Recent years have witnessed rapid growth of innovative and disruptive Internet services such as video streaming and peer-to-peer applications. As network traffic of these applications continues to grow, it has become a challenging task to understand their communication patterns and traffic behavior of end hosts engaging in these applications. This paper presents a novel approach based on behavioral graph analysis to study social behavior of Internet applications based on bipartite graphs and one-mode projection graphs. Through a vector of graph properties including coefficient clustering that capture social behaviors of end hosts, we discover the inherent clustered groups of Internet applications that not only exhibit similar social behavior of end hosts, but also have similar characteristics in the aggregated traffic. In addition, we demonstrate the usage of the proposed approach in detecting emerging applications and anomalous traffic patterns towards Internet applications.