Attack detection analysis in software-defined networks using various machine learning method

• Published in: Computers & electrical engineering Vol. 108; p. 108655
• Main Authors: Wang, Yonghong, Wang, Xiaofeng, Ariffin, Mazeyanti Mohd, Abolfathi, Masoumeh, Alqhatani, Abdulmajeed, Almutairi, Laila
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.05.2023
• Subjects: Attack detection; DDoS; Deep learning; Feature selection; Machine learning; Renyi joint entropy algorithm; Software denied networking; Deep learning; Feature selection; Attack detection; Software denied networking; Machine learning; Renyi joint entropy algorithm; DDoS
• ISSN: 0045-7906; 1879-0755
• DOI: 10.1016/j.compeleceng.2023.108655

The Software-Defined Network (SDN) provides a more flexible and effectively managed network design for next-generation networking. Network managers can easily manage and regulate the entire network using its programmable central controller architecture. This central controller serves as the focal point for numerous attack vectors due to its centralized structure. However, Distributed Denial of Service (DDoS) attacks against the SDN is the most prominent. The goal of this project is to use a machine learning method to categorize SDN traffic as either attack or normal traffic. Next, the Feature Selection method, such as the Filter-based Fisher score method, Wrapper-based method, and analysis of variables (ANOVA) f-test, is used for finely-granulated detection. Then, a rule-based detection method using the Renyi joint entropy algorithm is employed to detect DDoS attacks on SDN controllers. We manage a public "DDoS attack SDN Dataset" with 23 attributes overall. The dataset includes normal and attack traffic for the Internet Control Message Protocol (ICMP), User Datagram Protocol (UDP), and Transmission Control Protocol (TCP). Except for attributes that specify the target and source machines, the dataset, which contains more than 100,000 recordings, has statistical features such as byte count, duration sec, packet rate, and packet per flow. In the classification process, many classifiers such as Artificial Neural Network (ANN), XGBoost (XGB), Support Vector Machine (SVM), and k-Nearest Neighbor (k-NN) were used. The test results demonstrated the efficacy and efficiency of the suggested strategy using the analysis of variables (ANOVA), which performed better than competing methods across a range of evaluation parameters.