Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Threat modelling is a structured way of identifying what could go wrong with a system from a security perspective, so that risks can be addressed during design rather than after release.

What Is Threat Modelling?
Threat modelling is a structured process for understanding how a system could be attacked and deciding how to defend it. It is usually carried out during design, before code is written, when changing direction is cheap. Instead of waiting for a penetration test to reveal problems in a finished product, teams think like attackers early and build protections in from the start.
It applies to almost anything: a new web application, an API, a cloud migration, an AI-powered service, an internal business process, or a change to network architecture.
The Threat Modeling Manifesto, published by a group of practitioners in 2020, sums up the process in four questions:
These questions keep threat modelling practical and stop it from becoming an academic exercise.
Most teams start with a data flow diagram showing external entities (users, other systems), processes, data stores, and the flows between them. The most important element is the trust boundary: any point where data passes between areas with different levels of trust, such as from the internet into a web application, or from an application into a database. Threats cluster around these boundaries, so they deserve the closest attention.
STRIDE, developed at Microsoft in 1999, is the most widely used way of brainstorming threats. Each letter represents a category, paired with the security property it violates:
| Threat | Property violated | Example |
|---|---|---|
| Spoofing | Authentication | Using stolen credentials to impersonate a user |
| Tampering | Integrity | Changing the price of an item in a request |
| Repudiation | Non-repudiation | Denying an action because nothing was logged |
| Information disclosure | Confidentiality | An API returning more data than intended |
| Denial of service | Availability | Exhausting server resources with expensive requests |
| Elevation of privilege | Authorisation | A normal user accessing admin functions |
Other approaches suit different needs. PASTA (Process for Attack Simulation and Threat Analysis) is a seven-stage, risk-centric method that links threats to business impact. LINDDUN focuses on privacy threats, useful for UK GDPR data protection by design. Attack trees, popularised by Bruce Schneier, break an attacker’s goal into the different ways it could be achieved. Many teams also use the MITRE ATT&CK framework and the OWASP Top 10 as checklists of realistic attacks.
Imagine a team designing a password reset feature. Using STRIDE, they might identify that an attacker could request resets for other users’ accounts (spoofing), that reset tokens could be guessed if they are short (elevation of privilege), that the “email not found” message reveals which accounts exist (information disclosure), and that someone could flood the service with reset requests (denial of service). Mitigations follow naturally: long random single-use tokens, short expiry times, identical responses for existing and non-existent accounts, rate limiting, and logging of reset events.
Threat modelling works best as a regular habit rather than a one-off workshop:
Threat modelling is a core activity in secure software development frameworks such as the NIST Secure Software Development Framework and OWASP SAMM. It is also increasingly important for AI systems, where risks such as prompt injection need to be considered from the earliest design stage.
At Privia Security, our Infrastructure and Architecture Review Services include threat modelling of new systems and architectures, so security risks are addressed before they are built in.
Ideally during design, before significant code is written, because design flaws are much cheaper to fix at that stage. It should then be revisited whenever the system changes in a meaningful way, such as adding a new integration, exposing a new API, or moving to a different hosting platform. Many teams build it into regular design reviews.
No. Although it is most associated with software design, threat modelling is equally useful for cloud architectures, network changes, business processes, supplier integrations, and physical security. Any situation where you need to understand how something could be attacked and how to defend it can benefit from the same structured thinking.
Threat modelling is a design activity that identifies potential threats and mitigations before or during development. A penetration test actively attempts to exploit a built system to find real vulnerabilities. The two complement each other: threat models highlight where testing should focus, and test results show whether the planned mitigations actually work in practice.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.