Research On Cascading Failure Modeling And Optimization Technology For Multi-layer Information Network

The interdependent and dependent coupling relationship between multi-layer networks can better represent large networks in real life than a single independent network.However,due to the coupling relationship between multi-layer networks,the robustness of the entire system network will be reduced.And other researches have found that the coupling relationship between multi-layer networks being cascading failures easily will also increase.Once network nodes are attacked and fail,the inter-layer propagation of cascading failures between network layers will occur.Further,the propagation of overload faults within the network layer is triggered by inter-layer fault propagation.Those propagation processes continue to iteratively propagate until the network reaches a steady state due to cascading failures and the network collapses lastly.Therefore,this thesis aims at how to explore the evolution process of cascading failures propagation between networks and recover networks being cascading failures by constructing different topological multi-layer networks.These studies will provide reference value to serviceman of network security to maintain the network security,which is of great significance.Firstly,we construct a BA-BA network and study the effects of cascading failures on BA-BA network under different controllable parameters.The coupling relationship between BABA networks is one-to-one and one-to-multi,respectively.In the cascading failure model established,considering the relationship between the initial load and capacity of nodes,we add an attenuation coefficient ? and exponential coefficient ? to load-capacity model,it can be proportional to node degree,but not completely proportional and achieves the greater degree and stronger capacity,which is different from the existed model in related researches.By experiments,the results show that,whether it is the multi-layer network with a one-to-one coupling relationship or a one-to-multi coupling relationship,the controllable parameters have the same trend in the impact of networks cascading failures.The robustness against cascading failures is positively correlated with attack scale and tolerance parameters,and negatively correlated with redundant parameters.Secondly,for attackers,when they attack networks,network information is collected and the information is analyzed to obtain a breakthrough point for attacking the network.Therefore,this thesis proposes three different attack strategies MD/GHS/GLS to simulate an attacker attacking networks.Study the robustness of multi-layer networks against cascading failures under different attack strategies.In order to better reflect the real network,BA-BA,WS-WS, and BA-WS networks are set up.The experimental results indicate that,in most cases,for those networks,when the network is attacked by GHS,the network is less prone to cascading failures and collapses,at which time the network shows good robustness and has a longer lifetime;when attacking the network with GLS/MD,the network is more easily prone to cascading failures.At this time,nodes with large load carrying capacity play an important role in the network,and the network's robustness against cascading failures is very fragile and its lifetime is shorter.Finally,considering recover networks under cascading failures,this thesis proposes a Re GLS repair strategy based on the GLS attack strategy.According to the types of cascading failures that occur in networks,we divide the failure into full failure repair and partial failure repair to research and propose a repair and recovery process.The results show that,after the recovery strategy is added,when the cascading failure occurs in networks,the network scale attacked is larger than that without the repair strategy.