Intrusion Detection Approach Study On Data Dimensionality Reduction And Support Vector Machine

Intrusion detection is a positive method to cope with the problem of network security. The intrusion detection program based on data dimensionality reduction and SVM(Support Vector Machine) was built in this paper. Feature numerical, normalizing, data dimensionality reduction, feature acquisition, SVM classification were studied. Ultimately, LLE(Local Linear Embedding) algorithm was applying in intrusion detection by comparing the performance of the LLE, PCA(Principal Component Analysis), ICA(Independent Component Analysis) algorithms. With Matlab simulation experiment using KDD99 data, the approach of based on LLE and SVM can achieve higher detection rate, lower false positive rate and false negative rate. An improved LLE algorithm was applying in intrusion detection field for improving the inadequacies of the LLE algorithm. The experiment result shows that it has obvious advantage in the detection time.