Conflicts — Roles
The Conflicts — Roles screen lists the distinct pairs of roles that produce an SoD conflict on a connected application. One line per (Application, Role 1, Role 2) triplet, derived from the Details dataset.
Where Conflicts — Users drives short-term remediation (revoke or deactivate), Conflicts — Roles drives the structural fix: redesign the role pair so the combination cannot occur at all.
At a glance
Goal of the view
For each connected application:
- Spot the structurally incompatible roles. A pair appearing here means any user holding both will be in conflict. Trimming the role design once closes the issue for every current and future holder.
- Prioritise the redesign. Pairs producing many Details rows are the highest-impact fix. Use the Summary or Details counts to rank them.
- Plan the communication. The role pair is the unit of conversation with the security administrator and the role owners — far more actionable than discussing individual users.
The query simply distinct-counts the (CFD_APPS_ID, CFD_ROLE1_ID, CFD_ROLE2_ID) tuples in SOD_CONFLICT_DETAILS, sorted by Role 1.
Columns
| Column | Source | What it tells you |
|---|---|---|
| Application ID | CFD_APPS_ID — application identifier. | The connected application. |
| Role 1 | CFD_ROLE1_ID — role granting Activity 1. Looked up against the role catalog. | First half of the incompatible pair. |
| Role 2 | CFD_ROLE2_ID — role granting Activity 2. Looked up against the role catalog. | Second half of the incompatible pair. |
Tips & best practices
- Compare each pair with Roles matrix — if the same pair appears in a high-count combination, the conflict is broadly distributed and worth fixing structurally.
- The pair is symmetric. If
(A, B)appears, the redesign discussion is the same as for(B, A). Treat them as a single decision. - Three structural fixes are common: split one role into two, narrow one role's object scope, or move an object out of one role and require explicit user-level grant (with documented justification).
- Re-run the SoD scan after the redesign — proven and theoretical counts should both drop. A drop only in Details but not in Proven points to compensating controls that may still be needed.