Specifying and Reasoning about CPS through the Lens of the NIST CPS Framework

Thanh Hai Nguyen, Matthew Bundas, Tran Cao Son, Marcello Balduccini, Kathleen Campbell Garwood, Edward R. Griffor

TL;DR

This paper bridges the gap between high-level conceptual frameworks and rigorous engineering by providing a formal logic-based semantics for the NIST Cyber-Physical Systems (CPS) Framework. By encoding the framework’s ontology into Answer Set Programming (ASP), the authors enable automated reasoning to detect system conflicts, identify vulnerable components, and generate optimal mitigation strategies when things go wrong.

What Problem Does This Paper Address?

Cyber-Physical Systems—like autonomous vehicles, smart grids, and hospital robots—are incredibly complex. To help manage this complexity, the National Institute of Standards and Technology (NIST) developed the CPS Framework (CPSF), which organizes system requirements into "aspects" (e.g., trustworthiness, timing, data) and "concerns."

However, the NIST framework is primarily a conceptual reference; it lacks a mathematical foundation. Without formal semantics, engineers cannot easily automate tasks like:

  • Checking if a specific system configuration satisfies all safety concerns.
  • Identifying "hidden" conflicts between competing requirements (e.g., a security encryption requirement that breaks a real-time latency constraint).
  • Automatically generating a recovery plan after a cyber-attack or component failure.

Methods

The authors propose a formalization that combines three powerful tools:

  1. Ontology (Knowledge Representation): They leverage the existing CPS Ontology (RDF/OWL) to represent the vocabulary of the NIST framework.
  2. Action Language $\mathcal{B}$ (Dynamics): They use this to model how the state of a CPS changes over time when actions are performed.
  3. Answer Set Programming (ASP): This is the "reasoning engine." ASP is a declarative logic programming language well-suited for difficult combinatorial problems and planning.

Key Innovations: * Formal CPS Theory: They define a CPS system as a tuple $(CO, A, F, R, \Gamma)$, where $CO$ is a set of components, $A$ is actions, $F$ is fluents (state variables), $R$ maps components to properties, and $\Gamma$ defines how concerns are satisfied by logical formulas. * Likelihood of Satisfaction (LoS): A new metric that quantifies how well a system state meets its goals, allowing the reasoner to rank different mitigation plans. * Trustworthiness Metrics: A mathematical definition of component trustworthiness based on the ratio of satisfied concerns it impacts.

Datasets

The paper does not use traditional "big data" sets. Instead, it evaluates the methodology using two detailed CPS domain models:

  • Lane Keeping/Assist System (LKAS): A model of an advanced car featuring cameras, situational awareness modules, and batteries, used to test reasoning about cyber-attacks and energy constraints.
  • Smart Elevator Domain: Used to test the scalability of the reasoning engines.
  • Scalability Test: The system was also tested against a large-scale ontology containing over 150,000 triples, 85 classes, and 61,000 individuals.

Key Insights and Ideas

  • Trustworthiness is Quantifiable: Rather than a vague feeling, the authors define a component's trustworthiness as a ratio of the concerns it positively affects versus the total concerns it addresses. This allows for the automatic identification of "weak links" in a system.
  • Weak vs. Strong Non-Compliance: The paper distinguishes between a system that might fail under certain conditions (weakly non-compliant) and one that is guaranteed to fail regardless of the actions taken (strongly non-compliant).
  • Mitigation as Planning: By treating system recovery as a logic-based planning problem, the system can suggest the "best" sequence of actions (e.g., switching to a basic mode, enabling a specific patch) to restore the most critical concerns first.
  • Hybrid Reasoning: The authors show that you can successfully combine "loose-coupled" engines (SPARQL for the ontology + Clingo for the logic) to handle complex CPS reasoning in near real-time.

Key Results

  • Efficiency: For the LKAS and Smart Elevator domains, the system computed answers to complex queries (like generating mitigation plans or detecting non-compliance) in under 2 seconds across various reasoning engines (Naïve, Clingo-Python, DLVHex2, and Hexlite).
  • Scalability: The system successfully processed a massive ontology with 150k triples in under a minute, suggesting it is viable for industrial-scale CPS.
  • Decision Support: The authors demonstrated a prototype visualization tool (using sunburst diagrams) that allows designers to see exactly which "ring" of concerns is failing and why.

Limitations and Future Work

  • User Interface: The current implementation requires knowledge of ASP and ontologies. The authors acknowledge the need for a more user-friendly interface for non-expert designers.
  • Probabilistic Refinement: While they introduced "Likelihood of Satisfaction," they suggest that future work could integrate more granular probabilistic action languages (like Markov Decision Processes) to handle uncertainty more precisely.
  • Real-time Constraints: While 2-second reasoning is fast for design-time, it may still be too slow for some "on-the-fly" autonomous vehicle decisions, suggesting a need for further optimization in real-time execution environments.

Summary updated

Abstract

This paper introduces a formal definition of a Cyber-Physical System (CPS) in the spirit of the CPS Framework proposed by the National Institute of Standards and Technology (NIST). It shows that using this definition, various problems related to concerns in a CPS can be precisely formalized and implemented using Answer Set Programming (ASP). These include problems related to the dependency or conflicts between concerns, how to mitigate an issue, and what the most suitable mitigation strategy for a given issue would be. It then shows how ASP can be used to develop an implementation that addresses the aforementioned problems. The paper concludes with a discussion of the potentials of the proposed methodologies.
BibTeX
@article{nguyen2023specifying,
  author = {Thanh Hai Nguyen and Matthew Bundas and Tran Cao Son and Marcello Balduccini and Kathleen Campbell Garwood and Edward R. Griffor},
  title = {{Specifying and Reasoning about CPS through the Lens of the NIST CPS Framework}},
  journal = {Theory Pract. Log. Program.},
  year = {2023},
  doi = {10.1017/s1471068422000035},
}
Figure