Tracking the Evolution of Community in IP Networks

Abstract

Extracting underlying structures and significant communication patterns from Internet traffic data has become increasingly urgent and imperative for network operations and security management. In this paper, we proposed LPCT (Label Propagation based Community Tracking) to track the evolution of community in IP networks. In LPCT, we detect the community and preserve the labels of nodes for each snapshot by LAP (Label Propagation Algorithm), then initialize the labels of nodes as the preserved labels in the next community detection for next snapshot. By this way, we can track the evolution of community through the correspondence between label and community in two consecutive snapshots. We evaluate our method by using a NetFlow dataset collected from a boundary router in an actual environment. Experimental results show that our method outperform than other two community tracking methods (ALPA and CommTracker) in terms of NMI (Normalized Mutual Information) and speed. The NMI of LPCT is 30.6% more than that of ALPA and 50.3% more than that CommTracker. The tracking speed of LPCT is three times as fast as ALPA and twice as fast as CommTracker.