Authors : Oqbah Ghassan Abbas, Khaldoun Khorzom and Mohammed Assora
Abstract: In the last few years, big companies have been depending more and more on Software defined Networking “SDN” to fulfill their needs for programmable networks. But like other networks, SDN has some security issues. Many technologies are used to solve such problems and machine learning is considered one of the best. Machine learning has demonstrated its ability to find data patterns when other technologies failed. This makes it a perfect choice for intrusion detection system “IDS” in general and anomaly-based detection in particular. In this research, we propose a new anomaly-based IDS that benefits from the ability of SDN to provide statistical features about flows that pass through the network and passes these features to a voting system that consists of several machine learning algorithms. This technique gives the system the ability to study the user’s behavior and predict any possible intrusion. The voting system is trained and tested using NSL-KDD and KDDCup99 datasets and the results shows increasing in detection accuracy and decreasing in false positive rate.
Oqbah Ghassan Abbas, Khaldoun Khorzom and Mohammed Assora, 2020. Machine Learning based IDS for Software Defined Networking. Journal of Engineering and Applied Sciences, 15: 3354-3358.