Research And Application Of Intrusion Detection System Based On Twin Support Vector Machine

With the rapid development of the information system,the key business in the network was the explosive growth.Because of the openness and loopholes in the Internet,the computer system is exposed to the risk of intrusion,the network security system attracted more and more attention.Intrusion detection,as an active interception technology for passive defense,is becoming a hot research direction.But the traditional intrusion detection technology is the expert system based on rule matching,it is quite weak in the face of new intrusion,so the emergence of a large number of intrusion detection technology based on machine learning.Support vector machines stand out for their good classification effects for small samples and high dimensional data,but it also has three obvious disadvantages,the training time is not satisfactory in the face of a large amount of data,the data balance sensitive and two classification algorithm can not adapt to the requirements of multi classification.This thesis studies a lot of literature and relevant theoretical knowledge,based on a certain understanding of the current domestic research status.The multi classification twin support vector machine algorithm based on fruit fly algorithm,called FOA-TWSVM algorithm,which is based on fruit fly algorithm.The core idea of the algorithm is to adjust two penalty parameters and kernel parameters of TWSVM by using Drosophila algorithm according to the number of positive and negative samples,and train two classified hyperplanes for positive and negative samples,and then it achieve multi classification by OVO TWSVMs.In order to prove the performance of FOA-TWSVM algorithm,this paper uses the KDDCUP99 dataset widely used in intrusion detection field to do experiments,and compares it with traditional SVM algorithm and other multi classification algorithms,and takes the accuracy rate,false alarm rate and training time as evaluation indexes.The experimental results show that the proposed intrusion detection system based on FOA-TWSVM has been greatly improved,it can accurately give the specific type of attack is provided to the user in reducing the training time,proving the validity and feasibility of FOA-TWSVM.Finally,through simulation attack,we simulate the network environment of the system,apply the proposed intrusion detection technology based on FOA-TWSVM to the network security system,and show the simulation experiment with the visual interface.