National Cryptography Development Fund (No.MMJJ20180203);Open Fund of State Key Laboratory of Mathematical Engineering and Advanced Computing (No.2018A03);Open Fund of Key Laboratory of Information Assurance Technology (No.KJ-17-002)
REN Jiong-jiong, ZHANG Shi-wei, LI Man-man, et al. SAT-Based Automatic Search for Impossible Differentials and Zero-Correlation Linear Approximations in ARX[J]. Acta Electronica Sinica, 2019, 47(12): 2524-2532.
DOI:
REN Jiong-jiong, ZHANG Shi-wei, LI Man-man, et al. SAT-Based Automatic Search for Impossible Differentials and Zero-Correlation Linear Approximations in ARX[J]. Acta Electronica Sinica, 2019, 47(12): 2524-2532. DOI: 10.3969/j.issn.0372-2112.2019.12.010.
SAT-Based Automatic Search for Impossible Differentials and Zero-Correlation Linear Approximations in ARX
Xor) algorithms are based on three operations:modular addition
exclusive-OR and bitwise rotation
which execute very fast in both software and hardware. Impossible differential cryptanalysis and zero-correlation linear cryptanalysis are among the most powerful attacks for ARX ciphers. The key problem for the attacks is searching more and longer impossible differentials and zero-correlation linear approximations. Although there are many automatic search algorithms
these approaches do not fully utilize the non-linear component properties
which cannot reach their potential. A SAT (Satisfiability) -based model to automatic search for impossible differentials and zero-correlation linear approximations in ARX is proposed.By exploiting behaviors of every component
especially the differential and linear propagation properties through general modular addition and key modular addition operations
we generate computable and easily implementable constraints and establish the SAT-based model. As applications
we apply our model to Chaskey
SPECK and HIGHT. For Chaskey
we are able to find 13 4-round impossible differentials and 1 4-round zero-correlation linear approximation for the first time. For SPECK
we first find 10 6-round zero-correlation linear approximations of SPECK32 and 15 6-round zero-correlation linear approximations of SPECK48. Besides
we find 4 17-round impossible differentials and zero-correlation linear approximations of HIGHT in a few minutes. Compared with the existing results
both the number and the search time of distinguishers are significantly improved. In addition
by repackaging the output interface of the STP solver
we establish the automated SAT\\SMT model
which can give the upper bound of rounds of impossible differentials and zero-correlation linear approximations under special input and output differential and mask sets for ARX algorithm.