Optimizing optical chaotic sequences using GAN and the Fisher-Yates algorithm.

Abstract

An optical chaotic sequence optimization scheme combining deep learning and a special post-processing algorithm is proposed and demonstrated. The proposed scheme incorporates the Generative Adversarial Network into the traditional optical feedback chaotic system to optimize the optical chaotic sequence. Following this, the Fisher-Yates algorithm is applied as a post-processing step to further improve randomness. Finally, the optimized sequence is quantized into a random bit sequence. The key advantages of the proposed scheme include the integration of an artificial neural network into the random bit sequence optimization process, providing a novel perspective for future research. Experimental results demonstrate that the proposed scheme significantly improves the distribution characteristics and complexity of chaotic sequences, effectively suppresses the time-delay signature, and ensures that the optimized sequence successfully passes the NIST statistical test suite.