Analyzing the performance of intrusion detection model using weighted one-against-one support vector machine and feature selection for imbalanced classes

Imbalanced class is one of the main problems for intrusion detection models that use machine learning methods. The classifiers generally are designed to minimize the global error rates, which have not considered the condition of imbalanced class. The amount of training data for each type of imbalanced attack can cause those intrusion detection models to have high accuracy but can also lead to difficulty in identifying minority class attacks. In this research, we propose an intrusion detection model using a combination of the feature selection method for imbalanced class and weighted support vector machine classifier. We apply a composite performance index in the features selection and the optimization of the weight of the minority class. The experimental result using the NSLKDD dataset shows that this model produces overall accuracy, sensitivity, and specificity that reaches more than 99%, with false alarms below 0.5% and false-negative rates below 0.7%. The sensitivity of the U2R and R2L classes are 56% and 92%.