Research On Network Security Situation Prediction Model Based On BiLSTM

With technology iteration,the trend of industrial automation is becoming more and more obvious,and the era of Industry 4.0 led by smart manufacturing has arrived,and the development of industrial Io T is in full swing.However,the complexity of network environment and diversification of network attacks make the industrial Io T face security problems.The traditional network security devices all adopt the protection strategy of independent separation,which cannot realize message sharing and only reflect the security status of the network from one side.Therefore,the overall assessment of the overall security status of the industrial Io T and the prediction of the security trend of the network in the future period have become the focus of the prediction of the security posture of the industrial Io T.The main research elements are as follows:1.for the problem of incomprehensive selection of posture elements and single dimension of evaluation system,this paper selects 14 secondary indicators under the four dimensions of operation dimension,vulnerability dimension,stability dimension and threat dimension,and constructs an industrial Io T assessment index system,and quantifies the secondary indicators.2.To address the problem of low classification accuracy in industrial Io T security situation assessment,a least squares twin support vector machine based on arithmetic optimization algorithm optimization is proposed for industrial Io T security situation assessment model.To satisfy the multi-classification task of posture evaluation,a least-squares twin support vector machine based on a many-to-one strategy with improved regular terms is chosen in this paper.In order to solve the problem of unbalanced global and local search capability of the arithmetic optimization algorithm,cosine control factor and Corsi variance are introduced to improve it.The improved arithmetic optimization algorithm was applied to the least squares multi-classification twin support vector machine parameter search,and the experiments show that the classification accuracy of the model in this paper is higher compared with the existing models.3.To address the problem of low prediction accuracy in industrial Io T security posture prediction,an industrial Io T security posture prediction model based on BiLSTM neural network optimized by sparrow search algorithm is proposed.In order to deeply explore the hidden information in the industrial Io T time series data,this paper superimposes two BiLSTM layers,and then connects the dropout layer and the dense layer to achieve the purpose of improving the generalization ability and transforming the output dimension.In order to solve the problems that the initial population of sparrows is easily dispersed or concentrated,and the algorithm is easy to fall into local optimum,reverse learning strategy and Lévy flight strategy are introduced to improve it.The improved sparrow search algorithm is optimized for the parameters of the BiLSTM network,and the experiments show that the prediction accuracy of this model is higher compared with existing models.Based on the research content of this paper,a visual display page is built to visualize the prediction results.