Abstract
An intrusion detection system's main goal is to classify activities of a system into two major categories: normal and suspicious (intrusive) activities. Intrusion detection systems usually specify the type of attack or classify activities in some specific groups. The objective of this paper is to incorporate several soft computing techniques into the classifying system to detect and classify intrusions from normal behaviors based on the attack type in a computer network. Among the several soft computing paradigms, neuro-fuzzy networks, fuzzy inference approach and genetic algorithms are investigated in this work. A set of parallel neuro-fuzzy classifiers are used to do an initial classification. The fuzzy inference system would then be based on the outputs of neuro-fuzzy classifiers, making final decision of whether the current activity is normal or intrusive. Finally, in order to attain the best result, genetic algorithm optimizes the structure of our fuzzy decision engine. The experiments and evaluations of the proposed method were performed with the KDD Cup 99 intrusion detection dataset.
Original language | English |
---|---|
Pages (from-to) | 2201-2212 |
Number of pages | 12 |
Journal | Computer Communications |
Volume | 30 |
Issue number | 10 |
DOIs | |
Publication status | Published - 31 Jul 2007 |
Externally published | Yes |
Keywords
- Fuzzy
- Genetic algorithm
- Intrusion detection
- KDD Cup 99
- Neuro-fuzzy
- Soft computing