Immunization of Networks via Modularity Based Node Representation

Abstract

We propose an approach for immunization of networks via modularity based node representation. Immunization of networks has often been conducted by removing nodes with large centrality so that the whole network can be fragmented into smaller subgraphs. Since contamination is propagated among subgraphs (communities) along links in a network, besides centrality, utilization of community structure seems effective for immunization. However, despite various efforts, it is still difficult to identify true community labels in a network. Toward effective immunization of networks, we propose to remove nodes between communities without identifying community labels of nodes. By exploiting the vector representation of nodes based on the modularity matrix of a network, we propose to utilize not only the norm of vectors, but also the relation among vectors. Two heuristic scoring functions are proposed based on the inner products of vector representation and their filtering in terms of vector angle. Preliminary experiments are conducted over synthetic networks and real-world networks, and compared with other centrality based immunization strategies.