Threat Modeling (Navin Security)
Overview
Turn the discovered stack into assets, trust boundaries and attack scenarios before any scanner runs. Default posture is read-only and non-destructive: report evidence and a proposed fix, never exploit, extract data or change the target.
When it applies
- Triggers: always.
- Families: injection, authentication, authorization, multitenancy, code_execution, files.
Method
- List the assets worth stealing or breaking: identities, tenant data, money, secrets, infrastructure, the agent's own tools.
- Draw the trust boundaries: internet to app, user to user, tenant to tenant, app to database, app to cloud, model to tools.
- For each boundary write the attack scenario in one line (who, what they control, what they reach) and tag the controls that would stop it.
- Rank families by asset value times exposure; this order decides what the audit reads first.
- State what the model could not see (no running target, no second tenant) so the report can say it was not covered.
Evidence and status
potential: a rule or a suspicion with static evidence only.confirmed: a dataflow trace from a controlled source to the sink, or a dynamic observation on the running target.exploitable: a non-destructive reproduction showed the impact, in an authorized environment.- Every finding carries file and line or endpoint, the evidence, impact, a concrete fix and the control ids above.
Anti-patterns
- Listing every OWASP category without tying it to an asset of this project.
- Treating a framework default as a mitigation without reading the config.
- Treating a clean scan as proof of security: say which controls were not assessed.
