Research Of Intrusion Detection Based On Classifier Selection Ensemble

With the rapid development of computer network technology, computer networkhas penetrated into various fields and has brought great convenience to the people’s lifeand study, but the attendant is the network security problem, network security problemhas already become the focus of governments, enterprises and the people all of theworld. Intrusion detection technology is the important research field of networksecurity, intrusion detection system is composed of software or hardware and softwarecombination system, placed in the computer network environment, and then collectinformation form the key points of the network for analysis and detection, which canbe identified from external or internal to the computer network threat behavior.Intrusion detection system is an effective supplement to the traditional securitytechnology. However the current intrusion detection systems generally have lowdetection rates. In this paper, the classifiers ensemble technology of pattern recognitionwill be applied to intrusion detection system, in order to improve the detection rate ofthe detection module.Pattern recognition techniques can be applied to the intrusion detection module, itmainly use the attack characteristics of network data packets for detection and recognition. Traditional pattern recognition systems usually use one classifier forrecognition, but the single classifier can not do well for all samples, the accuracy ofmultiple classifier systems will be greatly improved if summarizing the results. Butwhen a large number of classifiers integrated, it will not only take up too much systemresources, but also decline the accuracy. Select part of the classifiers by a strategy oftencan get better classification results. Selection strategies need to consider both theclassifiers’ accuracy and diversity.In this paper we first introduce the development and background of intrusiondetection, the definition and principle of intrusion detection are introduced in detail,and then analyze and compare of different kinds of intrusion detection technologies foradvantages and disadvantages. Then the classifier ensemble and some classifiersensemble methods are introduced in this paper. This paper presents a classifierdiversity measure method and a classifiers ensemble method based on selectingclassifiers by accuracy and diversity. Through the experiments on the test data sets ofKDD CUP99we find that, considering the base classifiers’ accuracy and diversity isthe key for effective integration, apply this method to the detection module of intrusiondetection system can improve the detection rate of intrusion detection, and due to the selection of classifiers ensemble, the number of classifiers has reduced, so thedetection module occupies less system resources.