Hypothesis Testing Based Side-Channel Collision Analysis

Side-channel collision analysis has become a research hotspot since its first publication in 2003. Compared with differential power analysis (DPA) and correlation power analysis (CPA), collision analysis does not need to know the intermediate value and does not depend on a specific power model. Collision analysis includes collision detection phase and key recovery phase. After collisions have been detected, the cryptographic key can be recovered from these collisions using such cryptanalytic techniques as linear collision attack and algebraic collision attacks. Therefore, the success rate of the entire attack depends largely on the success rate of collision detection. In this paper, we propose an efficient collision detection method, which combines side channel information with hypothesis testing in statistics. With the traces gathering by repeatedly encrypting a certain plaintext that is unnecessary to be known, we are able to evaluate the cumulative distance between two collision S-boxes, then, define a null hypothesis H
₀
and calculate the rejection region accordingly. Given some traces collected from repeatedly encrypting any identical plaintext, collisions can be detected by determining whether the cumulative distance between the corresponding two S-boxes falls into the rejection region or not. We compare our work with those related to side channel collision attack in recent years from perspective of four hot issues. Comparing with current methods, our proposed collision detection method is simpler and more feasible. Furthermore, the experiments show that the more accurate the constant power is, the higher the success rate of collision detection is.