With the rapid development of Internet technology, social networks show a trend of explosive growth. As the traditional analysis on static networks becomes more and more difficult to achieve satisfactory results, dynamic network analysis has turned into a research hotspot in the field of social network data management. Node betweenness centrality measures the ability of a node to control the shortest paths between other nodes in the graph, which is useful for mining important nodes in social networks. However, the efficiency will be low if the betweenness centrality of all nodes needs to be calculated each time while the graph structure changes frequently. To address the difficult problem of computing node betweenness centrality in dynamic networks, a community based betweenness centrality updating algorithm is proposed in this paper. By maintaining the shortest distance sets between communities and communities, as well as between communities and nodes, the node pairs which are not affected during the dynamically updating process can be quickly filtered out, thus greatly improving the updating efficiency of node betweenness centrality. Experimental results conducted on real-world datasets and synthetic datasets show the effectiveness of the proposed algorithms.