Free on every plan

Threat model examples for any infrastructure

Five finished threat models, from a company's own data center to hosting and public cloud. Each has a data flow diagram, two attack trees and a scored risk register. Find the one closest to where your system runs and make it yours.

Where does your system run?

  • No credit card
  • Opens as an editable draft
  • Doesn't use up a Free forest until you save it

What is a threat model example?

A threat model example is a finished threat model of a realistic system that you can study or copy. Each of these five has a data flow diagram, two attack trees and a scored risk register, and they cover an on-premise data center, shared hosting, AWS, Azure with Google Cloud and a managed cloud. Each opens free in ThreatTree as a starting point for your own.

Examples by infrastructure

Each opens free in ThreatTree as a draft you can explore, edit and save. The example apps are only there to make each model concrete. What carries over to your system is the infrastructure and its threats.

On-premise

Ironbridge Corporate data center

A company that runs its IT in its own server room. An edge router and perimeter firewall face the internet, with a VPN gateway for staff and a support vendor. A DMZ holds the WAF, load balancer, mail gateway and public DNS. Behind the internal firewall are the core switch, Active Directory, internal DNS and DHCP, app servers, databases, file server, hypervisors, SAN and backups, plus a staff LAN and a management network.

  • Routers
  • Firewalls
  • VPN gateway
  • Load balancer + WAF
  • Core + access switches
  • Wi-Fi
  • Active Directory
  • Hypervisors
  • SAN
  • Backups
  • SIEM
Data flow diagram of Ironbridge: internet users, remote staff and a support vendor outside; an internet edge with the edge router, perimeter firewall and VPN gateway; a DMZ with the WAF, load balancer, mail gateway and public DNS; a data center with the internal firewall, core switch and a server VLAN holding app servers, databases, Active Directory, internal DNS and DHCP, a file server, hypervisors, SAN and backups; a staff LAN with PCs, Wi-Fi and access switches; and a management network with the jump host, SIEM and network monitoring.
The Ironbridge data flow diagram, as it opens in ThreatTree.

Key risks in the model

ThreatScoreFirst fix
Ransomware reaches servers and backups20 CriticalKeep an immutable or offline backup copy off site
Flat network: any PC can reach every server16 CriticalRoute between VLANs through the internal firewall
An unpatched VPN gateway gives direct access15 HighPatch internet-facing devices within days of a critical advisory
Shared admin password on firewalls and routers15 HighNamed admin accounts through central sign-in with MFA

1 data flow diagram · 2 attack trees · 12 scored risks (3 critical)

Shared hosting

Quillpad AI writing assistant

One hosting plan holding the whole app: the web app, MySQL, uploads and backups side by side, managed through a control panel. Everything outside the plan, such as DNS, payments, email and an AI API, is reached over the internet. The threats centre on the server, the hosting account and the secrets stored with the code.

  • Shared or VPS hosting
  • hPanel / cPanel
  • MySQL
  • Registrar & DNS
  • Site backups
Data flow diagram of Quillpad: a customer, the registrar and DNS, and the founder outside a shared hosting account; inside it the Laravel web app, uploads in public_html, the .env file, MySQL, the hosting panel and site backups; the OpenAI API, Stripe and SMTP email outside.
The Quillpad data flow diagram, as it opens in ThreatTree.

Key risks in the model

ThreatScoreFirst fix
Download .env or a backup zip15 HighServe only the public/ folder; deny dotfiles and archives
Take over the hosting panel or FTP15 High2FA everywhere, SSH keys instead of FTP

1 data flow diagram · 2 attack trees · 12 scored risks (2 critical)

AWS

Lexibase AI document assistant

A single AWS account with a VPC. CloudFront and WAF sit in front, Cognito handles sign-in, ECS Fargate runs the API, and S3, SQS and Lambda move documents through a pipeline to OpenSearch, RDS and Amazon Bedrock. GitHub Actions deploys through OIDC. The threats centre on IAM roles, trust policies and what each service can reach.

  • CloudFront + WAF
  • Cognito
  • API Gateway
  • ECS Fargate
  • S3 · SQS · Lambda
  • OpenSearch
  • Amazon Bedrock
  • RDS
Data flow diagram of Lexibase on AWS: a customer user outside the AWS account; inside it CloudFront and WAF, API Gateway, Cognito, S3, SQS, Secrets Manager, Amazon Bedrock and CloudTrail; a VPC with the ECS Fargate API service, RDS Postgres, OpenSearch vectors and the Lambda ingest worker; GitHub Actions deploying from outside.
The Lexibase data flow diagram, as it opens in ThreatTree.

