Blog. Access governance.
What a permission map can tell you, and what it cannot
Co-founder and CEO · November 6, 2026 · 4 min read

On this page
What a permission map shows, in four lines
- A permission map shows who could read sensitive data. It does not show that anyone did.
- Role names are labels. Two roles with the same name can hold very different grants.
- Activity lives in a separate record, often in a different system, and only if logging was on and kept.
- For an incident or audit question, keep allowed reach and observed activity in separate columns.
A permission map answers one question well: which identities could reach this sensitive data. Who actually read it is answered by a different record, the source’s audit log, and only where the source keeps one.
Both questions matter. Mixing them up turns a careful review into a confident, wrong sentence in an incident report.
Two roles with the same name hold different reach
Take two database identities in a constructed example. Both are called reporting_ro, and both belong to the analytics team. On a spreadsheet of roles they look identical. Only the grants tell them apart.
| Identity | Database | Allowed reach | Evidence | Decision |
|---|---|---|---|---|
| reporting_ro | Billing | Select on three summary views | Grants read November 9, 2026 | Keep |
| reporting_ro | Customer | Select on the whole schema, including the national ID table | Schema-wide grant made two years ago for one extra column | Downgrade to the named tables |
| etl_load | Customer | Insert and update on staging tables | Grants read November 9, 2026 | Keep |
| dba_owner | Customer | Owner of every table | Role membership read November 9, 2026 | Time-limit and name the approver |
| audit_tmp | Archive | Permissions not available | The source does not expose permissions | Excluded from the map; owner to confirm by hand |
NIST’s guide to attribute-based access control describes privileges as “the authorized behavior of a subject,” defined by an authority in policy or rules. It also notes that failing to remove or revoke access over time “leads to users accumulating privileges.” The second reporting_ro is that sentence in practice.
A permission map answers four questions
- Which users, service accounts, roles and groups can read or change a sensitive table, file or bucket.
- Which rights are wider than the job needs, such as admin, owner or schema-wide grants.
- Which access has gone unused, where the source records activity.
- Where to start a review: the widest reach to the most sensitive data.
That is the starting point for least privilege, which NIST’s Cybersecurity Framework 2.0 builds into how access permissions are defined, managed and reviewed (PR.AA-05).
Pull the allowed reach from the database itself
On PostgreSQL, the information schema lists table privileges. The view shows privileges granted to or by a role that is currently enabled, so run it as the table owner or a superuser. Column-level grants sit in a separate view.
SELECT grantee, table_schema, table_name, privilege_type
FROM information_schema.table_privileges
WHERE table_schema = 'customer'
AND table_name = 'national_ids'
ORDER BY grantee, privilege_type;Expect a row for each grant. A grantee named PUBLIC means every role holds that privilege. This query reads grants and says nothing about use.
Activity is a different kind of evidence
A grant is permission given in advance. It doesn’t show that a read happened, what an export job copied out of the system, or who sat behind an application that reads through its own account. For a source that exposes no permissions, the entry is “permissions not available.”
NIST SP 800-53A lists what an audit record should establish: the type of event, when and where it occurred, its source, its outcome and the identity involved. That record exists only if the event type was selected for logging and the log was kept long enough. Reads of a sensitive table are not always among the events logged. Where they weren’t, the permission map is the best evidence available, and the report should say so.
Two columns keep the evidence apart
For any question about access to sensitive data, label one column “allowed reach” and the other “observed activity.” Each entry carries its source, the time it was taken and the person who reviewed it.
| Field | Allowed reach | Observed activity |
|---|---|---|
| Statement | reporting_ro in the customer database can select the national ID table | None available. Select statements on this table were not logged |
| Source | Database grants | Database audit settings |
| Taken on | November 9, 2026 | November 9, 2026 |
| Reviewed by | Database owner | Security lead |
The sentence to avoid merges the columns: “the reporting role accessed customer IDs.” It might be true. Without the log it is a guess written as a finding. Three wordings keep each claim to its record.
- Reach: “reporting_ro could select the national ID table from the date of the grant.”
- Activity: “Database logs show reporting_ro ran select statements on the table on these dates.”
- Gap: “No activity evidence exists for this table, because reads were not logged.”
Three records settle an investigation
An access question rarely stops at the database. Identity records show who sits behind an account and when they joined, moved or left. Application logs show which user triggered a query. Database audit records show which statements ran, when and with what result.
Each has a different owner and a different retention period. Write both down for each sensitive system now, because finding out during an incident is slower.
IRIS shows who can reach each sensitive dataset and flags excessive access, including broad rights on sensitive data, where the source exposes permission data.
IRIS reads the audit logs of the sources it supports and shows who accessed sensitive data and when, where the source keeps an audit log. Capabilities vary by edition.
The access mapping page lists the sources, and access review with IRIS shows how decisions are recorded.
Check what each source exposes before the review
For each source in scope, write down two facts: whether it exposes permissions, and whether it logs reads of the sensitive tables. The answers decide which column of your report you can fill.
The database access review guide walks through the queries, and the glossary entry on least privilege defines the principle behind them.
Sources
- NIST: SP 800-162, Guide to Attribute Based Access Control (ABAC) Definition and Considerations (January 2014, updated August 2, 2019; opened November 9, 2026)
- NIST: SP 800-53A Rev. 5, Assessing Security and Privacy Controls, controls AU-02 and AU-03 (January 2022; opened November 9, 2026)
- NIST: The NIST Cybersecurity Framework (CSF) 2.0, subcategory PR.AA-05 on access permissions (February 26, 2024; opened November 9, 2026)
- PostgreSQL Global Development Group: PostgreSQL 18 documentation, table_privileges (opened November 9, 2026)
Start with a free assessment of one database you choose.
Get a free assessmentProtect your brand. Start with your data.
Book a demo, or start with a free assessment of one database.


