Research On Security Situation Prediction Method For Industrial Internet

Industrial Internet,as an important part of the next generation information infrastructure,promotes the development of digital,intelligent and networking of industry,but it also faces numerous security problems.The existing security situation prediction method for Industrial Internet has low accuracy and high time complexity.Therefore,it is very important to study the accurate and efficient security situation prediction method for the complex Industrial Internet environment.So as to solve the above problems,this thesis mainly uses the long and short term memory network,convoluted neural network,bidirectional gated circulation unit network et al,to carry out research on the key technology of Industrial Internet security situation prediction,which is as follows:1.In industrial production processes,traditional predictive models often struggle to accurately and efficiently forecast network security situations due to the large amount of data,high dimensionality,and complex time series involved.Therefore,this thesis proposes an adaptive Industrial Internet security situation prediction model.This model uses entropy correlation method to quantify the situation values of multiple cycles,segments the situation values using sliding window mechanism,predicts the situatio0 n values based on adaptive triple exponential smoothing method,and corrects the situation values by using time-varying weighted Markov chain prediction error.Finally,the Industrial Internet security situation prediction model in this thesis is established based on this model.2.Due to the influence of various network security factors on the time-related Industrial Internet security situation value,this thesis proposes an Industrial Internet security situation prediction method based on the fusion of nonlinear dynamic particle and improved artificial fish swarm algorithm optimize LSTM network.This method first uses the LSTM network to mine the time-relatedness of Industrial Internet security situation data and uses the analytic hierarchy process to calculate the situation value.The sigmoid weighted linear element is introduced to handle the gradient problem in back propagation,and the input value is multiplied by the sigmoid activation function to enhance the structure of the LSTM network and improve the prediction accuracy.Then,the NDPSO and the IAFSA are combined to optimize the hyperparameters of the LSTM to improve the local optimization ability and optimization speed.3.Due to the fact that mainstream network security situation prediction methods are based on a single neural network model,which is not conducive to judging the security situation of Industrial Internet networks,this thesis proposes an Attention mechanism-based CNN-BiGRU method for predicting the security situation of Industrial Internet networks.This method uses an attention mechanism to assign different weight values to different situation attributes and uses particle swarm optimization to optimize the hyperparameters of the BiGRUSimulation results show that the proposed model and prediction methods can meet the needs of Industrial Internet security situation prediction,can predict the security of the Industrial Internet and provide some help for network security administrators.