"Can a customer see another customer's invoices if they change the ID in the URL?" A tester asks this three days before launch, and the answer is yes. The fix touches every endpoint in the billing service. Ten minutes at a whiteboard during design would have caught it, which is the whole case for threat modelling.
Most product teams know they should do it and don't. It sounds like a specialist job that needs a security architect, a week of documents and a tool nobody can drive. So features ship with design decisions nobody looked at from an attacker's side: a webhook that trusts any caller, an internal API with no authorisation check, an admin action with no audit trail.
None of that needs a week. With a tight agenda, the right five to eight people and STRIDE as a prompt, a product team can produce a useful model in an afternoon. This is the agenda we use.
Four questions and a whiteboard
Threat modelling is a design activity, so it pays off before code exists. The Threat Modeling Manifesto frames it around four questions: what are we working on, what can go wrong, what are we going to do about it, and did we do a good enough job? One of its principles is that the best use of threat modelling is to improve security and privacy through early and frequent analysis.
The OWASP Threat Modeling Cheat Sheet turns that into a repeatable process: decompose the application, identify and rank threats, decide mitigations, then review and validate. It recommends starting during design and maintaining the model alongside the system as it changes.
The pressure to do this has gone up. AI coding tools let teams build features faster than anyone can review the design behind them. And enterprise customers and insurers increasingly ask for evidence of a secure SDLC, not just a clean pen test report.
STRIDE threat model basics
STRIDE lets people without deep security training ask "what can go wrong?" in a structured way. Microsoft's Threat Modeling Tool documentation describes the tool as a core part of its Security Development Lifecycle, designed with non-security experts in mind, and says STRIDE categorises threats and simplifies security conversations.
| Letter | Threat | Property at risk | Question for your team |
|---|---|---|---|
| S | Spoofing | Authentication | Could someone act as another user or service? |
| T | Tampering | Integrity | Could data be changed in storage or in transit? |
| R | Repudiation | Non-repudiation | Could a user deny an action and we couldn't prove otherwise? |
| I | Information disclosure | Confidentiality | Could someone read data they shouldn't see? |
| D | Denial of service | Availability | Could someone make this unusable for others? |
| E | Elevation of privilege | Authorisation | Could a normal user gain admin or system rights? |
The OWASP cheat sheet suggests applying the categories to each component, data flow and trust boundary and recording which ones each threat affects. That's what makes an afternoon workable: walk the diagram and ask six questions at each stop.
Who's in the room
Five to eight people. More than that and the session turns into a meeting about the meeting.
- The product owner for the feature.
- A tech lead or architect, plus one or two engineers who'll build it.
- Someone from QA. Testers are often best at spotting abuse cases.
- A facilitator: a security engineer or an experienced tech lead who isn't also the main designer.
- Optionally, someone from support or operations who knows how real users and incidents behave.
Beforehand, the facilitator collects any existing architecture diagram, the API list (an OpenAPI spec is ideal), the data the feature stores and the third-party services involved. Book a room with a big whiteboard or a shared canvas in Miro or draw.io. Laptops closed, except the scribe's.
A half-day threat modelling agenda
About three and a half hours, with a break.
- Scope and goals (15 minutes). Pick one feature, service or integration. Write down what you're protecting: customer data, payments, admin functions.
- Draw the system (45 minutes). Build a data flow diagram together with external entities, processes, data stores, data flows and trust boundaries, the elements the OWASP cheat sheet highlights. Accuracy beats neatness.
- Mark the trust boundaries (15 minutes). Circle every point where data crosses from less trusted to more trusted: browser to API, API to database, your system to a third party, staff tools to production.
- Break (15 minutes).
- Walk the diagram with STRIDE (60 minutes). At each boundary and flow, ask the six questions. Capture every plausible threat as one line: what, where, how. No debating fixes yet, or you'll run out of time at the second box.
- Rank (20 minutes). Score likelihood and impact as high, medium or low. Agree the top items.
- Decide responses (30 minutes). For each top threat, choose to mitigate, eliminate, transfer or accept it, the four options in the OWASP cheat sheet. Name who owns each decision.
- Wrap up (10 minutes). Confirm owners, agree when you'll revisit, and ask the fourth question: did we do a good enough job?
From whiteboard to backlog
A threat model that lives in a slide deck changes nothing. Put the output where your team already works:
- Acceptance criteria on existing stories, for example "API returns 403 when the account ID doesn't belong to the caller".
- New stories for mitigations that need real work, such as audit logging or rate limiting on login.
- Test cases, so mitigations are checked on every release. Our article on shift-left testing covers fitting these into the pipeline.
- Accepted risks with an owner and a review date, so the decision stays visible.
Revisit the model when the design changes materially, when you add an integration, or before a major release.
Making it a habit
One workshop is useful. A habit is worth far more. Teams that keep getting value set a simple trigger, such as "any feature that adds a data store, an external integration or a permission level gets a session", and keep the diagram in the repo next to the code (a draw.io file or a Mermaid diagram in docs/threat-model.md works fine).
Threat modelling also feeds the rest of a secure SDLC. Decisions about which packages and services to trust lead straight into software supply chain security, and the controls you find often map to baselines like the Essential Eight for delivery pipelines. For mobile, pair the workshop with the release checks in our mobile app security checklist.
How JRC Software helps
Our secure development and DevSecOps team runs facilitated sessions. The week before, we sit with your tech lead for an hour to draft the diagram, so the workshop starts from something real rather than a blank board. On the day we facilitate the agenda above.
Afterwards you get a written model with ranked threats, agreed responses and backlog items you can import into Jira or Azure DevOps, plus our template and a one-page facilitator guide so your team can run the next one without us. We'll also tell you what not to do: don't try to model the whole platform in one session. Pick the feature with the most sensitive data and start there.
If you'd like to run your first session with an experienced facilitator, book a threat modelling workshop.
Frequently asked questions
What is threat modelling?
Threat modelling is a structured way to find security and privacy problems in a system's design before they're built. A team describes the system, usually with a data flow diagram, identifies what could go wrong, decides how to respond to each risk, and checks the result. It works best early in design and should be revisited when the system changes.
What does STRIDE stand for?
Spoofing, tampering, repudiation, information disclosure, denial of service and elevation of privilege. Microsoft uses STRIDE in its Threat Modeling Tool to categorise threats. Each category maps to a security property, such as authentication, integrity or confidentiality, which gives a product team six practical questions to ask at each part of a system diagram.
How long does a threat modelling session take?
About three and a half hours is usually enough for one feature, service or integration, including ranking threats and agreeing responses. Larger systems are better split into several sessions, each with a clear scope. One long workshop covering everything tends to lose energy before it reaches the riskiest parts.
Do we need a security expert to run a threat modelling workshop?
No. A tech lead who knows the system and follows a clear agenda can run a good session with STRIDE. An experienced security facilitator helps for the first few sessions, for high-risk systems such as payments or health data, and for teaching the team to spot threats it might otherwise miss.
Cyber Security
Book a threat modelling workshop
Tell us where you're stuck and we'll come back with practical next steps.
Get in touch