AWS Security ChangesHomeSearch

AWS clean-rooms: Document analysis-rule controls enforced only on SQL queries, not PySpark

Service: clean-rooms · 2026-11-06 · Documentation medium

File: clean-rooms/latest/userguide/custom-considerations.md · Type: authz

Summary

Added a new "Controls that apply only to SQL queries" section explaining that differentialPrivacy, disallowedOutputColumns, additionalAnalyses, aggregationThresholds, and comparisonControls are not enforced on PySpark jobs, that rule creation fails when only jobs are allowed, that StartProtectedJob rejects jobs referencing such tables, and how to separate tables for SQL vs PySpark workloads.

Security assessment

The added text documents the enforcement boundary of privacy/authorization controls (differential privacy, aggregation thresholds, comparison controls, disallowed output columns, additional analyses) and explicitly notes they are not enforced for PySpark jobs, with StartProtectedJob rejecting jobs referencing such tables. This is security-relevant guidance about control enforcement and job rejection behavior, but no specific vulnerability, CVE, or incident is referenced.

Evidence

+The following controls apply only to SQL queries. AWS Clean Rooms doesn't enforce them on PySpark jobs, because a job runs custom code that AWS Clean Rooms can't check against these controls.

Diff

diff --git a/clean-rooms/latest/userguide/custom-considerations.md b/clean-rooms/latest/userguide/custom-considerations.md
index 3ad8ad5..72c1bf5 100644
--- a//clean-rooms/latest/userguide/custom-considerations.md
+++ b//clean-rooms/latest/userguide/custom-considerations.md
@@ -7 +7 @@
-Assumptions about potentially adversarial queriesLimitationsDifferencing attacks
+Assumptions about potentially adversarial queriesLimitationsControls that apply only to SQL queriesDifferencing attacks
@@ -57,0 +58,28 @@ Note the following limitations:
+## Controls that apply only to SQL queries
+
+The following controls apply only to SQL queries. AWS Clean Rooms doesn't enforce them on PySpark jobs, because a job runs custom code that AWS Clean Rooms can't check against these controls.
+
+  * `differentialPrivacy`
+
+  * `disallowedOutputColumns`
+
+  * `additionalAnalyses`
+
+  * `aggregationThresholds`
+
+  * `comparisonControls`
+
+
+
+
+How AWS Clean Rooms handles these controls depends on what the rule allows:
+
+  * If the rule allows only jobs (`ANY_JOB` without `ANY_QUERY`, or only PySpark analysis templates), you can't set these controls. Creating or updating the analysis rule fails with a validation error.
+
+  * If the rule allows both SQL queries and jobs, AWS Clean Rooms enforces the controls on SQL queries. `StartProtectedJob` rejects any job that references the table.
+
+
+
+
+To use these controls on SQL queries and also allow PySpark jobs on the same data, create a separate configured table for jobs, with a custom analysis rule that doesn't set these controls.
+