A cost sensitive learning algorithm for intrusion detection

In this paper, a novel cost-sensitive learning algorithm is proposed to improve the performance of the nearest neighbor rule for intrusion detection. The goal of the learning algorithm is to minimize the total cost of misclassifications in leave-one-out test. This is important since in intrusion detection systems, the performance of the classifier on test data is usually evaluated by computing the total misclassification cost instead of the number of misclassified patterns. In our approach, the distance function is defined in a parametric form. The free parameters of the distance function (e.g. features and instances weights) are learned by our proposed method that attempt to minimize the average cost per example. The KDD99 dataset is used to assess the performance of the proposed method.