Network Security Situation Assessment Method Based Eigenvector Centrality

Traditional network security situation assessment methods rely too much on expert systems, and with networks that gradually show dynamic and heterogeneous characteristics, such as vehicle networks and aviation networks, the general assessment methods can no longer be used well. To address these problems, this paper proposes a method that uses a neural network approach to identify threats and based on this, quantifies the importance of network devices using eigenvector centrality indicators, combining with the severity and impact of the attack to assess the network security situation. The method uses sparse encoder to extract network flow features for identification; uses a combination of BiLSTM with attention mechanism to identify the attacks present in the network; modeling the correlation relationship between hosts in the network by classifying different hosts and quantifying the impact of various attacks on the network space to obtain the situation value. The results from the experiment demonstrate that the model performs superiorly to the compared model, and the quantification technique is characterized by increased objectivity and precision.