Research On Intrusion Prevention Based On Semi-Supervised Fuzzy Clustering

As the application of the computer and network technology is popularized more and more, various kinds of network security issues have become increasingly obvious. Traditional network security technologies are data encryption, access control, firewall and intrusion detection system, but these methods have some disadvantages. Intrusion prevention system is one of the network security technologies in recent years,it is paid great attention by researchers and becomes a hot research topic in the information security field.By studying the intrusion prevention system, an intrusion prevention system model based on semi-supervised clustering is presented in this paper. It includes data packet getting module, intrusion prevention module, message register module and central control module. Function of every module is presented. The detection algorithm of intrusion prevention module is the key in the intrusion prevention system.Traditional cluster is one kind of hard division method. Each object was strictly divided into one class. It reflects the nature of one or the other. The result of fuzzy cluster is that the uncertainty degree which samples belongs to various classes. It expresses the intermediary generic of samples and makes the result of cluster be more realistic. It becomes the mainstream in cluster analysis field. In this paper.fuzzy clustering algorithm based on data weighting is proposed by optimizing objective function of traditional fuzzy clustering. It makes the cluster centers update according to the fuzzy membership and the data, it also increased the optimization of the fuzzy exponential factor, which makes the clustering results more reasonable.Semi-supervised learning is one of new research of many hot topics in machine learning field, which attains joint probability distribution of labeled data and unlabeled data to improve classifier's performance.An intrusion detection algorithm based on semi-supervised clustering is presented by combining semi-supervised learning with fuzzy cluster. The algorithm instructs lots of unlabeled data to cluster by generating correct sample model using few labeled data. It improves correctness of classification.The semi-supervised clustering algorithm is the key detection algorithm in the intrusion prevention system.The proposed algorithm was simulated with KDD cup 99 datasets. Experiment results indicate that this algorithm has high detection rate with low false positive rate.