Optimization of Cryptographic Properties Using Metaheuristic Algorithms

Abstract

With the rise of the Internet of Things (IoT), ensuring strong and adaptable cryptographic mechanisms has become a major challenge, particularly for resourceconstrained devices. In this context, this thesis focuses on the optimization of cryptographic properties using metaheuristic algorithms to enhance the robustness of symmetric encryption schemes. This thesis proposes a hybrid optimization method for generating substitution boxes (S-Boxes), key components in symmetric encryption, by combining two metaheuristic algorithms: Particle Swarm Optimization (PSO) for global exploration and Simulated Annealing (SA) for local refinement. Two variants are studied: a deterministic initialization using the Elliptic Curve Diffie-Hellman (ECDH) protocol, and a dynamic generation inspired by the Advanced Encryption Standard (AES). Evaluation is based on a multi-criteria fitness function incorporating nonlinearity, the Strict Avalanche Criterion (SAC), the Bit Independence Criterion (BIC), and the Differential and Linear Approximation Probabilities (DAP and LAP). The experimental results are significant: nonlinearity reaches 110, average SAC is 0.505, and DAP remains below 0.039, complying with secure 8 × 8 S-Box standards. The hybrid approach outperforms standalone PSO and SA in terms of overall performance and convergence stability.