Presenting a new method of intrusion detection in computer networks by classification and feature selection methods

Intrusion detection systems have special importance in computer networks. In present situation, we need high accuracy and relatively good speed systems. We can improve these types of systems by using machine learning algorithm and data mining tools. The purpose of intrusion detection is to recognize unauthorized use, misuse or damage to systems and computer networks by both groups of internal users and foreign attackers. In detecting misuse, recognized intrusion patterns are used for intrusion detection. While in detection methods of unusual behavior, usual behavior of users is considered and as a result each different behavior with it is recognized as attempt for intrusion to system. Decision tree is one of the most famous and oldest methods of data mining to construct classification model. In classification algorithms based on decision tree, external knowledge is presented as a tree from different states of features. Decision trees are considered because their results are interpretable and they do not need to input parameters. Also processing their structures is relatively fast and flexible. Efficiency of a system severely depends on selection method of features. Since by increasing features number, computing cost of a system is also increased, design and implementation of systems seem necessary with the least number of features.