Optimizing Guessing Strategies for Algebraic Cryptanalysis with Applications to EPCBC

In this paper we demonstrate how to use Mixed Integer Linear Programming to optimize guessing strategies for algebraic cryptanalysis with applications to the block cipher EPCBC. Using our optimized guessing strategy we are able to attack 5 rounds of EPCBC96 and 8 rounds of EPCBC-48 faster than brute force using one and two known plaintexts resp. Finally, we are able to identify a class of weak keys for which the attack is faster than brute force for up to 7 rounds of EPCBC-96. Alongside results on EPCBC we believe that the proposed technique of optimized guessing is a useful tool in a more general context of algebraic cryptanalysis.