2 6 A ug 2 00 3 Epidemics in Hierarchical Social Networks
D. Zheng,P. Hui,S. Trimper,B. Zheng
2003-01-01
Abstract:Epidemiological processes are studied within a recently proposed social network model using the susceptible-infected-refractory dynamics of an epidemic. Within the network model, a population of individuals may be characterized by H independent hierarchies or dimensions, each of which consists of groupings of individuals into layers of subgroups. Detailed numerical simulations reveal that for H > 1, global spreading results regardless of the degree of homophily of the individuals forming a social circle. For H = 1, a transition from global to local spread occurs as the population becomes decomposed into increasingly homophilous groups. Multiple dimensions in classifying individuals (nodes) thus make a society (computer network) highly susceptible to large scale outbreaks of infectious diseases (viruses). PACS numbers: 89.75.Hc, 87.23.Ge, 89.75.-k Typeset using REVTEX Epidemics of all kinds are costly. The influenza in the early 20th century and the recent outbreak of SARS claimed precious lives, and the spreading of viruses like the “Blaster” and “SoBig” on the internet in recent months has led to much economic lost when computer systems of governments and airlines were attacked. Epidemiological processes, together with percolation and problems on searchability, are the most interesting dynamical problems in a system of connected entities [1–3]. There are two important aspects in the study of epidemic spreading, namely how infection and recovery occur and how people or computers are connected. The latter is best described in terms of networks or graphs [2,3] in which the nodes representing people, communities, or computers are connected by links representing the acquaintance among people or the connections between computers. The two popular models for studying epidemiological processes are the susceptible-infected-susceptible (SIS) and the susceptible-infected-refractory (or recovered/removal) (SIR) models [1,4,5]. The most important question is the extent to which a disease can propagate in a population. Since infections take place through direct contact between infected (I) and susceptible (S) nodes, the underlying structure of the network plays a determining role on the spread of diseases. For SIS model, each node takes on one of two possible states. An infected node has a probability ν of infecting a connected susceptible neighbor and has a probability δ of recovering and becoming a susceptible node per unit time. The ratio λ = ν/δ turns out to be crucial. Traditional epidemic models, which ignore the realistic connections between nodes, reveal a threshold for the spread of a disease [1]. Recent work reveal, however, that the underlying connectivity between nodes may lead to qualitative changes in the extent of a spread [6–8]. In particular, it was shown that a threshold of λ does not exist for epidemics in a growing scale-free network with preferential attachments to popular nodes [6], while a threshold exists for growing networks with high-clustering nodes [7]. The SIR model, which was proposed in the 1920’s, has been widely used in different forms for studying the spread of influenza [9] and, most recently, SARS [10], and also for the spread of rumors [11]. In its simplest form, a node takes on one of three states (S, I or R). The R-state represents the case in which an infected node becomes recovered and immunized or, unfortunately