Analysis of Classification Techniques for Intrusion Detection System

Abstract: Duplicate and unimportant features exist in dataset will cause a long-term problem in classification of network traffic. The existing duplicate features not only reduce the processing speed of classification but they also prevent the classifier from classifying the data, and also losses the trust of providing accurate decisions especially when working with huge collection of data. By considering all these drawbacks a novel system is designed, this system uses two algorithms FMIFS and FLCFS for feature selection and for the classification of data. Here KDD Cup 99 dataset is used for selecting and classifying of dataset. The LS-SVM classification algorithm is used by the two algorithms and it is evaluated for KDD dataset. The evaluation result shows the most relevant features for classification of the dataset and classifies the dataset by sorting out normal data and attacked data.

Keywords: Intrusion Detection System, FMIFS, FLCFS, Feature Selection, Classification, LS-SVM.