Effective Industrial Internet of Things Vulnerability Detection Using Machine Learning

Abstract

Protecting the industrial internet of things (IIoT) devices through vulnerability detection is critical as the consequences of attacks can be devastating. Machine learning (ML) has assisted several works in this regard, improving vulnerability detection accuracy. Based on established vulnerability assessment, development and performance comparison of various ML detection algorithms is essential. This work presents a description of the IIoT protocols and their vulnerabilities. The performance of the ML-based detection system was developed using the WUSTL-IIoT-2018 dataset for industrial control systems (SCADA) cy-bersecurity research. The approach was validated using the ICS-SCADA and CICDDoS2019 datasets, a recent dataset that captures new dimensions of distributed denial of service (DDoS) attacks on networks. The evaluation and validation results show that the proposed scheme could help with high vulnerability detection and mitigation accuracy across all evaluated datasets.