An effective network intrusion detection based on truncated mean LDA

The network traffic data employed to build an intrusion detection system (IDS) is always large with ineffective information, that what decrease it efficiency. To deal with this issue, we need to remove the worthless information from the original high dimensional data by using a feature extraction method. The most famous technique which fulfills this role is Linear Discriminant Analysis. However, this method has an important limitation. The class mean vector employed in this method is always estimated by the class sample average. That is not enough to approximate the class mean, specially with the presence of outliers. In this paper, we suggest to use the truncated mean to estimate the class mean vector in LDA modeling. Many experiments on KDDcup99and NSL-KDD indicate the superiority of the proposed technique.