Impossible differential analysis is widely used as a cryptographic tool for key recovering or distinguishing attacks on block ciphers. There are mainly two methods for searching impossible differentials: one is case-by-case analysis and the other is automatic search. In this paper, the authors propose a new method for finding impossible differentials in an automatic way. The key idea behind the method is to unify the representations of the transformation during the round function. Thus, the program can exhaustively search the whole space for possibilities of mismatch. The method uses matrix representations, so the memory and computation costs are feasible.

But the unified method has disadvantages as well. First, the method seems to only be effective on Feistel structures, especially for generalized Feistel networks. Second, the method cannot look into the details of the round transformation, which means the results might not be optimal. How to improve the method to fix these two disadvantages will be interesting future research.