Ensemble wrapper filter based feature Selection and stacking model with Significant rule power factor classifier For intrusion detection system

Abstract

An Intrusion Detection System (IDS) is a well-established security mechanism with the purpose of being implemented through Information Technology (IT) infrastructure and computer systems. Machine learning techniques have helped correctly identify the intrusions in IDS. Although there is much work on IDS, still some issues in this area need further attention from researchers. The challenges related to classification techniques include: (1) no single-classification technique is capable enough to detect all classes of attacks due to a higher false alarm rate, low rate of detection, and detection accuracy; (2) existing techniques are not capable enough to model correct hypothesis space of problems and imbalanced data distribution; (3) some existing techniques are unstable in nature such as neural networks that show different results with different initializations due to the randomness inherent in the training procedure; (4) different techniques trained on the same data may not only differ in their global performances, but they also may show strong local differences. This has motivated us to address the new feature selection and classification methods which were introduced in order to detect the attacks accurately. Three major contributions are made in this proposed work which are discussed in the following section. newlineFirst contribution of the work, is the Classifier Ensemble Based Intrusion Detection Systems (CEBIDS) proposed for the IDS in the UNSW-NB15 dataset. Before applying a CEBIDS classifier, we need to balance the samples for increasing the detection rate. This can be performed by oversampling and under sampling. Ensemble sampling framework based on K-means and Synthetic Minority Over-sampling TEchnique (ESKSMOTE) technique. Because of the number of data samples in the minority class is less, ESKSMOTE only clusters the majority class into k sub-clusters newline