Selection of Relevant Features using Hybrid Artificial Intelligent Techniques and Heuristic Method in Intrusion Detection System

摘要

Speed of intruder's detection in intrusion detection systems is one of the most important factors, which gives the system the ability to minimize intruder activities. The work in this paper aims at improving the detection process by introducing a hybrid of artificial intelligence techniques to provide a good features subset selection process; this hybrid consists of rough set theory, decision tree, and a heuristic methods. The classifier-learning algorithm based on good features selection that maximizing an information measure and providing a good classifier. In this work, we introduce five methods for selecting the relevant feature. KDDCUP99data set used to evaluate the goodness of the hybrid methods abilities of selecting relevant features. Neural network classifier built to differentiate between the different methods used. The experiments design and results analysis is presented. The results show that hybrid of AI methods is better than using each one individual and the hybrid of decision tree and rough set is better than the hybrid of rough set and decision tree. The heuristic method usedin this work provides good results as rough sets and decision tree; but still need improvement.