Human vs Machine Identities
People, services, CI/CD, agents and automation need separate identities so access can be scoped, attributed and revoked independently.
Frame the problem
Security starts with a concrete asset, attacker capability and trust crossing.
Why the system fails
Shared credentials erase accountability and force every actor to inherit the broadest required permission.
The important question is not “what is Human vs Machine Identities?” but “which assumption let untrusted data or an over-scoped identity cross distinct actor → distinct credential and policy?” Trace the decision at the boundary, then constrain what can happen after the first control fails.
Design the control in layers
Start with the control closest to the interpretation or privilege boundary: Issue a unique identity per human and workload Then add a control that reduces blast radius and telemetry that proves the decision was enforced.
The resulting design is not labelled secure. Record the identified controls, the known failure paths, the remaining exposure, and the evidence you would need during an incident.
| Prevent | Detect | Recover |
|---|---|---|
| Issue a unique identity per human and workload · Ban shared static production credentials · Bind automation to its environment and purpose | One identity used from unrelated workloads or locations | Contain the affected identity or component, scope impact from audit evidence, and preserve a regression test. |
Key points
- Asset: Attribution and revocation of every privileged action.
- Boundary: Distinct actor → distinct credential and policy
- Primary control: Issue a unique identity per human and workload
- Detection signal: One identity used from unrelated workloads or locations
- Always ask what limits damage when the primary control fails.
Follow the attack
Safe conceptual simulation: capability → missing control → crossed boundary → asset impact.
- 1Attacker starts with: Someone who compromises one person or workload.
- 2Shared credentials erase accountability and force every actor to inherit the broadest required permission.
- 3The weak or missing boundary control is crossed: Distinct actor → distinct credential and policy
- 4Impact: A compromise is hard to contain or attribute and affects every credential user.
- A compromise is hard to contain or attribute and affects every credential user.
Defend, detect, recover
One prevention is a single point of security failure. Layer it and make failure observable.
- • Issue a unique identity per human and workload
- • Ban shared static production credentials
- • Bind automation to its environment and purpose
- • One identity used from unrelated workloads or locations
- • Contain the affected identity or component.
- • Scope access from audit evidence.
- • Fix the boundary and add a regression test.
- • Misconfiguration and new access paths can bypass the intended control.
- • A privileged insider or compromised control plane may still reach the asset.