet been thoroughly explored. In this paper, we propose two practical evaluation methods for fault analysis utilizing the fault sensitivity. The first detects data-dependent fault occurrences using Welch’s t-test, and the second assesses fault sensitivity on each share using Shannon entropy. We then conduct fault injection experiments on a TI-protected AES implementation to demonstrate the efficacy of our proposed methods. The results show that the t-test evaluation can successfully detect data-dependent faults, and the entropy-based evaluation reveals differences in fault sensitivity among the shares.