A Study On Intrusion Detection Algorithm Based On Rough Set Theory And Improved LS-SVM

Intrusion detection is a kind of active defense technology, which can detect the normal and abnormal network behavior and identify the types of abnormal behavior. It is a supplementary of passive defense technology such as the firewall and a means of network security.This thesis analyzes the characteristics of various intrusion detection methods and proposes intrusion detection algorithm based on rough set and the improved least squares support vector machine in the network environment with the large flow and many characteristics. The algorithm reduces sample attributes by discernible matrix using rough set theory, removes redundant attributes to reduce the dimension of the data samples. It improves the least squares support vector machine by a sparse algorithm, so it can improve the veracity of data sample classification with the sparse characteristic and rapid detection. On the one hand the combined algorithm has the advantages that rough set can reduce the data effectively and the support vector machine can classify accurately, and on the other hand it avoids the poor generalization while the rough set is in the noise environment and overcomes the limitations when support vector machine identifies effective data and redundant data. With the parameter optimization of the kernel function and the training of the samples, the fine classifier of support vector machine is set up to achieve high efficiency and practical test. Experiment results show that intrusion detection algorithm based on rough set and the improved least squares support vector machine has high detection accuracy, low false positive rate and false negative rate and short detection time which show the validity of the algorithm.