Research On Network Survivability Based On Community Discovery And Community Evolution

Facing the large scale of network and rich application of networks, it is veryimportant for network can provide essential service in time and in effect, when networksystem suffers attack, failure and unexpected accident. Research on networks survivabilitycan significantly benefit to society and economy.In order to measure and evaluate thesurvivability of network, we introduce the community structure characteristics whichembody survivability by inherent requirement of survivability definition.From the inherent requirement of network survivability. This paper presents a set ofcommunity evolution model to analyze the network survivability. First, we establish therelationship between the security status of the node/communication link and the evolutionbehaviors of community. Then we put forward quantitatively indicators to track theevolution behaviors and these indicators can evaluate the network survivability; Finally,we evaluate the survivability of real communication network and we can track theevolution events of growth, contraction, spliting and merging by carrying out thesimulation experiments.This paper study community discovery based on network survivability in order todiscover the community structure characteristics which embody survivability. First, thetraditional label propagation algorithm (LPA) cannot guarantee the high quality of thecommunity division and stable results of community division, we presents a nodedegree-based improved label propagation algorithm (NDI-LPA). Simulation results showthat compared with the traditional label propagation algorithm (LPA), our NDI-LPA hasbetter quality and more stable results of community division, while the time complexity ofalgorithm is not increased significantly and it is still close to linear complexity O (n+m).