Research On Method Of Community Tracking Based On Communication Relationship

As an important communication network in real life,IP network has community structure in it.When it is security-supervised,analyzing the network behavior of the entire community is more efficient than analyzing the network behavior of each IP communication node separately.Meanwhile,tracking changes in communities in IP network can help achieve anonymous network discovery and network anomaly detection.This paper takes IP network as the research object and studies the methods of discovering and tracking network communities.The main research contents of this thesis can be summarized as follows:(1)Research on community detection method based on communication relationship:Design and implement a community detection method named CDSCB based on the similarity of communication behavior.This method first introduces the concept of communication behavior similarity,which is used to describe the similarity between IP nodes,and is based on the communication behavior of nodes,can effectively characterize the similarity between nodes and is simple to calculate.Based on the similarity,the method adds all nodes in the network to the community with the highest similarity by iteratively.The experimental results show that this method can obtain better modularity on the public dataset and the modularity is closer to that of real community partitioning results on the real IP network dataset than the other four classical community detection methods.At the same time,CDSCB can also achieve a higher score on the NMI(Normalized Mutual Information)indicator that measures the accuracy of the result of community detecting than the other four methods.(2)Resear-ch on community tracking method based on communication relationship:This paper expands the label propagation algorithm,meanwhile,designs and implements a community tracking method named LPCT which is based on label propagation by using the correspondence between labels and communities and the changing of labels of communities in two consecutive community detecting results.The experimental results show that the method can better track the changes of the community in the IP network comparing with the other two community tracking methods,ALPA and CommTracker.This method can obtain better community detecting results(higher NMI score)when tracking the community,and the processing speed is 2 times and 3 times faster than ALPA and CommTracker,respectively.Finally,this paper designs a ser-ies of experiments to track and analyze the communities in the IP network.The experimental results show that the proposed method can effectively detect the community structures of some network applications and network services,and can track individual or multiple communities separately to obtain the evolution relationship between communities.These results will help anonymous network discovery and network anomaly detection.