In complex networks, an attack on a single node can drastically change the communication pattern between other nodes. To investigate this effect, we measure the betweenness centrality (BC) changes after single node removal. Then we construct a corresponding secondary network based on this response characteristic of the network under perturbation. We find that the changes of BC are proportional to the BC of a removed node. We use a minimal spanning tree and a percolation cluster method for network construction and find that the degree distribution of secondary networks follows a power-law distribution in both cases.