Robustness on topology reconfiguration of complex networks: An entropic approach

Abstract

Study on complex networks illustrates systems of real-world in disparate realms that incorporates a range of biological networks to technological systems and has, over the past years, become one of the most important and fascinating fields of the interdisciplinary research center. These complex networks share many topological features such as the small-worldness, scale-freeness, the existence of motifs and graphlets and self-similarity. In most cases, complex and real-networks are very large, and the description and analysis of them in explicit form is often faced with difficulty. We manage to head off aforementioned troubles by examining successful models amongst communication networks in some particular aspects, including important factors such as cost, security, integrity, scalability, and fault tolerant. The last factor is distinctly important for each communication network. Recently, some methods and mechanisms have been proposed to increase and improve the robustness of network by modifying its topology. The rewiring is the mechanism amongst the defensive strategies to increase the resilience of attacked networks in which the affected nodes are disconnected from faulty nodes and, possibly, connect to another profitable node with a specific probability. In this paper, a rewiring mechanism based on Shannon entropy concept is proposed to streamline the complex networks configuration in order to improve their resiliency. Network entropy is a quantitative criterion for describing its robustness and is acknowledged as one of the topological characteristic criteria. In practice, this quantity is related to the capacity of the network to tolerate changes in its configuration under various environmental constraints. We evaluate the network robustness based on the spectrum of degree distribution, heterogeneity, as well as the average size of the largest connected cluster during removing nodes with a sequence of systematic attacks based on the degree, betweenness, and Dangalchev’s closeness centralities. The proposed rewiring strategy is applied over six synthetic networks and six real datasets, and then we verified that through approximately 30% swapping of links, the overall robustness of networks can be reached.