Research On Time Delay Signature Recognition Of Optical Chaotic System Based On Convolutional Neural Network

With the rapid development of long-distance,large-capacity,and high-speed optical fiber communications,the security of communication systems is increasingly being valued.As a security mechanism,chaotic secure communication has the high randomness of its physical entropy source,ensuring the security of the communication system,that is,the attacker cannot determine the correspondence between the ciphertext and the plaintext.The optical chaotic system based on the delay feedback structure has the unique advantage of large physical parameter space,and it can generate chaotic signals with high bandwidth,high dimension and complexity.But this kind of chaotic signal will have time delay signature(TDS),and the parameter security of the optical chaotic system is threatened by the leakage of TDS.At present,the widely used TDS evaluation methods mainly include auto-correlation function(ACF)and delayed mutual information(DMI),which are mainly based on the linear correlation of time series.In the case of strong system nonlinearity,it is easy to fail.In order to solve the difficult problem of TDS security assessment of optical chaotic systems under strong nonlinearity,a TDS recognition method based on Convolutional Neural Network(CNN)is proposed by using the advantages of neural networks in dealing with nonlinear problems and developed to be a corresponding TDS identification software.The main research contents are as follows:(1)A CNN-based optical chaotic TDS recognition scheme has been proposed.The one-dimensional chaotic time series is used to generate two-dimensional image samples by the delayed reconstruction method,and the excellent image recognition ability of CNN is used to extract the TDS features contained in the two-dimensional images.(2)The effectiveness of the proposed TDS has been verified under the results of the opto-electronic oscillator(OEO)chaotic experiment.Through the numerical simulation of the OEO chaotic system,the effectiveness of the proposed TDS identification scheme under different nonlinear intensities and different signal-to-noise ratios has been explored.(3)The effectiveness of the proposed TDS identification scheme in the optical chaos simulation system based on external optical feedback is verified.The three optical feedback optical chaotic systems with TDS suppression mechanism,namely open loop injection,closed loop injection and phase modulation Sagnac loop mutual coupling,are further studied and when both ACF and DMI methods fail,the effectiveness of the CNN-based TDS identification method has been verified in these systems.(4)A CNN-based TDS recognition software has been designed.Through integrating the process of one-dimensional time series delay reconstruction to generate image samples,the operation of CNN training process and TDS recognition process,it has been compiled and generated to be an easy-to-use TDS recognition software based on MATLAB development environment.