Genetic Algorithm Assisted State-Recovery Attack on Round-Reduced Xoodyak

Zimin Zhang, Wenying Zhang, Hongfang Shi

Abstract: Genetic algorithm (GA) has led to significant improvements in many challenging tasks, including combinatorial optimization, signal processing, and artificial life. It shows enormous potential for cryptanalysis. This paper designed a heuristic algorithm based on GA for the known-plaintext attack on round-reduced Xoodyak, a finalist of the NIST lightweight cryptography project, under the nonce-respecting setting. To accomplish this, we firstly remodel Xoodoo, the underlying permutation of Xoodyak, portraying it as a function whose input and output are continuous variables defined in [0, 1], representing the likelihood that each bit is equal to 1 and describing the goal of cryptanalysis as an objective function optimized with GA secondly. Consequently, we can abstract the potential information of the unknown state of Xoodyak from the results given by GA. Compared with traditional methods, ours requires less knowledge about complex cryptanalysis as GA can work well with lower complexity, both in time complexity and data complexity, and can be carried out under more restricted conditions.