An analysis of the ways to reduce the vulnerability of networks based on the sequential removal of key elements

Abstract

The research focuses on the metrics that allow assessing the stability of a graph and centrality measures. Their calculation underlies the percolation of the key elements of the graph. The experiment involved methods for calculating the average path length of vertex connectivity, clustering coefficient, and graph efficiency based on graph theory. To determine the optimal network stability metric, the principal component method was used, based on the theoretical provisions of mathematical statistics. In this study, the authors solved two scientific tasks: the main and auxiliary ones. The latter was to review the existing network stability metrics, which allowed them to choose the optimal one. The choice of the metrics was carried out using the principal component method. As a result, the average path length proved to be optimal. The solution of the auxiliary problem enabled the authors to analyze the ways to reduce the network stability based on the sequential removal of key elements, which is the main scientific task of the study. The analysis revealed that the nodes whose importance is expressed based on the measurement of centrality by degree are best suited for reducing the network stability. To estimate the stability of networks, an original complex two-criterion coefficient was developed. The analysis of the ways to reduce the stability was carried out by measuring this coefficient in model and real networks. Thus, testing of the proposed methods confirmed their efficiency and enabled their application in various fields of science and technology, e.g., sociology, medicine, physics and radio engineering.

Keywords

• vulnerability
• network
• principal component analysis
• centrality measure
• graph
• percolation