Optimization of S-Boxes by Evolving

The Substitution Box (S-Box for short) is the core component of a block cipher. The strength of a block cipher basically depends on the cryptography property of the S-Box it employs. Designers use "nonlinearity", "autocorrelation" and "difference" as the S-Box designing criteria. In this paper, we focus on optimizing S-Boxes by two heuristic methods - Genetic Algorithm and Simulated Annealing. Quite a lot of 8×8 S-Boxes with a joint feature value (Nonlinearity, Autocorrelation) = (102, 80) can be gotten just by our GA or SA with a steady probability. Both our methods are comparable to J. Clark\´s hybrid method (Simulated Annealing combined with Hill Climbing).