Chaos key enhanced physical layer secure transmission method based on the convolutional long short-term memory neural network

Abstract

In this paper, we propose a method for training a key-enhanced chaotic sequence using the convolutional long short term memory neural network (CLSTM-NN) for secure transmission. This method can cope with the potential security risk posed by the degradation of chaotic dynamics when using chaotic model encryption in traditional secure transmissions. The simulation results show that the proposed method improves the key space by 1036 compared to traditional chaotic models, reaching 10241. The method was applied to orthogonal chirp division multiplexing (OCDM). To demonstrate the feasibility of the proposed scheme, we conducted transmission experiments of encrypted 16 quadrature amplitude modulation (QAM) OCDM signals at a speed of 53.25 Gb/s over a 2 km length of 7-core optical fiber and test different encryption schemes. After key enhancements, the overall number of keys in the system can increase from 18 to 105.The results show that there is no significant difference between the bit error rate (BER) performance of the encryption method proposed in this paper and the traditional encryption method. The maximum performance difference between the different systems does not exceed 1 dBm. This fact proves the feasibility of the proposed scheme and provides new ideas for the next generation of secure transmission.