Check an architecture for security risks

The threat model runs fixed checks on an architecture - data crossing a boundary without TLS, a reachable database, a service without sign-in, shared secrets - and turns findings into cards.

In short

The threat model reads your architecture and flags what a security review would ask about. It looks for data crossing a boundary without encryption and a database reachable from outside. It also looks for people reaching a service with no sign-in, and a database or secret store shared by several services. The checks are fixed rules - no AI decides them.

Before you start

• Who: anyone who can open the architecture reads the findings. "Create a card" makes a card on the board you choose. Accepting a risk needs a reason.
• Plan: the full list of findings comes with the Business and Enterprise plans; other plans show the first three. "Explain with AI" needs your own AI key on this workspace.
• First: an architecture with named connections. The checks read the words you wrote.

Steps

1. Open an architecture and choose "Show" and then "Threats". The panel opens and the canvas paints each part with its worst finding. "Review" and then "Threat model" opens the panel alone.
2. Read each finding: its severity (High, Medium or Low), "Why it matters:" and "How to fix it:". Press "Show on the canvas" to find it. In the Outline view, components and connections show a Threats column with the worst open finding about each while the panel is open.
3. Fix it in the design - for example write https or TLS on the connection - and the finding goes.
4. Or press "Create a card" to file the work, or "Accept the risk" with a reason.

Result: "No findings. Every check passed." - or each open finding has a card or an accepted reason.

What it checks

Good to know

• The checks read the words you wrote. An HTTPS connection with no label is flagged until you name it.
• An accepted risk stays listed apart, with who accepted it, when and why, and is not painted on the canvas. "Take back" reopens it.
• "Create a card" makes one card per finding on the board you choose, linked to the design. The finding then names its card, and while Threats is on the canvas shows a small card link on each part and connection it is about - only to people who may open that card.
• "Explain with AI" writes the findings in plain words. It changes no finding and saves nothing.
• The full list comes with a paid plan (Pro, Business and Enterprise); Free shows the first three findings.

Write the protection on the connection A connection labelled "HTTPS" or with the technology "gRPC over TLS" passes the boundary check. Name what is really there.

If something goes wrong

• "N more findings are shown on a paid plan."
• "This needs your own AI key on this workspace." - press "Set up AI", or read the findings without AI.
• "Say why first." - accepting a risk needs a reason.

Related

• /help/designer/start-a-design
• /help/designer/live-match
• /help/organizations/security-and-data-report