An Integrated Cybersecurity Risk Management (I-CSRM) Framework for Critical Infrastructure Protection

Risk management plays a vital role in tackling cyber threats within the Cyber-Physical System (CPS) for overall system resilience. It enables identifying critical assets, vulnerabilities, and threats and determining suitable proactive control measures to tackle the risks. However, due to the increased complexity of the CPS, cyber-attacks nowadays are more sophisticated and less predictable, which makes risk management task more challenging. This research aims for an effective Cyber Security Risk Management (CSRM) practice using assets criticality, predication of risk types and evaluating the effectiveness of existing controls. We follow a number of techniques for the proposed unified approach including fuzzy set theory for the asset criticality, machine learning classifiers for the risk predication and Comprehensive Assessment Model (CAM) for evaluating the effectiveness of the existing controls.
The proposed approach considers relevant CSRM concepts such as threat actor attack pattern, Tactic, Technique and Procedure (TTP), controls and assets and maps these concepts with the VERIS community dataset (VCDB) features for the purpose of risk predication. Also, the tool serves as an additional component of the proposed framework that enables asset criticality, risk and control effectiveness calculation for a continuous risk assessment. Lastly, the thesis employs a case study to validate the proposed i-CSRM framework and i-CSRMT in terms of applicability. Stakeholder feedback is collected and evaluated using critical criteria such as ease of use, relevance, and usability. The analysis results illustrate the validity and acceptability of both the framework and tool for an effective risk management practice within a real-world environment.
The experimental results reveal that using the fuzzy set theory in assessing assets' criticality, supports stakeholder for an effective risk management practice. Furthermore, the results have demonstrated the machine learning classifiers’ have shown exemplary performance in predicting different risk types including denial of service, cyber espionage, and Crimeware. An accurate prediction can help organisations model uncertainty with machine learning classifiers, detect frequent cyber-attacks, affected assets, risk types, and employ the necessary corrective actions for its mitigations.
Lastly, to evaluate the effectiveness of the existing controls, the CAM approach is used, and the result shows that some controls such as network intrusion, authentication, and anti-virus show high efficacy in controlling or reducing risks. Evaluating control effectiveness helps organisations to know how effective the controls are in reducing or preventing any form of risk before an attack occurs. Also, organisations can implement new controls earlier. The main advantage of using the CAM approach is that the parameters used are objective, consistent and applicable to CPS.