verify

The purpose of verify is to detect verification errors in your threat model document.

The verify actions allows:

  • uses who are not experts at creating threat models to get feedback on a partially complete threat model document, telling them about issues that need to be resolved before a threat model should be submitted for approval

  • an approver of a threat model to get a degree (but not complete) assurance that a threat model has been populated and is free of various common mistakes/omissions/errors

verify takes the following parameters

  • scheme - the scheme name to use. This defines how to convert the threat model document to the model format.

  • docloc - the document ID of the threat model document (in the document location (e.g. Confluence, Google Doc) specified by scheme)

  • doctemplate - the document ID of the threat model template document (in the document location (e.g. Confluence, Google Doc) specified by scheme)

  • reports - what reports to return in addition to the verification issues

    • none - (the default) do not return any reports

    • assets - return a report showing the controls covering each asset per (in-scope) storage location

    • controls - return a report showing which assets each control covers per (in-scope) storage location

    • all - return all (i.e. assets and controls) reports

Examples:

Using verify with threatware as a lambda, to verify a threat model document in Confluence:

https://<lambda-url>/threatware?action=verify&scheme=confluence_1.0&docloc=123456&doctemplate=67890&reports=all

Note: because the scheme confluence_1.0 has been specified, threatware will access the document with ID 123456 in the configured Confluence location.

Using verify threatware as a CLI, to verify a threat model document in Google Docs:

python3 -m actions.handler verify -scheme googledoc_1.0 -docloc 123456 -doctemplate 67890 -reports all

Note: because the scheme googledoc_1.0 has been specified, threatware will access the Google document with ID 123456.