Firewall Anomaly Detection Based on Double Decision Tree

To solve the problems regarding how to detect anomalous rules with an asymmetric structure, which leads to the firewall not being able to control the packets in and out according to the administrator's idea, and how to carry out an incremental detection efficiently when the new rules are added, anomaly detection algorithms based on an asymmetric double decision tree were considered. We considered the packet filter, the most common and used type of First Matching Rule, for the practical decision space of each rule and the whole policy. We adopted, based on the asymmetric double decision tree detection model, the policy equivalent decision tree and the policy decision tree of anomalies. Therefore, we can separate the policy's effective decision space and the anomalous decision space. Using the separated decision trees can realize the optimization of the original policy and the faster incremental detection when adding new rules and generating a detailed report. The simulation results demonstrate that the proposed algorithms are superior to the other decision tree algorithms in detection speed and can achieve incremental detection. The results demonstrate that our approach can save about 33% of the time for complete detection compared with the other approaches, and the time of incremental anomaly detection compared to complete detection is about 90% of the time saved in a complex policy.