A threat model is a structured account of who might attack a system, how, and which risks the design is meant to stop. It lists the assets worth protecting, the possible attackers and their abilities, the ways in and the defenses in place.

It matters because security measures are easier to judge against a stated purpose. A control that is strong against one kind of attacker may be irrelevant against another, and writing down the assumptions shows which risks are covered and which are accepted.

A threat model is not complete or permanent. It reflects what the authors know at the time, and new attack methods, features and changes in how a system is used can make it out of date. Disagreements about what to include are common. A model that states what it does not defend against is more honest and more useful than one that claims full coverage.

To make one useful, keep it short and specific, include the people and systems that interact with the product, review it when the design changes and test the defenses against the threats it lists.