Using Fuzzy C-means Algorithm for Improving Intrusion Detection Performance

Using the fuzzy logic for sampling may be a suitable method for data preprocessing and it improves the efficiency of intrusion detection system. This paper has shown how to use a method based on fuzzy clustering, the training samples will be clustered and separated the inappropriate data from the clusters. Accordingly, the remaining samples are supposed to be very suitable representative of different classes and can have a positive influence on the classification but inappropriate data will not be removed or deleted. In proposed method, the inappropriate data will be labeled with Abnormal Class then in training and test phase we will have one extra class that we called Abnormal. Evaluation of the proposed method is performed by KDDCup99 dataset. Our experimental results indicate that intrusion detection system with the proposed preprocessing has performed better than other systems without preprocessing in the case of classification, precision, recall, f-measure, detection and false alarm rate.