Accuracy Improvement of Anomaly-Based Intrusion Detection System Using Taguchi Method

Abstract

In development of anomaly-based Intrusion Detection System, improving detection accuracy is important. It is considered a kind of optimization problem of the thresholds for dividing normal and potential attack. We applied the Toguchi method, which is a technique for quality engineering, for this purpose. We also use "the standardized signal-to-noise ratio of digital data" approach in order to improve simultaneously the two opposite characteristics, false detection rate and true detection rate. The false detection rate can be more decreased by giving more positive training data. By these techniques, 0.186% or the less false positive rate is attained in real data.