Software defined network and graph neural network-based anomaly detection scheme for high speed networks

Abstract

In recent years, the proliferation of Software-Defined Networking (SDN) has revolutionized network management and operation. However, with SDN’s increased connectivity and dynamic nature, security threats like Denial-of-Service (DoS) attacks have also evolved, posing significant challenges to network administrators. This research uses the GraphSAGE algorithm to improve DoS attack detection using SDN and Graph Neural Network (GNN) to address the abovementioned problems. The study further explores the effectiveness of four anomaly detection techniques - Histogram-Based Outlier Score (HBOS), Cluster-Based Local Outlier Factor (CBLOF), Isolation Forest (IF), and Principal Component Analysis (PCA) - to identify and mitigate potential DoS attacks accurately. Through extensive experimentation and evaluation, the proposed framework achieves an better accuracy of detecting the anomalies than one without GraphSAGE model underscoring its potential to strengthen the security of SDN architectures against DoS attacks.