A traffic-aware top-N firewall approximation algorithm

Packet classification is widely used in various network security and operation applications. Two of the main challenges are the increasing number of classification rules, amount of traffic and network line speed. In this paper, we investigate an approximation algorithm for selecting the top-N most frequently matched subset of rules from the original ruleset. The goal is to obtain Top-N rules that covers as much traffic as possible while preserving the dependency relationships. Through simulations, we show that our approaches the optimal while runs in seconds, allowing online adaptation to changing traffic patterns.