The crucial problem of ever increasing high traffic encountered by an IDS can be tackled by classifying the network traffic and distributing the analysis among several IDSes ensuring faster detection. Besides, each IDS equipped with only the required functionalities can provide sharper analysis of the traffic. We propose in this paper a new classification algorithm that constructs a Direct Acyclic Graph (DAG) to split the traffic using security policies and IDS characteristics. The method divides different classfication rule features into several bytes and sorts them by considering explicit values before masked one thereby reducing overlaps between rules ensuring smaller DAG and easier way to classify packets during runtime.