This paper, presented at the 2017 IEEE International Systems Conference, proposes a shift in how we secure Cyber-Physical Systems (CPS). Rather than playing a reactive game of "whack-a-mole" with individual software vulnerabilities, the authors argue for a top-down, mission-centric approach that treats security as a subset of system safety.
TL;DR
The authors introduce a methodology that combines stakeholder elicitation (the "War Room") with Systems-Theoretic Process Analysis for Security (STPA-Sec) to derive security requirements based on mission success. By focusing on preventing unacceptable mission outcomes rather than just patching component threats, they provide a framework for building more resilient and cost-effective systems.
What Problem Does This Paper Address?
Traditional security for Cyber-Physical Systems—like drones, autonomous vehicles, or industrial controllers—often borrows too heavily from standard IT security. This "bottom-up" approach focuses on identifying and mitigating threats to individual components. However, in complex, tightly coupled CPS, a security breach is often a safety failure in disguise.
The authors identify three major gaps in current practices:
- Mission Agnosticism: Traditional methods don't distinguish between a vulnerability that crashes the system and one that has zero impact on the specific mission at hand.
- Late Implementation: Security is often "bolted on" after the design phase, leading to higher costs and lower effectiveness, as illustrated in Figure 1.
- The Safety-Security Blur: In the physical world, the line between a malicious hack and a system failure is thin. Both can lead to the same catastrophic physical outcomes.
Methods
The proposed methodology consists of two primary phases:
1. The War Room
This is a structured elicitation exercise involving diverse stakeholders: system designers, military commanders, operators, and maintenance technicians. The goal is to define the "Concept of Operations" (ConOps) and identify what constitutes an unacceptable loss for a specific mission. This ensures the security analysis is grounded in real-world operational needs rather than abstract technical threats.
2. STPA-Sec Modeling
The outputs from the War Room are fed into a modified version of STPA-Sec (Systems-Theoretic Process Analysis for Security), which is based on Nancy Leveson’s STAMP model. Instead of looking for "broken" components, STPA-Sec views security as a control problem. The process follows a logical chain:
- Identify Unacceptable Losses: (e.g., "Loss of the UAV in enemy territory").
- Identify Hazards: System states that lead to those losses (e.g., "UAV strays into inappropriate area").
- Analyze Control Actions: Identifying the commands (human or automated) that could lead to these hazards.
- Derive Constraints: Creating specific requirements to prevent those unsafe control actions.
Datasets
The paper does not use a large-scale statistical dataset. Instead, it validates the methodology through a qualitative case study of a small Unmanned Aerial Vehicle (UAV) conducting a tactical reconnaissance mission.
Key Insights and Ideas
- Security is a Safety Problem: In CPS, the ultimate goal of security is to ensure the system continues to provide its expected service (resiliency) and doesn't cause harm (safety).
- Context is King: A vulnerability that is critical for a strategic reconnaissance drone might be negligible for a small farming drone. A mission-centric approach allows for smarter resource allocation.
- Traceability: By modeling the system as a hierarchy of control (from mission requirements down to physical servos), analysts can trace how a high-level mission failure might be triggered by a low-level hardware or software manipulation.
Key Results
Using the UAV case study, the authors successfully derived a set of concrete security and safety constraints. For example:
- Unacceptable Losses: Identified loss of resources (L1), loss of classified info (L2), and loss of control (L3).
- Hazards: Identified "Absence of information" (H1) and "Loss of control in unacceptable area" (H3).
- Derived Constraints: One resulting constraint (CA 1.1) specifies that "The mission planner shall clearly define the area of interest to be congruent with or a superset of the area of operations..." to prevent confusion that could lead to a mission-critical hazard.
Limitations and Future Work
The authors acknowledge that this paper focuses on the elicitation of requirements. The current process is largely manual and qualitative.
Future directions include: * Formal Verification: Implementing these derived constraints into a system model and using formal methods to check if they can be violated. * Closing the Loop: Testing specific security and resiliency solutions (like encryption or redundant control loops) to see how effectively they satisfy the mission-centric constraints. * Automation: Developing tools to help organize the vast amount of natural language data generated in the "War Room" into formal models more efficiently.
Summary updated