Improving decryption quality of optical chaos communication using neural networks

Abstract

Optical chaos communication is a promising secure transmission technique because of the advantages of high speed and compatibility with existing fiber-optic systems. The deterioration of chaotic synchronization quality caused by fiber optic transmission impairments affects the quality of recovery of information, especially high-order modulated signals. Here, we demonstrate that the use of a convolutional neural network (CNN) with a bidirectional long short-term memory (LSTM) layer can reduce the decryption BER in an optical chaos communication system based on common-signal-induced semiconductor laser synchronization. The performance of a neural network is investigated as a function of network parameters and chaos synchronization coefficient. Experimental results show that the BER of 16-ary quadrature-amplitude-modulation (16QAM) signal after 100-km fiber transmission is decreased from 3.05 × 10−2 to below the soft-decision forward-error-correction (SD-FEC) threshold of 2.0 × 10−2.