Introduction
Process Control Systems (PCS) form the backbone of critical infrastructure, including power grids and water treatment plants. Historically, these systems were "air-gapped" and relied on proprietary hardware. However, modern PCS are increasingly integrated with standard IT networks and protocols. While this connectivity improves efficiency, it introduces a dangerous "cyber dimension" to industrial safety.
The primary challenge addressed in this research is fault diagnosis. When a physical abnormality occurs—such as a tank overflow—traditional tools focus on mechanical failures like stuck valves. In a connected environment, however, the same symptom could be caused by a malicious actor injecting network commands. This paper introduces a formal ontological framework to distinguish between routine mechanical breakdowns and cyber-induced faults.
Methodology: Bridging the Physical and Digital
The authors propose a dual-layer approach that treats cyber and physical elements as equal parts of a unified system.
1. Ontological Modeling
Based on Bunge’s ontological principles, the system is modeled as a collection of "components." Crucially, the ontology treats both physical hardware (e.g., a distillation column) and conceptual cyber entities (e.g., the DNP3 communication protocol or an IP network) as components with specific properties and "lawful states." Any deviation from these states is flagged as an "unlawful event."
2. Impact Digraphs
To understand how faults propagate, the researchers map relationships between components using a directed graph (digraph). If a change in one component affects another, a directed edge is drawn between them. This creates a map of dependencies across the cyber-physical boundary.
3. The Diagnosis Algorithm
The framework utilizes an adapted version of the Guan-Graham fault diagnosis algorithm. By using reachability matrices and level partitions, the algorithm works backward from a detected error to identify the most likely root cause. It can also incorporate "propagation probabilities," allowing operators to prioritize certain diagnostic paths based on experience.
Case Study and Results
The methodology was validated using a laboratory-scale distillation column system. This setup included 12 distinct components, ranging from physical accumulators and valves to the Master Terminal Unit (MTU) and the DNP3 protocol.
The effectiveness of the model was demonstrated through two distinct scenarios:
- Scenario A (Mechanical): A stuck distillate flow valve caused a system error. The algorithm correctly identified the physical valve as the primary component to investigate.
- Scenario B (Cyber): A simulated hacker injected DNP3 traffic to force valves to 100% capacity. Although the physical symptoms (flooding and temperature spikes) were identical to Scenario A, the algorithm successfully bypassed the physical components and identified the IP Network and DNP3 Protocol as the root causes.
Conclusion and Future Work
By treating network protocols as "first-class citizens" within the system ontology, this framework provides operators with critical context during an incident. Instead of wasting time repairing functional hardware, they can immediately pivot to cyber-defense measures.
Future research aims to integrate inputs from traditional IT security tools, such as Network Intrusion Detection Systems (NIDS), directly into the ontology. Additionally, the authors propose exploring "forward propagation" to predict how a compromised digital component might impact physical safety in the future.
Summary updated