Identity & Access · Chapter 5
Privileged access — why admin is different
The single highest-value control surface in an enterprise, and what PAM products actually sell.
A privileged account is one that can change the system rather than merely use it — domain admin, root, cloud account owner, database superuser, the CI/CD service account that can deploy anything. These deserve a separate regime, because compromising one converts an intrusion into a total loss.
What PAM tooling actually does
- Vaulting — credentials stored centrally, never known by the human. The oldest capability and the least interesting.
- Rotation — automatic change after use, so a captured secret has a short life.
- Brokered sessions — the admin connects through the PAM system, which injects the credential and records the session. The human never handles the secret.
- Just-in-time elevation — the important one. No standing privilege at all; access is requested, approved, granted for a window, and expires automatically.
- Session recording — keystroke or video capture of privileged sessions, which is as much a deterrent and an evidence source as a preventive control.
CyberArk and HashiCorp Vault are not the same product
| Centre of gravity | Typical buyer | Primary use | |
|---|---|---|---|
| CyberArk | Human privileged access | Security / IAM team | Admins reaching servers, network kit, databases; session recording; compliance evidence |
| HashiCorp Vault | Machine and application secrets | Platform / engineering team | Applications fetching database credentials and API keys; dynamic short-lived secrets; encryption as a service |
They overlap and both vendors have grown into the other's territory, but the distinction holds in practice and is worth stating precisely in an interview. CyberArk's classic problem is the human administrator; Vault's classic problem is the application that needs a database password at runtime. Vault's most distinctive capability is dynamic secrets — it creates a credential on demand with a short lease, so the credential your application uses did not exist an hour ago and will not exist in an hour.
Where PAM programmes actually fail
Not in the technology. They fail because the vault is deployed, the admins keep their standing accounts “for emergencies”, and eighteen months later half of privileged activity still bypasses the system entirely.
So the metric that matters is not “is PAM deployed?” but “what percentage of privileged sessions went through it last month, and what is the trend?” Ask for that number. Very often nobody has computed it, and that is itself the finding.