Enhanced Network Security Through Optimized Feature Subset Selection Using GTO Algorithm

Abstract

Several attacks are carried out daily to steal sensitive data or make servers inaccessible. Currently, Optical Burst Switching (OBS) networks are among the most widely used in the world. Hackers regularly resort to Burst Header Packet Flooding (BHPF) techniques due to vulnerabilities in the network architecture. Identifying BHPF attacks prevents server applications from being disrupted or stopped. Our solution comprises three main steps: learning, detection, and diffusion of the model. We used an Extreme Learning Machine (ELM), a highly accurate and fast classifier. We proposed a new feature selection algorithm that combines the Fisher score to calculate variable relevance and the Gorilla Troops Optimizer (GTO) to avoid exhaustive searches. The type of attack is shared using the MQTT protocol to enhance network security. The experimental results show that our approach achieves the best precision while maintaining competitive accuracy, compared to Ant-Tree, Naive Bayes, Nearest Neighbor, Artificial Neural Networks (ANN), SVM with Linear Kernel (SVM-LN), and SVM with Radial Basis Function (SVM-RBF).

Keywords