Trust domains and names that identify principals
Bind workload names to an authority before using them in policy or audit records.
On this page
The question for today
What you’ll understand
- Distinguish a principal name from evidence that authenticates it.
- Preserve issuer or trust-domain context when comparing subjects.
- Explain why cryptographic provenance does not grant resource access.
A string such as payments-worker does not identify one workload everywhere. Another cluster, tenant, or issuer can use the same string. Infrastructure identity needs a namespace and an authority that binds a principal name to acceptable evidence. Authentication then evaluates that binding; authorization decides what the authenticated principal may do. Keep those layers visible before a friendly service name becomes an overly broad policy rule.
Names live inside an authority boundary
A SPIFFE identifier includes a trust domain and a workload-specific path. The complete identifier matters: a matching path in another domain is not automatically the same principal. Trust domains organize authority over identities and associated trust material. Their names may resemble DNS names, but a spelling or a DNS lookup alone does not establish the verifier's trust relationship. The verifier needs deliberately configured authority information and the relevant identity-document validation rules.
For token-based identities, an issuer-subject pair commonly distinguishes principals that share the same subject text. Record that pair in policy and audit mappings instead of stripping the issuer because it seems repetitive. Namespace changes, mergers, and environment moves require migration rules rather than silent string substitution. A staging workload called the same thing as its production counterpart should receive production privileges only if an explicit authorization decision permits that relationship.
Evidence proves a scoped assertion
An identity document can bind a name to a public key or carry signed claims for a bearer token. Verification must establish the intended issuer, signature or chain, validity conditions, and intended use under the selected format. A valid signature from an unrelated authority does not authenticate the desired principal. Likewise, choosing keys based only on attacker-controlled token fields can defeat the binding. Use established protocol validation and vetted implementations rather than constructing an improvised cryptographic check.
Successful authentication answers a bounded question about presented evidence. It does not show that the workload's current code is uncompromised, that a deployment remains approved forever, or that every request should be allowed. A copied bearer token may authenticate possession of that token without demonstrating the intended process's identity. Key possession and bearer presentation have different threat assumptions. Carry validated principal context forward, then evaluate resource-specific authorization and request constraints separately.
Make namespace collisions a test case
Create test identities with the same subject in distinct authorities and verify that policy does not merge them unintentionally. Include unknown issuers, removed trust domains, and environment-specific names. Audit logs should preserve enough validated context to distinguish those cases without logging secret token contents. Plan for renaming and decommissioning so old names cannot quietly regain privileges when a namespace is reused. Ownership of a service label and ownership of signing authority are different operational responsibilities.
The vectors lab offers only a decomposition contrast. Reset and set both angle 0 degrees and magnitude 2, predict x=2, y=0, then Step. Reset and re-set angle 90, magnitude 2 to predict x=0, y=2 up to floating rounding. The two coordinates are mathematical components, not an authority and subject, and the lab performs no authentication. Use the contrast to explain why dropping one part of a scoped identity loses information even when another visible part remains familiar.
Work it through
Twenty identical subject strings
Five trust domains each register four subject names, reusing the same four subject strings. Assume each full authority-subject pair is a distinct principal and no federation mapping merges them.
There are 5×4=20 scoped principals. Keeping only subject text produces four labels and collapses five distinct authorities behind each label.
A policy granting one subject string without checking its authority can therefore match up to five principals instead of the intended one under this example's naming model.
Preserving the pair restores the distinction, but does not automatically authorize any principal. The resource policy must still select the intended domain, subject, action, and resource under validated evidence.
Names become useful security inputs only when their authority and comparison rules remain attached.
The living notebook
Explore the model
Interactive lab · vectors
Two components, one explicit comparison
Reset, set angle 0°, magnitude 2;predict(2,0). Reset and re-set 90°,2;predict(0,2), allowing floating rounding. This geometric decomposition is not an identity, token, or cryptographic verifier.
- x component
- 1.393
- y component
- 0.975
- Magnitude
- 1.7
Playback traces a fixed vector and its projections. Angle and magnitude set the geometry; the moving marker does not change them. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.
Look forDo not discard authority context when a principal's subject text looks familiar.
Pause & reason
Check your understanding
Your turn
Try it yourself
A policy strips the trust-domain portion from SPIFFE IDs and compares only the path. Explain a collision involving production and staging, then name the separate decision still required after full-ID validation.
Take a moment to make a prediction or work it out on paper. Open a hint when you need one.
Give me a hint
Use identical workload paths under different authorities and keep authentication separate from access policy.
Compare with the solution
Production and staging can each issue a path such as/service/payments under different trust domains. Comparing only the path merges those principals. Preserve and validate the complete authority-bound identity; then authorize the particular requested action and resource. A successfully authenticated staging principal does not acquire production access merely because its workload path is familiar.
Keep the idea
Take it with you
- Principal names need an authority context.
- Validated provenance is a bounded assertion, not universal runtime integrity.
- Authentication and resource authorization remain separate responsibilities.
Read further
Original sources and deeper explanations to follow at your own pace.
Optional progress, saved in this browser when available. Revisit any lesson at any time.
Lesson not marked complete.