A dynamic intrusion detection system for critical information infrastructure

Abstract

The inflow of cyber-attacks on the services of critical information infrastructure (CII) has necessitated adequate attention to their security, functionality and continuous existence. Current attack detection strategies for CII have challenges of high false positive rate, low detection accuracy and lack of immediate response to the security breach. This work presents a Dynamic Intrusion Detection System (DIDS) to control the flow of cyber-attacks on CII and provide a dynamic response to every detection in real-time using a multiclass support vector machine (m-SVM) learning algorithm. Principal component analysis was used for feature reduction to enhance the classification accuracy of m-SVM. Hyper-parameter-tuning was conducted with a support vector classifier and the lowest cost was obtained at a parameter value (c) of 100 with a gamma value of 0.1. The highest performance accuracy of 97.64% and the detection rate of 99.20% were recorded by DIDS when compared with related approaches. DIDS demonstrated a strong capability for securing CII against malicious activities with comparative advantages in terms of efficiency, accuracy and effectiveness.