Research on the Semi-Supervised Fuzzy Clustering Algorithm with Pariwise Constraints for Intrusion Detection

Traditional FCM algorithm has the problems of sensitivity to initialization, local optimal and the Euclidean distance is only applied to handle the dataset of spatial data structure for the super-ball. Hence a semi-supervised Fuzzy C-Means algorithm based on pairwise constraints for the intrusion detection is proposed. The pairwise constraints can be used to improve the learning ability of the algorithm and the detection rate. The KDDCUP99 data sets were selected as the experimental object. The experiment result proves that the detection rate and the false rate can be more efficiently improved by the semi-supervised FCM clustering algorithm than the traditional FCM algorithm.