Speech Encryption Scheme Based on Chaotic Memristor Neural Network and S-Box

In this study, we propose a novel speech encryption scheme based on a chaotic memristor neural network (MNN) and S-box. Firstly, we consider a magnetic-controlled memristor model and a new MNN model, and explore the chaotic behaviors of the new MNN under different memristor strengths. By iterating the new chaotic MNN, we generate an S-box through bit AND and modulo operations. Subsequently, the original speech is converted into a one-dimensional array, which is split into an integer part array and a fractional part array. The integer part array of the above one-dimensional array is converted into a binary matrix, which is separated into two binary matrices. These two binary matrices undergo XOR operations with two chaotic sequences and a sequence generated by the S-box. We convert the obtained matrix into a new array, and perform scrambling operation on the new array by using another chaotic sequence. Then, the encrypted speech can be obtained. Finally, the proposed encryption scheme is evaluated through experimental simulations and security analyses, demonstrating its high level of security.