Trike
An open-source framework that approaches threat modeling from the defender's requirements rather than the attacker's imagination. Trike formalises what each actor is allowed to do with each asset, then treats anything outside that as a threat — and assigns an explicit, agreed level of acceptable risk to each.
What it is
Trike is a threat-modeling methodology with a strong emphasis on risk management and requirements. Its distinctive idea is to start from a security model of what should be allowed: which actors may perform which actions (create, read, update, delete) on which assets. Threats are then defined as violations of that model. Crucially, Trike asks stakeholders to assign an acceptable level of risk to each asset up front, so mitigation is measured against a stated risk appetite rather than an open-ended "make it secure".
How it works — the architecture
Trike is organised around an actor–asset–action matrix and a set of connected models:
Define actors, assets, intended actions and rules — and the acceptable risk level the business assigns to each asset.
Map which actors may perform which CRUD actions on which assets — the allowed-behaviour baseline.
Derive threats systematically as violations of that matrix (largely elevation-of-privilege and denial-of-service classes), then attacks against them.
Score and weigh threats against the stated acceptable-risk levels to decide which controls are actually required.
How we audit your systems with Trike
- We make permissions explicit — the actor–asset–action matrix surfaces over-broad access that ad-hoc reviews miss.
- We agree acceptable risk first — so "secure enough" is a defined target, not a moving one.
- We derive threats systematically — from violations of the intended model, reducing reliance on the assessor's imagination.
- For AI agents — the matrix is a natural fit for scoping tool access and preventing excessive agency.
How we implement it
Trike suits teams that want threat modeling tied to access requirements and a stated risk appetite — useful when authorization is the dominant concern, or when you need a defensible record of what risk was knowingly accepted. We often use its actor–asset–action matrix to scope least-privilege for agents and integrations, then complement it with STRIDE for broader threat discovery. Outcomes map to NIST AI RMF, ISO/IEC 42001 and the EU AI Act.
Trike is an open-source threat-modeling framework. This page reflects NexusFinLabs' practice and is general guidance, not legal advice.