الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 138 |  |  |
| | | from | 138 | | |
Abstract
Intrusion detection systems are no longer a concern only for traditionally security-conscious organizations, such as military and financial institutions, but for every organization and individual who uses computers. What we are really needed is robust intrusion detection systems that are able to categorize each attack type precisely and take the reasonable action against these attacks. These intrusion detection systems have to be independ on specific patterns of intrusions or specific actions of attackers, in contrary we need autonomous intrusion detection systems that have the ability to learn. In addition, these systems have to be light weight, fast, noninvasive, and have the ability to handle qualitative data. In intrusion detection systems field it is very valuable to get both high detection rate and explainable rules since this can improve our knowledge about the nature of intrusions. We can achieve all of these requirements using a modern learning algorithm called Rough Set Classification in combination with two of the most comprehensive techniques of Evolutionary Computation techniques called Genetic Algorithms and Genetic Programming. Rough Set Classification satisfying the general goal of maximizing the accuracy of the intrusion detection systems while minimizing the associated measurement costs. In addition, it yields both explainable detection rules and high detection rate for attacks. Genetic Algorithms and Genetic Programming are effective algorithms for reduction computation. So it can be used in combination with Rough Set Classification to improve accuracy and reduce the complexity of intrusion detection systems by truncating irrelevant and possibly redundant features of attacks.