Active Queue Management (AQM) plays an important role in network congestion control. Most of AQM algorithms aim to achieve flow fairness. However, with an investigation to congestion-cause traffic, the anomalous packets vary in different aggregating methods, not just within limited high-bandwidth flows. Therefore the previous AQM would lost the ground to penalize the misbehaved traffic scattered in distributed flows. In this paper, an escalated and mix matching of flow ID and IP algorithm (named MxCHOKe) has been proposed for various network congestion avoidance based on their statistical behavior. Simulation results suggest it gained an expected performance of protecting well-behaved traffic in more realistic congested situations