International Conference on Cyber Security and Computer Science

CLASSIFICATION BHP FLOODING ATTACK IN OBS NETWORK WITH DATA MINING TECHNIQUES

V.N. UZEL E. SARAÇ EŞSİZ

Abstract

oday, almost everything is done through networks. Especially, Networks are widely used for transportation of data. Various methods are used to move the data from one place to another. One of these methods is Optical Burst Switching (OBS). When carrying data in OBS, some of the threats may be encountered as a result of security shortcomings. Some of these threats are Spoofing, Replay Attack, Circulating Burst Header Attack and Burst Header Packet (BHP) Flooding Attack. Detection of threats is difficult but it is very important to our safety. Therefore, using Machine Learning (ML) methods to detect threats will give us flexibility, time and accuracy. In this study, we will classify BHP Flooding Attack data that have four class labels with ML methods. Our class labels are as follows: Misbehaving-Block (Block), Behaving-No Block (No Block), Misbehaving-No Block (NB-No Block), and Misbehaving-Wait (NB-Wait). Methods used in classification are Decision Tree (J48), Logistic, Multilayer Perceptron (MLP), Random Tree (RT), Reduce Error Pruning (REP) Tree and Naive Bayes (NB). Since there are 22 properties in the data set, the results of feature selection are also examined using the same classification methods. As a result, J48 and RT have been found to achieve the best results with 100% accuracy.



Conference
International Conference on Cyber Security and Computer Science
Keywords
Machine Learning Data Mining Network Attacks Optical Burst Switching (OBS) Network Burst Header Packet (BHP) Flooding Attack

Language
English

Subject
Computer Science

Full Paper (PDF)

162 views
348 downloads