Abstract: The chaos characteristic of traffic flows in the congested network environment was utilized to develop a novel shared congestion detection method based on a weighted first order localregion arithmetic. In this method, the endtoend sample delay series of network flows are analyzed by the chaos signal process technique to quickly detect whether the network flows pass the common congestion links. The simulation results show that this method has better performance than other related methods in the network environments with long TCP, short TCP, and UDP background flows respectively. The utilization of this technique is very promising in the cooperation congestion control in peertopeer networks and the topology construction of intelligent overlay networks.