Key risks in the model

ThreatScoreFirst fix
SSRF via "import from URL" steals the task role15 HighFetch URLs from an isolated worker with no IAM role
GitHub OIDC trust allows any repository10 HighPin the trust policy's sub to your repo and environment

1 data flow diagram · 2 attack trees · 10 scored risks (1 critical)

Azure + Google Cloud

Teamscope People analytics

Two clouds working as one system: the application and Microsoft Entra ID sign-in on Azure, the data platform and Vertex AI on Google Cloud, joined by workload identity federation instead of stored keys. The threats centre on the trust between the two clouds and on who can sign in.

  • Front Door + WAF
  • App Service
  • Entra ID
  • Key Vault
  • Workload Identity Federation
  • Cloud Run
  • BigQuery
  • Vertex AI
Data flow diagram of Teamscope: customer staff and the customer HRIS outside; an Azure subscription with Front Door, App Service, Entra ID, Azure SQL, Key Vault, the HRIS sync function and Sentinel; a Google Cloud project with workload identity federation, Cloud Run, Cloud Storage, BigQuery, Vertex AI and Cloud Audit Logs.
The Teamscope data flow diagram, as it opens in ThreatTree.

Key risks in the model

ThreatScoreFirst fix
A static Google service-account key leaks15 HighFederation only; block key creation by policy
Any Microsoft tenant can sign in15 HighAllow-list each customer's tenant ID

1 data flow diagram · 2 attack trees · 10 scored risks (1 critical)

Managed cloud

Invoicely SaaS invoicing app

The common starting point for a web application: an API, an admin console for support staff and a background worker in one cloud account, with a data tier holding a database and file storage, and a payment and an email provider outside. It isn't tied to one cloud provider, so it adapts to most managed platforms.

  • Web API
  • Admin console
  • Postgres
  • File storage
  • Stripe
  • Email provider
Data flow diagram of Invoicely: a customer and a support agent outside the Invoicely cloud account; inside it a Web API, an admin console and a PDF worker; a data tier with Postgres and file storage; Stripe and an email provider outside.
The Invoicely data flow diagram, as it opens in ThreatTree.

Key risks in the model

ThreatScoreFirst fix
Credential stuffing on login16 CriticalRate limits, breached-password checks, two-factor sign-in
Forged Stripe "paid" webhook10 HighVerify webhook signatures and reject replays

1 data flow diagram · 2 attack trees · 10 scored risks (2 critical)

Building a SaaS product?

Quillpad, Lexibase, Teamscope and Invoicely are also SaaS products. The SaaS page covers them in more depth, including tenant isolation and AI features.

See the SaaS examples

Which one should you start from?

If your system…Start from
runs in your own data center or server room, with your own network equipmentIronbridge
runs on one hosting plan or VPS, managed through cPanel or hPanelQuillpad
runs in AWS, using services such as IAM roles, S3, Lambda or containersLexibase
spans two clouds, or signs in with Microsoft Entra IDTeamscope
is a web app on any other managed cloud, with a database and third-party servicesInvoicely
FAQ

Threat model examples, answered

What teams ask before opening one.

Are these threat model examples really free?

Yes. All five open on every plan, Free included. Each one opens as a draft that is not added to your forests, and does not count towards the Free plan's 3 forests, until you choose to save it.

My system isn't a SaaS product. Can I still use these?

Yes. Choose by where your system runs, not by what the example app does. An internal business application on AWS can start from Lexibase and remove the customer-facing parts. If you do build SaaS, the SaaS threat model examples page goes deeper on those four.

We run a mix of on-premise and cloud. Which one should we pick?

Start from the side that holds the most sensitive systems, then add the other. For example, open Ironbridge and add your cloud services to the diagram with a trust boundary around them. Rename and delete components, add your own, and rescore the risks to match.

Can I edit an example to match my own system?

Yes, that is what they are for. Rename and delete components, add your own, rescore the risks and assign real owners. Opening the same example again takes you back to your copy rather than starting over.

Which frameworks do the examples use?

Every risk is tagged with STRIDE and, where it fits, OWASP Top 10, the OWASP Top 10 for LLM Applications (2025), LINDDUN or CAPEC, with MITRE ATT&CK and MITRE ATLAS technique IDs. All of these are available on the Free plan.

Is an example a substitute for threat modeling my own system?

No. It gives you a realistic starting point and shows what a finished model looks like, but your system has its own data, integrations and decisions. Use it to get to a first version of your own model quickly, then keep it up to date as the system changes.

Your system's threat model, from a finished example.

Free plan, no credit card. Open the example closest to your infrastructure, or start a blank model.