Network Security Situation Prediction Based on Dempster-Shafer Evidence Theory and Recurrent Neural Network

Due to characteristics of evaluation conflicts and subjective biases in conventional fuzzy evaluation system, and problems of low accuracy and robustness performance for the network security situation prediction caused by such characteristics, a novel network security evaluation algorithm based on Dempster-Shafer(D-S) evidence theory and recurrent neural network was proposed. The roles of network security based on the evaluation data of different experts were constructed, and then the triangle fuzzy function was applied to obtain the experts’ evaluation index. The weight D-S evidence theory was used to screen infer, and correct the evaluation index, and the loss and possibility matrixes were constructed. Lastly, the loss and possibility matrixes were imported in RNN model to finish network security evaluation. The results of simulation experiments show that the D-S evidence theory is proved to be able to deal with experts’ collision problems, and the RNN model can improve the precision and robustness of network security situation prediction results.