A Kind Of Network Intrusion Detection Research Based On Rough Sets And Genetic Algorithm To Improve The SVM

The progress of science and technology provides a goodenvironment for development of Internet, at the same time, networksecurity operation become the bottleneck of the development.Intrusion detection as a security network security is a kind ofeffective method, the performance of the network safety perfectsignificance. Intrusion detection is essentially classification, instatistics, machine learning, neural network and expert system, dataclassification has been widely applied further study. Support VectorMachine (Support Vector Machine, SVM) is put forward in1992byVapnik a Machine learning method. SVM is based on VC dimensiontheory and structural risk minimization on the basis of a new dataclassification method, in the treatment of high dimensional data,nonlinear data show a good classification characteristic, therefore,was the widespread concern and become a machine learningclassification algorithm in the typical representative of. Highdimensions and multiple attribute is data of natural property, at thesame time for classifier speed and accuracy have great influence.In order to improve the performance of support vector machine(SVM) classification and generalization ability, this paper mainlystudies the genetic algorithm based on rough set and improved support vector machine (SVM), combined with the classificationstandard database kdd99data sets of the theoretical analysis andexperimental research, in view of rough sets and genetic algorithm,the advantages of the design based on rough sets and geneticalgorithm of support vector machine (SVM) classification model forclassification of performance analysis and research, and achievedvery good classification effect.The main work and achievements in the following aspects:First, analyzed the current situation of network intrusiondetection the related technology, focusing on support vector machine(SVM) classification technology development situation is discussed.After analysis and support vector machine (SVM) classification oftraining data, the existing high dimensional sex and multiple attributeproblem processing method, most cases without considering the pooror consideration, the classification performance cause certaininfluence, especially in the classification of time and classificationaccuracy.Then, this paper introduces the rough set theory and geneticalgorithm to the relevant basic knowledge. Based on rough sets andcharacteristics of genetic algorithm, and found that the rough set inthe dimension reduction which great advantages; and the geneticalgorithm in dealing with multiple attribute problem effect is better.On this basis, puts forward the genetic algorithm based on rough setand improved support vector machine (SVM) classification method.Secondly, the research based on rough set theory and geneticalgorithm improved support vector machine (SVM) method. The roughset for dimension reduction, so as to remove invalid data to speed up the training speed, Then genetic algorithm is applied attributereduction, to remove the effect of independent attributes, and speedup the training speed, and at the same time, the classificationaccuracy is improved, so the classification performance is greatlyincrease.Finally, the algorithm used in standard kdd99data concentration,part of the experimental analysis shows that this algorithm in networkintrusion detection speed and precision on the increase.