Efficient Intelligent Network Intrusion Detection for SDN Using XGBoost

G. Balu Narasimha Rao, Amit Kumar, Nitish Kumar, Prem Raj

Abstract: Networking defined by software (SDN) has arisen as an encouraging path for the internet’s future expansion, providing enhanced flexibility and transparency to centrally managed networks. However, these advantages come with inherent vulnerabilities, posing significant risks such as network disruptions, system paralysis, and cybercrimes like online banking fraud and theft. Addressing these challenges is paramount for organizations, enterprises, and economies. In recent times, the incorporation of clever machine learning techniques goes into systems for detecting intrusions (IDS) through SDN has garnered considerable interest. In this manuscript, we introduce a fresh IDS called XGBoost tailored to SDN environments. Initially, a hybrid feature selection and extraction algorithm are employed to reduce data dimensionality and acquire an optimal feature subset. This approach involves utilizing a correlation-based feature extraction (CFE) algorithm followed by random forest recursive feature elimination. Subsequently, the XGBoost algorithm is employed for the detection and classification of varied forms of attacks. Experimental findings conducted on the dataset NSL-KDD illustrate that the suggested system surpasses current methods in F-measure, recall, accuracy, and precision.

External IDs:dblp:conf/icccnt/RaoKKR24