Research Of Intrusion Detection Based On Combination Of Multiple Classifiers

With the popularity of computer networks, the network security receive more attention,IDS(Intrusion Detection System) is an active defense system as the line of network defense. Asnetwork attack mode being diversity, more and more intelligent technology is utilized in intrusiondetection system, particularly the machine learning technology development bring new mentalityfor the intrusion detection method improvement. Semi-supervised learning is the hotspot ofmachine learning but its application in IDS is still very rare.This paper carries on the analysis and research to present intrusion detection system andvariety of methods based on semi-supervised learning, applies semi-supervised classificationtechniques to intrusion detection. First, it introduces the current status of intrusion detectionsystematically, sums up the problems and limitations existing in the current intrusion detection,and looking forward to the future trends. Then, it reviewed the background of semi-supervisedlearning and related technologies, and also elaborated on the Naive Bayes, BP Neural Networkand Support Vector Machine classification method. In view of the current semi-supervisedlearning being usually based on single classifier, since the integration of multiple classifiers hassome incomparable advantages than the former. Combined with the integrated learning this paperproposed a mixture classification model based on semi-supervised learning--NBS (NeuralNetwork+Bayes+Support Vector Machine), and also improved BP neural network and Bayesianclassification algorithm.Finally the writer design the experiment for the improved algorithm and NBS model, theexperiment data carry on the sampling from KDD Cup99’s data set, and takes intrusion detectionrate and the rate of false alarm as the standard to examine algorithm performance. The algorithmanalysis and experimental results show that NBS model could obtain good detection performance,and it has the practical value.