We consider the problem of identifying battlefield events using sensors deployed in the area. The goal is to alert centralized headquarters about the occurrence of significant events so that it can respond appropriately to the events. We propose a mechanism using which the sensors can exchange information using signatures of events instead of data to save on transmission costs. Further, we propose an algorithm that dynamically generates phases of information exchange based on the cost and selectivity of each filter.We present simulation results that compare the proposed algorithm to other alternatives. Our results show that the proposed algorithm detects events while minimizing the transmission and processing costs at sensors.