Current Proceedings on Technology
Yazarlar: Susan M. Al Naqshbandi, Venus W. Samawi, Jaap van den Herik
Konular:-
Anahtar Kelimeler:Intrusion,Fuzzy Logic,Genetic Algorithms
Özet: Intrusion Detection Systems (IDSs) are used to establish if someone has made an intrusion into the network or is trying to make one. Many techniques are available to construct IDS using genetic algorithm, but all are based on a fixed length rule. In this paper, we propose to improve the IDSs by using dynamic length rule with an automatic feature selection. The proposed improvement accounts for the complexity of the data by using two of the most popular methods of soft computing, namely Fuzzy Logic and Genetic Algorithm. For a proper determination of the rule length we apply iterative rule learning based on a fuzzy rule-based genetic classifier. We distinguish five main classes, viz. Normal, User-to-Root (U2R), Probe, Remote-to-Local (R2L), and Denial-of-Service (DoS). The first aim of the paper is to suggest an automatic method for producing rules (chromosomes) of dynamic length. The chromosome length represents number of features involved in the corresponding rule. The second aim is to evolve comprehensible rules that improve the classification rate for each of the five classes. In the paper the performance of the evolved rules is given per class. The obtained results provide the detection rate with regards to the lowest number of features for each class.