全光通信网络非线性突变频率干扰检测算法

Nonlinear Mutation Frequency Interference Detection Algorithm for All Optical Communication Networks

【目的】针对全光通信网络非线性突变频率干扰问题进行分析时,主要依托于估计信道协方差直接展开干扰检测,并未展开网络通信信号变换处理,使得检测结果的F1⁃score值较低,为此提出了基于时域有限差分的全光通信网络非线性突变频率干扰检测算法。【方法】结合数据包抓包和镜像两种方式采集全光通信网络流量数据,并进行清理、转换和规约处理。依托于时域有限差分工作原理,在时域和频域空间内描述信号的时宽和带宽,应用导数与傅里叶变换算法对实时采集信号进行处理,利用变换后的信号分析非线性突变频率干扰情况,再针对变换后的信号进行检测分析。在时间⁃频率联合特征分析方法辅助下,提取干扰信号的时域、频域特征,依托于反向传播算法和最小化损失函数,简化非线性突变频率干扰检测过程,采用特征距离函数替换网络损失函数,并将其输入基于孪生网络的干扰识别模型中,得出非线性突变频率干扰检测结果。【结果】在不同噪声条件下,所提算法非线性突变频率干扰检测结果的F1⁃score值保持在0.95以上,检测时间低于40 ms。【结论】应用时域有限差分方法的新型检测方法,能够更准确地反映当前通信网络的干扰情况,保证通信网络正常运行,更好地满足了全光通信网络干扰检测要求。

【Objective】When analyzing the nonlinear mutation frequency interference problem in all optical communication net⁃works,interference detection mainly relies on estimating channel covariance directly,without expanding network communication signal transformation processing,resulting in low F1⁃score values of the detection results.Therefore,a nonlinear abrupt frequency interference detection algorithm based on finite difference time domain is proposed for all⁃optical communication networks.【Methods】We collect all optical communication network traffic data by combining packet capturing and mirroring methods,fol⁃lowed by cleaning,conversion,and protocol processing.Relying on the working principle of finite difference time⁃domain method,we first describe the time width and bandwidth of the signal in time⁃domain and frequency⁃domain space.Then we apply derivative and Fourier transform algorithm to transform the real⁃time collected network communication signal.Next,we use the transformed signal to analyze the nonlinear abrupt frequency interference.Finally the transformed signal is detected and analyzed.With the aid of the time⁃frequency joint feature analysis method,the time⁃domain and frequency⁃domain features of the interference signal are extracted.Relying on the back propagation algorithm and the minimization of the loss function,the detection process of nonlinear abrupt frequency interference is simplified.The characteristic distance function is used to replace the network loss function,and it is input into the interference recognition model based on the twin network to obtain the detection results of nonlinear abrupt frequency interference.【Results】The experimental results show that under different noise conditions,the F1⁃score value of the proposed algorithm's nonlinear mutation frequency interference detection results remains above 0.95.The detection time is less than 40 ms.【Conclusion】The new detection method using the finite difference time⁃domain method can more accurately reflect the interference situation of the current communication network.It ensures the normal operation of the communication network,and meets the interference detection requirements of all⁃optical communication network.