THREAT‑AWARE INDUSTRY 5.0 ATTACK GRAPH FRAMEWORK FOR CICIDS2017‑BASED CYBER‑PHYSICAL RISK ASSESSMENT

Abstract

The swift development of Industrial Internet of Things (IIoT), cyber-physical systems, artificial intelligence and smart manufacturing technologies, the cybersecurity issues in Industry 5.0 are growing. In this study, a Threat-Aware Industry 5.0 Attack Graph (TAI5-AG) Framework is proposed for cyber-physical risk assessment from CICIDS2017 dataset. The proposed system consists of data preprocessing, Information Gain-based feature selection, multi-class threat detection, attack graph generation, Time-to-Compromise (TTC) analysis, threat likelihood estimation and restoration factor evaluation, which are all combined to evaluate security risks in interconnected industrial infrastructure. The framework illustrates propagation of attacks between critical assets like smart sensors, PLC controllers, industrial gateways, edge servers and digital twins. Through experimental evaluation, the proposed model is shown to be more accurate (99.12%), precise (99%), precise (99%) and F1 (99%) than the conventional machine learning approaches. The critical attack paths found using TTC analysis had compromise times between 25 and 5 minutes, with the highest threat likelihood score of 0.94 coming from the Botnet attacks, and the highest cyber-physical risk score of 0.91 from the same attacks. In addition, the restoration factor rose from 1.12 in the case of low impact attacks to 2.36 in the case of severe attacks. The results have verified that the proposed framework effectively enhances the threat prioritization, cyber resilience, and proactive risk management of cyber-physical systems in Industry 5.0.