Paper of Interest

[richs27]

Hi Marcus,

Thought you might like this:

Bitsliced Implementation of Non-Algebraic 8 × 8Cryptographic S-Boxes Using x86-64 ProcessorSIMD Instructions

https://www.semanticscholar.org/paper/Bitsliced-Implementation-of-Non-Algebraic-8×8-Using-Sovyn-Khoma/2346f7dcd1fd74c3952eee5f1096de9202a5f8b5

@dansarie Updated 3 May 2024
Further to the above paper, I downloaded the sboxes and run logs of the ternary examples illustrated in the paper and was surprised to find the time taken to generate Sbox0, an 8x8 bit one with 209 ternary instructions, was 6 hours 10min.
I made contact with one of the authors, to find out how much resource was engaged in this task.

I was told a single 4 core CPU, which threw me. He also pointed me to another paper they have done, refining the heuristics used, which further reduces the LUT count to 178, in about the same time, but using GPU for the searches. This paper is here: https://elibrary.kubg.edu.ua/id/eprint/45690/1/Y_Sovyn_V_Khoma_I_Opirskyy_V_Kozachok_3421_CPITS.pdf

Certainly food for thought.

[dansarie]

Sorry for the late response. The algorithm currently used by the program is definitely not the most efficient one possible, both in terms of time and solution size. One day, when time allows, I hope to be able to implement a more efficient algorithm.