Skip to main content

Controlling federated data analysis settings

Joining FLNet does not mean every participant can run arbitrary analysis against your client. A core responsibility of the client is to define what is allowed, for whom, and under which conditions.

What you are controlling

In practice, access management for federated analysis includes questions like:

  • Which external users or groups can see that your client exists?
  • Which metadata is visible for discovery?
  • Which tools or workflow classes are allowed to target your client?
  • Which datasets or cohorts can participate?
  • Which actions require manual approval?

Start with the most conservative setup that still supports your intended collaboration.

  • allow only the minimum metadata needed for discovery
  • restrict execution to explicitly approved users or groups
  • enable federated workflows only for datasets that are prepared and reviewed
  • keep auditability for who requested and who approved what

Expand only when the operational model is clear

Broader openness can be useful, but only after you understand:

  • which tools are trustworthy in your environment
  • how users are authenticated and grouped
  • who is responsible for reviewing requests
  • what monitoring or logging is available after execution

Questions to settle before enabling access

  1. Which user roles are allowed to request analysis?
  2. Which local roles are allowed to approve or reject it?
  3. Which datasets are in scope?
  4. Are there workflows that are safe for discovery but not for execution?
  5. What should happen if a request is incomplete or ambiguous?

Operational best practices

  • document the approval rule, not just the technical setting
  • test access rules with a non-production dataset first
  • review role mappings whenever authentication or realm settings change
  • align tool permissions with the maturity of your local data preparation