DOMAIN 8 · Software Development Security
Threat Modeling: Turn Design Assumptions into Security Decisions
Model a system and its trust boundaries, develop concrete threat scenarios, and connect mitigations to verifiable requirements.
Begin with a decision, not a diagram
Section titled “Begin with a decision, not a diagram”Threat modeling helps a team reason about how a system could fail under adversarial conditions and choose an appropriate response. A diagram is useful when it makes those decisions easier to examine; producing a polished diagram is not the outcome by itself.
OWASP organizes the work around understanding the system, identifying what could go wrong, choosing responses, and checking their adequacy. NIST’s SSDF also includes risk modeling during design and maintaining the resulting requirements and decisions.12
Set scope and expose assumptions
Section titled “Set scope and expose assumptions”Define the feature, data, actors, dependencies, and outcomes under discussion. A model of one login flow cannot establish the security of an entire organization. Record assumptions such as who controls an identity provider, where encryption keys are held, and whether a background worker can access multiple customers’ data.
Keep the model understandable to the people making the decision. Use a high-level view for shared understanding and additional detail where the consequences justify deeper analysis. OWASP describes data-flow diagrams as a common way to represent components, data movement, and external interactions.1
Follow data across trust boundaries
Section titled “Follow data across trust boundaries”A trust boundary marks a change in assumptions about identity, privilege, control, or validation. Follow a request through the browser, API, queue, worker, and data store. Consider what each component knows and which information it merely accepts from another party.
In a fictional export service, a browser submits a customer identifier to an API, which queues a job for a shared worker. The key design question is how the worker obtains an authoritative customer context. Trusting the submitted identifier because the browser hid the field would leave an important assumption untested.
Use prompts to develop specific threats
Section titled “Use prompts to develop specific threats”STRIDE prompts consideration of spoofing, tampering, repudiation, information disclosure, denial of service, and elevation of privilege. It is one technique, not a requirement to force every scenario into one category.1
For the export service, write a concrete scenario: a customer alters a queued job’s customer context, causing the worker to export another customer’s records. Record the entry point, necessary conditions, affected information, and business consequence. This is more actionable than a list containing only “tampering” and “data breach.” The scenario is an original teaching example, not a reported vulnerability.
Turn responses into requirements
Section titled “Turn responses into requirements”For this example, candidate responses include deriving customer context from trusted server-side state, binding it to the job, checking authorization at execution, limiting worker access, and recording the decision. The team must decide which controls fit its architecture and what residual risk remains.
NIST SSDF PW.1.1 supports risk modeling; PW.1.2 addresses maintaining requirements, risks, design decisions, and justified exceptions. A threat record should therefore lead to an owner, a treatment decision, and evidence that the selected design addresses the scenario.2
| Threat record element | Illustrative export-service entry |
|---|---|
| Assumption | Job customer context is authoritative |
| Failure | A caller substitutes another customer’s context |
| Treatment | Bind trusted context and check access at execution |
| Verification | A tampered job is rejected without producing another customer’s export |
Review changes and verify the model
Section titled “Review changes and verify the model”Review whether planned mitigations were actually implemented and whether tests cover the relevant boundary. NIST’s SSDF design-review practice connects the design with security requirements and risk information. Revisit decisions when a new integration, privilege, data type, or deployment model changes the assumptions.2
Suppose the export feature later allows scheduled reports sent to external addresses. The old model did not consider that recipient relationship. Reusing its diagram without analyzing the new delivery path would preserve documentation while missing the changed risk.
Common misunderstandings
Section titled “Common misunderstandings”- “Threat modeling is vulnerability scanning.” It can analyze a proposed design before executable software exists.
- “A STRIDE list proves coverage.” Useful coverage depends on the system model and concrete scenarios.
- “The security team can do it alone.” Developers and business owners often hold essential knowledge about behavior and consequences.
- “A threat model is completed once.” Changed assumptions require renewed analysis.
Connect this topic
Section titled “Connect this topic”Secure development embeds these decisions in the lifecycle. Access control develops the example’s authorization rules, and security testing supplies ways to verify their enforcement.
References
Section titled “References”Footnotes
Section titled “Footnotes”-
OWASP, Threat Modeling Cheat Sheet, overview, system modeling, threat identification, and review sections. Accessed November 21, 2026. ↩ ↩2 ↩3
-
NIST, SP 800-218, Secure Software Development Framework Version 1.1, February 2022; Table 1, PW.1.1, PW.1.2, and PW.2.1. ↩ ↩2 ↩3
Independent, unofficial study notes. Not endorsed or reviewed by ISC2. CISSP is a registered trademark of ISC2.
Editorial policy & attribution