COMPARATIVE ANALYSIS OF FIREWALL RULE SET USING CLASSIFICATION ALGORITHMS

Abstract

This study focuses on comparative analysis of firewall rule set using classification algorithms based
on the fundamental concept of data mining to evaluate the accuracy and performance of several
classification algorithms. Rule sets grow to large numbers written by different network administrators.
This condition will cause increase the rule set policy and complexity poses problem among other
inconsistencies in the firewall configuration. This led to firewall poses overload and used high
process performance. The Knowledge Discovery in Database (KDD) is adopted as research
methodology to illustrate how this study was conducted. In this study, classification algorithms
namely JRIP, J48, Naïve Bayes, Random tree and Random forest were used for the classification of
dataset. Waikato Environment for Analysis Knowledge (WEKA) was used in comparing these
algorithms. Two firewall dataset were used, KUIPSAS 1098 dataset and PSDC 1024 dataset as
training and testing data on different classification algorithms. The experiment used dataset that have
been formatted into ARFF 10 folds cross validation and the results were compared for accuracy.
Based on the comparative analysis, it can be concluded that using two different datasets from
different sources indicated that the Random Tree algorithm shows the best performance in terms of
accuracy which are 99.70% for PSDC and 99.80% for KUIPSAS.