Abstract: Intrusion Detection Systems (IDSs) rely on feature selection algorithms when selecting the most important features; this has an effect on both the accuracy and the time it takes to classify data. Several of these algorithms deal with a number of classes to classify the data. In this study we will evaluate several methods relating to feature selection which utilise adifferent number of classes of the classification in order to determine the optimal number of classes that deliver the best results basedon two criteria the overall accuracy and the time it takes to completethe classification.We utilised WEKA 3.8.0 software for datamining as well as to analyse two types of datasets which are KDD-CUP and NSL-KDD the datasets are each divided into three types based on (23, 5 and 2) classes. The reason behind choosing these numbers of classes is due to the fact that these datasets are available to the researchers on the internetat no cost.It was observed that through minimising the number of classes in classification algorithms, theresults arehighly accurate while training requires only ashort period of time; moreover, there are fewer selected features therefore the processing time is shorter.
Adil M. Salman and Safaa O. Al-mamory, 2016. An Evaluation Study on Performance Enhancement of Intrusion Detection Systems. Research Journal of Applied Sciences, 11: 910-920.