A Situation Awareness Approach for Network Security Using the Fusion Model

Abstract

Aiming at the limited learning ability of a single model, the objective of this paper is to investigate situational awareness of the network security which is established on the fusion model. In this paper, a convolutional neural network (CNN) and long short-term memory (LSTM)-based model for situational assessment of the network security condition are provided. According to different fusion methods, the parallel and serial CNN-LSTM fusion models were constructed to evaluate the UNSW-NB15 data set, and both the situation values and levels were obtained. The investigational outcomes illustrate that the evaluation accuracy of the two models can reach up to 85.19% and 92.59%, respectively. A situation prediction model called IPSO-ABiLSTM is suggested and is based on improved particle swarm optimization (IPSO) and attention fusion bidirectional long short-term memory (ABiLSTM). The IPSO has the characteristics of faster convergence speed to optimize the ABiLSTM network parameters and obtain the optimal parameters for situation prediction. The investigational outcomes illustrate that the suggested IPSO-ABiLSTM model has a fitting degree of up to 0.9922, which can effectively achieve the situation prediction in the network security.

Research on the Implementation Method of Situational Awareness Strategy by Fusing Computational Migration Models

“…Currently, many scholars have conducted in-depth research on situational awareness strategies and achieved certain research results. Literature [15] proposes a decision tree-based situational prediction model for Vehicular Network Security, by categorizing specific attributes and using the information gain rate to construct a decision tree, thus realizing the situational prediction of Vehicular Network Security, which improves the accuracy of prediction, but ignores the long-distance nature of the vehicle traveling data; Literature [16] proposes a Long Short-Term Memory (LSTM)-based and Multiple Attention Approach (MADA) hybrid model for predicting a given time series, which outperforms most of the tested methods in terms of symmetric mean absolute percentage error, but fails to extract features from the time series data, and has low warning accuracy; Literature [17] proposes a spatio-temporal neural network (GCN-DHSTNet) model, which is modeled by a graphical convolution network, and dynamically learns, based on global spatial relations of traffic among nodes, the spatio-temporal characteristics of traffic data, and then simultaneously deal with the complex and dynamic spatio-temporal dependence of traffic flow, the model effectively captures the dynamic temporal correlation, but it can only be used when the node distances are small; Literature [18] based on the Gram's angular disparity field theory to understand the temporal correlation of deviation changes, and establish a convolutional neural network model to predict the future trend of the system, the method of predicting the subsequent postures with high accuracy but without considering the computational load; literature [19] proposes a situation prediction model named IPSO-ABiLSTM, which is based on Improved Particle Swarm Optimization (IPSO) and Attention Fused Bidirectional Long and Short-Term Memory (ABiLSTM), which is able to converge the parameters of the neural network quickly, but without processing the data itself.…”