TL;DR
This paper introduces a formal framework to analyze whether security policies, combined with physical and digital controls, effectively protect against socio-technical threats in complex systems. They achieve this by modeling information flow and using formal methods to verify policy efficacy against an adversary.
What Problem Does This Paper Address?
Organizations today face sophisticated threats that exploit not just digital vulnerabilities but also the human and physical elements of their systems. Simply implementing security policies (like "clean desk" or "encrypt emails") and physical controls (locked doors, secure storage) doesn't guarantee protection. The critical question is: do these measures, when combined, actually work against real-world adversaries who can exploit the interplay between people, technology, and the physical environment? This paper tackles this challenge by proposing a formal method to rigorously analyze the effectiveness of such combined security strategies.
Methods
The authors build upon existing work on modeling Cyber-Physical Socio-Technical Systems (STPS) using labeled multi-graphs. They extend this model to explicitly represent data flow, introducing "digital objects" and "digital object carriers." The core of their approach involves:
- Formal Modeling: Representing STPS states as graphs where nodes are physical spaces, objects, or agents, and edges represent doors or containment. Labels on these nodes and edges capture properties like object movability, door lock status, and agent behavior protocols.
- Policy Formalization: Defining security policies as specific types of security statements (using a temporal logic language) that constrain system executions.
- Adversary Modeling: Implicitly modeling an adversary by considering system executions that don't adhere to the policy (specifically, by excluding intruder actions from "honest traces").
- Policy Constrained Semantics: Developing a method to derive a new system model (
S|π) that represents the original system with the policy enforced. This involves modifying the transition system to reflect policy constraints. - Algorithm for Analysis: Proposing an algorithm (
Reduce) to transform the original STPS model into the policy-constrained model. This algorithm aims to remove states or transitions that violate the policy.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- The Interplay is Key: The paper emphasizes that security cannot be analyzed in isolation. The effectiveness of policies hinges on their interaction with both digital controls and the physical environment, and how these are exploited by human actors (including adversaries).
- Formalizing "Efficacy": A significant contribution is the formal definition of what it means for a policy to be "effective." They define "affectedness" of a requirement by a policy, distinguishing between requirements whose validity changes due to policy enforcement and those that remain unaffected.
- Honest Traces: The concept of "honest traces" is crucial. These are system execution paths that exclude actions taken by an intruder, allowing for the analysis of how policies constrain the behavior of legitimate users.
- Policy as Constraint: Policies are treated not just as guidelines but as formal constraints on system transitions, effectively pruning undesirable execution paths.
Key Results
The paper presents a formal framework and definitions for analyzing policy efficacy. They define concepts like "affectedness" and introduce a method (Reduce) to derive a policy-constrained system model. While they don't present extensive empirical results, they propose that their framework can be used with model checkers (like PRISM) to verify security properties. They also provide a theoretical complexity analysis for their Reduce algorithm, suggesting a worst-case time complexity of O(|S|² · check(π)).
Limitations and Future Work
- Algorithm Efficiency: The authors acknowledge that their
Reducealgorithm is inefficient and suggest that more optimized approaches are needed. - Policy Language Expressiveness: The current policy language is kept simple for conceptual clarity. Future work will explore extensions to manage more complex, real-world policies.
- Probabilities and Costs: The current framework does not incorporate probabilities or costs associated with actions or attacks. The authors suggest this as a direction for future research, which would allow for more nuanced analysis of intruder strategies and policy trade-offs.
- Practical Experience: While they mention running proof-of-concept scenarios with PRISM, detailed reporting of practical experiences is deferred to future work.
Summary updated