From risk assessment to action

Prev Next

Darwinium can assess an interaction and help determine what should happen next. To change the customer's experience, connect that assessment to an action in your application, at the CDN, or in an operational workflow.

For example, a login step might produce a Challenge result. Your integration must then start the appropriate verification flow, or execute a configured edge action. The result alone does not launch MFA or stop a request.

Choose where the action belongs

You need to… Use Configure next
Let your backend choose the response Request-header enrichment or an Event API response Map the assessment data and implement the corresponding application behavior.
Intervene before a request reaches the origin Request-time proxy action Define the condition, required data and action before forwarding.
Change a response sent to the client Response-time proxy action or response mapping Define the modification and ensure it fits the client and content type.
Create a case or notify another system Incident workflow or an external API call Define the trigger, credentials, payload and failure handling.

Start with Build your first decision if you do not yet have an assessment to consume.

Send assessment data to your application

Enrich an HTTP request header

For a CDN integration, open Workflows → your journey file → the step → Data Mappings → Request → Header Mappings. Configure an injection for the attribute your application will consume. For example, your backend can receive the event's identifier in a header you name dwn-identifier and use it to correlate the request with an investigation.

If a header carries a model or rule result, make its mapping wait for the workflow that produces that value. Test that the result exists at the point the header is injected. See data mappings and dependencies.

Your backend should trust assessment headers only from the protected integration path. Ensure client-supplied values cannot impersonate them and that alternate routes cannot bypass the intended enforcement. Define how your application handles missing data and timeouts.

Consume an Event API response

The Event API submits an event to a configured API step and returns its assessment according to that step's configuration. Your backend authenticates the request, supplies the mapped context, reads the response and takes the appropriate action. Use the API examples to establish the request path first.

Sending an identifier does not, by itself, request all stored data about a previous journey. Configure any required history, linkage or feedback explicitly. The Update API serves a different purpose: supplying supported updates to an existing event.

Act at the CDN

Open the step's Workflows tab. Under Proxy Actions, choose the phase:

  • Before Request to Proxied Destination: act before forwarding the request to the origin.
  • Before Response to end-user: act on the response sent back to the client.

Configure the action, condition and dependencies. Follow Proxy actions for the supported operations and Digital Speedbump for that mitigation option.

Timing matters. A request-time action can use information available before forwarding. A response-time action can use the origin's result, but cannot undo a payment or account change the origin has already completed. Avoid dependencies that require the response before a pre-origin action can run.

Choose a response appropriate to the client. A browser redirect or interactive challenge may work for a website but not for a machine API client. For MCP traffic, plan protocol-compatible handling with the application team.

Trigger follow-up work

Use incident management for cases your team will work in Darwinium. Use a Marketplace integration or Call URL to enrich an assessment or contact another system. Add conditions so the external operation runs only when required, and configure dependencies when a decision needs its result.

A Review or Alert result expresses a policy outcome. The associated queue, notification or case creation still needs to be configured.

Verify both the assessment and the effect

Deploy to a test environment through the deployment lifecycle. Exercise at least one interaction for each intended result, plus a missing-data or failure case.

In Investigations, inspect the inputs, signals and result. Then check the application or downstream system: was the header received, request stopped, verification started or case created? Capture the event identifier with the test outcome so your team can trace the whole path.