Designing Role-Based Access Control Without Permission Drift
How modeling an Access Control Matrix (ACM) as a first-class domain artifact prevents security regressions in enterprise CAPEX approval systems.
When building enterprise financial software, permissions rarely start out complicated. On day one, you have three roles: Requester, Approver, and Admin. Six months later, after audit reviews and regional exceptions, you have fourteen roles and dozens of conditional UI visibility rules scattered across pages.
That slow sprawl is permission drift — when what the interface hides no longer matches what the server enforces.
What Clicked Today
While refining our CAPEX Management System at CIMB Niaga, structuring permissions around a single Access Control Matrix (ACM) changed how we reason about every feature:
- Enforce at the data and workflow boundary, not just the UI. Hiding a button is convenience; validating entity-level access rules and microflow transitions is security.
- Treat approval state as part of authorization. A user might be an authorized requester, yet lose edit rights the moment a proposal transitions from
DrafttoIn Review. - Auditability beats cleverness. Explicit role-to-action tables look verbose at first glance, but during enterprise security reviews they let both engineers and compliance stakeholders verify rules in minutes.
State | Requester | Dept Head | Finance Reviewer | Admin
------------|-----------|-----------|------------------|------
Draft | R / W | R | - | R
Submitted | R | R / Appr | R | R
Approved | R | R | R / Baseline | R / WClean authorization isn't about writing clever conditionals — it's about making invalid state transitions structurally impossible.
Key Takeaway
Whenever a feature request says "only people from team X should see this button," pause and translate it into the Access Control Matrix first. Your future self (and your security auditor) will thank you.
