Network intrusion detection is one of the hottest issues in the world. An increasing number of researchers and engineers deal with this problem by using machine learning methods. However, how to improve the identification accuracy of all the attack classes remains unsolved since the dataset is an imbalanced one with high imbalance ratio. This thesis work intends to build a classifier to achieve high classification accuracy. It proposes an undersampling Genetic Algorithm-Support Vector Machine (GA-SVM) method to handle this problem. It applies an undersampling method in GA-SVM. To solve the multiclassification problem with a binary classifier, this work proposes to utilize the undersampling GA-SVM with several classic structures. After adjusting the parameter in genetic algorithm and undersampling ratio in each support vector machine, this work concludes that the proposed undersampling GA-SVM improves the performance of an intrusion detection system. Among its variants, the decision tree-based undersampling GA-SVM offers the best performance.
If you have any questions please contact the ETD Team, libetd@njit.edu.