Research On The Network Anomaly Detection Model Based On Data Fuse Method

The data fusion method is crucial to the performance of intrusion detectionmodel, and also is one of the research hotspots in the field of network security. Thisthesis research on the data fuse methods and gives a method of combining the roughset with support vector machine to design intrusion detection model. The mainresearch contents are as follows:(1) Research on the basic theory of rough set, analyze the existing positiveregion extension methods and point out their shortages. Aiming at these shortages,this thesis improvises the existing positive region extension methods so that we canget more in line with the cognitive world expansion results. Subsequently, theattribute reduction algorithm is given based on improved positive region extension.(2) Research on the support vector machine related knowledge. And this thesisuses the support vector machine as classifier to design intrusion detection model.(3)In order to improve the performance of the classifier for data processing,including the accuracy of classification and the classification of expenses in time,the data should be preprocessed before being classified. Delete duplicate, redundantand unimportant attributes under remaining the useful information. This thesisproposes an intrusion detection model that combined with rough set and supportvector machine RS-SVM. Use the attribute reduction algorithm to processes of thedata to improve the data quality; and use support vector machine to fusion andclassify the inputted data. Finally, the decision result is given.(4)Experimental results demonstrate the effectiveness of the improved attributepositive region extension method, indicating that the combination of rough sets andsupport vector machine intrusion detection model is more validity than using asingle support vector machine in time performance.