Research On Network Immunization Methods In Complex Networks

Epidemic attack such as worms and viruses spreading in complex networks like theInternet is a serious problem faced by network security researchers. Designingeffective network immunization strategies is important to defend against them. Thisdissertation investigates network immunization methods in complex networks,including four parts: research on the metric for evaluation and design of networkimmunization methods, research on using graph partitioning methods to designnetwork immunization methods, research on using both the network topology andspreading parameters to design network immunization methods and research on theeffects of network topology, immunization node number and spreading parameters tonetwork immunization methods.For the standard model of spreading epidemics: Susceptible-Infected-Susceptible(SIS) model, much research works proposed using the amount of drop in λ forevaluation of network immunization strategies and regard the strategy which canachieve this as the optimal strategy: OPT strategy. By proposing an immunizationstrategy and demonstrating its superiority over the OPT strategy, the author, in Chapter2, points out that the aforementioned metric is not appropriate. The author points outthat the appropriate metric for evaluation of network immunization strategies is usingthe infected node number in the steady state of SIS spreading.In Chapter3, the author designs two immunization strategies based on graphpartitioning methods. The first proposed immunization strategy aims at partitioning thenetwork so that the largest segment of the immunized network contains as few nodesas possible. The second proposed immunization strategy aims to disconnect thedifferent network segment as much as possible. The author compares the proposedstrategies in various simulation scenarios with the widely used High Degree First(HDF) immunization strategy. The author finds the proposed immunization strategiesbased on graph partitioning methods are better than the HDF immunization strategy inall of the simulation scenarios.The Equal Graph Partitioning (EGP) strategy and the "max–Δλ" strategy are tworepresentative network immunization strategies which are designed based only on thenetwork topology the epidemic spreads in. In Chapter4, the author first points out thelimitations of these strategies. The author identifies the critical role of epidemicspreading parameters in the evaluation and design of network immunization strategies.The author proposes the idea of taking not only the network topology the epidemicspreads in but also the epidemic spreading parameters together in designing a networkimmunization strategy. Based on this, the author designs a new network immunization strategy. In all of the simulation scenarios, the new immunization strategy performsbetter than the EGP immunization strategy.Through the research work presented in the previous chapters, the author recognizesat least three ingredients which affect the network immunization methods: the networktopology, the immunization node number and the spreading parameters. In Chapter5,the author researches on the relationship of the three ingredients and the effects onchoosing and designing of the network immunization methods. First, by comparing theEGP method and the HDF method, the author points out that when choosingimmunization methods, considering carefully the network environment and thespreading parameters is necessary. Then, by comparing the EGP method and the "max-Δλ" method, the author points out that one should choose the better immunizationstrategy according to the network topology and the immunization doses at hand.Finally, in Chapter6, the author concludes the whole dissertation.