CVE-2026-17183: 
Grafana vulnerability analysis and mitigation

CVE-2026-17183 is an Incorrect Authorization vulnerability in Grafana OSS and Grafana Enterprise that allows authenticated low-privileged users to bypass datasource query permission enforcement via manipulated alert rule queries. Discovered by researcher czarflix and reported via Intigriti (report GRAFANALABS-F996ATD7), it was published on August 19, 2026. Affected versions span Grafana OSS and Enterprise from 8.4.0 up to (but not including) 12.3.11, 12.4.0–12.4.9, 13.0.0–13.0.7, and 13.1.0–13.1.4. It carries a CVSS v3.1 base score of 7.1 (High) (GitHub Advisory, Grafana Advisory).

The root cause is an inconsistent query classification between Grafana's authorization and execution paths, classified as CWE-863 (Incorrect Authorization). When an alert rule query has its client-controlled queryType field set to "__expr__" (or "expr"), the datasource authorization helper skips datasources:query permission enforcement, treating the query as a server-side expression. However, the query evaluator subsequently resolves and executes the query against the real datasource identified by the datasourceUid parameter — which is also client-controlled and not validated against the spoofed queryType. The attack requires only network access, low privileges (alert rule create/edit permission in any accessible folder), and knowledge of the target datasource UID; no user interaction is needed (GitHub Advisory).

Successful exploitation exposes data accessible through Grafana's configured datasource credentials to users who lack datasources:query permission for that datasource, resulting in a High confidentiality impact. Additionally, depending on the datasource backend and the credentials configured in Grafana, the attacker may be able to execute state-changing (write) operations, resulting in a Low integrity impact. No availability impact has been demonstrated. The scope is unchanged, meaning the impact is contained within Grafana's own security authority, but sensitive monitoring data — potentially including infrastructure metrics, logs, or business data — could be exfiltrated (GitHub Advisory, Grafana Advisory).

As of the publication date, there is no public proof-of-concept exploit and no evidence of in-the-wild exploitation (GitHub Advisory). The EPSS score is 0.0, and the vulnerability is not listed in the CISA Known Exploited Vulnerabilities (KEV) catalog. NVD SSVC assessment classifies exploitation as "none" and the attack as not automatable, reflecting the requirement for authenticated access and knowledge of a target datasource UID (Feedly). The vulnerability was originally reported under a Cloud/Enterprise-specific context and subsequently reproduced under default OSS permissions by the same researcher.

1. Reconnaissance: Authenticate to the target Grafana instance as a low-privileged user with permission to create or edit alert rules in at least one folder. Enumerate available datasource UIDs via the Grafana API (GET /api/datasources) or by inspecting existing dashboards and alert rules — even if the user lacks datasources:query permission, datasource UIDs may be discoverable through the UI or API metadata.
2. Craft malicious alert rule: Create or edit an alert rule in an accessible folder. In the rule's query definition, set the queryType field to "__expr__" (the server-side expression sentinel value) while simultaneously setting datasourceUid to the UID of a real datasource the user is not authorized to query.
3. Submit the alert rule: Save or evaluate the alert rule via the Grafana API (e.g., POST /api/ruler/grafana/api/v1/rules/{folderUID} or the equivalent UI action). The authorization layer inspects queryType, classifies the query as a server-side expression, and skips datasource permission enforcement.
4. Trigger query execution: Trigger alert rule evaluation (e.g., by setting a short evaluation interval or using the "Test" functionality). The evaluator resolves the real datasource from datasourceUid and executes the query using Grafana's stored datasource credentials.
5. Exfiltrate data: Observe the alert rule evaluation results, which contain data returned from the unauthorized datasource. Depending on the datasource type and credentials, the attacker may also craft queries that perform state-changing operations (GitHub Advisory).

• Logs: Grafana server logs showing alert rule evaluation requests where queryType is "__expr__" or "expr" but datasourceUid references a real (non-expression) datasource UID; audit log entries for alert rule creation or modification by users with low privilege levels.
• Network: Unexpected datasource query traffic originating from the Grafana backend to datasource endpoints, correlated with alert rule evaluation events from low-privileged users.
• Application Behavior: Alert rules in folders owned or editable by low-privileged users that reference datasource UIDs those users are not authorized to query directly; alert rules with mismatched queryType and datasourceUid values (e.g., queryType: "__expr__" paired with a Prometheus, Loki, or other real datasource UID).
• API Activity: Repeated calls to datasource enumeration endpoints (GET /api/datasources) or alert rule APIs from accounts that do not normally interact with those endpoints (GitHub Advisory).

Grafana has released patched versions that properly enforce datasource query permissions regardless of the queryType value: 12.3.11, 12.4.9, 13.0.7, and 13.1.4 for both Grafana OSS and Grafana Enterprise. Upgrading to one of these versions is the recommended remediation (Grafana Advisory, GitHub Advisory). As interim mitigations: restrict alert rule creation and editing permissions to trusted users only; audit existing alert rules for queryType: "__expr__" entries paired with real datasource UIDs; and review datasource access logs for anomalous query patterns from low-privileged accounts.

The vulnerability was discovered by researcher czarflix via the Intigriti bug bounty platform (report GRAFANALABS-F996ATD7) and required a second submission to demonstrate the issue under default OSS permissions after an initial Cloud/Enterprise-specific report was archived as Not Applicable (GitHub Advisory). No significant broader media coverage or notable community commentary beyond the advisory has been identified at this time.