Combating imbalance in network intrusion datasets

Daniel Cieślak, Nitesh V. Chawla, Aaron Striegel

2006 · 281 citations · 16 references

Concepts

TL;DR

Network intrusion detection systems rely on machine learning and data mining, yet many suffer from high false alarm and missed intrusion rates. The study aims to improve intrusion detection rates while reducing false positives by applying rule‑learning with RIPPER to highly imbalanced datasets. The authors employ RIPPER as a rule classifier, augmenting it with oversampling (replication, synthetic generation, and clustering‑based synthetic instances) and undersampling, and evaluate these techniques on two Notre Dame traffic‑derived intrusion datasets. ROC analysis demonstrates that synthetic minority oversampling outperforms replication and RIPPER’s loss ratio method, and the clustering‑based approach yields further improvement in intrusion detection.

Abstract

An approach to combating network intrusion is the development of systems applying machine learning and data min- ing techniques. Many IDS (Intrusion Detection Systems) suffer from a high rate of false alarms and missed intrusions. We want to be able to improve the intrusion detection rate at a reduced false positive rate. The focus of this paper is rule-learning, using RIPPER, on highly imbalanced intrusion datasets with an objective to improve the true positive rate (intrusions) without significantly increasing the false positives. We use RIPPER as the underlying rule classifier. To counter imbalance in data, we implement a combination of oversampling (both by replication and synthetic generation) and undersampling techniques. We also propose a clustering based methodology for oversampling by generating synthetic instances. We evaluate our approaches on two intrusion datasets — destination and actual packets based — constructed from actual Notre Dame traffic, giving a flavor of real-world data with its idiosyncrasies. Using ROC analysis, we show that oversampling by synthetic generation of minority (intrusion) class outperforms oversampling by replication and RIPPER's loss ratio method. Additionally, we establish that our clustering based approach is more suitable for the detecting intrusions and is able to provide additional improvement over just synthetic generation of instances.

References

16