CVE-2026-11945: PostgreSQL Anonymizer Privilege Escalation via Malicious JSON

Technical summary

The vulnerability exists in how PostgreSQL Anonymizer processes JSON input during database rule and role rule imports. Attackers can embed arbitrary code within JSON key-value pairs that evade sanitization checks. When a superuser invokes import_database_rules() or import_roles_rules() functions on attacker-controlled JSON, the embedded code executes in the superuser's security context. The root cause is improper input validation and insufficient code isolation during JSON parsing and execution. CWE-89 (SQL Injection) classification suggests the injection vector may involve SQL statement manipulation embedded in JSON structures.

Business impact

Successful exploitation allows a standard database user to escalate privileges to superuser, enabling them to read, modify, or delete any data, create backdoor accounts, or sabotage database operations. For organizations using PostgreSQL Anonymizer in compliance-sensitive environments (healthcare, finance, PII handling), this undermines the extension's core function of protecting sensitive data. The attack requires superuser action, making it most dangerous in scenarios involving data imports from third parties, shared development environments, or scenarios where superusers routinely process rule files. Compromise could expose masked data or allow exfiltration of unmasked records.

Affected systems

PostgreSQL Anonymizer versions prior to 3.1.1 are vulnerable. The extension runs as part of PostgreSQL and affects all installations using Anonymizer for data masking. The vulnerability is platform-agnostic (Linux, Windows, macOS) and impacts any PostgreSQL instance where the extension is installed and superusers interact with user-supplied JSON rule files. Organizations should verify their current Anonymizer version against the vendor's release notes.

Exploitability

Exploitability is moderate. The attack requires three conditions: (1) the attacker must have authenticated database access, (2) the attacker must create a malicious JSON file and deliver it to a superuser, and (3) the superuser must actively call one of two specific import functions on that JSON. This is not a zero-click or remote unauthenticated attack. However, in environments where rule files are regularly imported (e.g., automated compliance policy updates, database migrations, or shared development workflows), the bar for exploitation lowers significantly. The requirement for direct superuser action prevents worm-like propagation but enables targeted privilege escalation against database administrators.

Remediation

Upgrade PostgreSQL Anonymizer to version 3.1.1 or later. Dalibo (the extension maintainer) has patched the JSON parsing logic to properly validate and sanitize input before code execution. Organizations should verify patch availability and test in non-production environments before deployment. Until patching is complete, restrict import_database_rules() and import_roles_rules() invocations to JSON files from trusted sources only, and audit historical imports for suspicious activity.

Patch guidance

Consult the PostgreSQL Anonymizer release notes and Dalibo's official advisory to confirm version 3.1.1 availability for your PostgreSQL version and platform. Verify patch compatibility with your current Anonymizer version and any dependent applications before applying. Test the update in a staging environment to ensure compliance rules and masking functions remain intact. Deploy during a maintenance window. Monitor PostgreSQL logs for any anomalies post-patching. If version 3.1.1 is not yet available for your PostgreSQL release, contact Dalibo support or consider temporarily disabling Anonymizer in high-risk environments.

Detection guidance

Monitor for suspicious invocations of import_database_rules() or import_roles_rules() in PostgreSQL audit logs, especially when called by superusers or database administrators. Inspect JSON files passed to these functions for embedded code patterns, unusual escape sequences, or payloads that deviate from expected rule structure. Establish a baseline of legitimate rule imports and alert on deviations. Review PostgreSQL error logs for SQL injection attempts or code execution errors during JSON import operations. If your organization stores historical rule files, audit them for malicious patterns. Enable PostgreSQL statement logging for these specific functions to capture the full JSON payloads in forensics scenarios.

Why prioritize this

This vulnerability rates as medium severity (CVSS 6.4) but warrants priority attention because it enables direct privilege escalation in database systems that handle sensitive data. The requirement for superuser action makes it a targeted attack vector rather than a mass-exploitation risk, but for organizations relying on Anonymizer for compliance and PII protection, privilege escalation is a worst-case outcome. The CVE does not yet appear on CISA's KEV list, suggesting active exploitation is limited, but this should not delay patching given the severity of superuser compromise. Prioritize patching based on whether your superusers regularly import external rule files and your organization's data sensitivity.

Risk score, explained

The CVSS 3.1 score of 6.4 (MEDIUM) reflects high confidentiality, integrity, and availability impact if exploited, but is tempered by high attack complexity (AC:H) and the requirement for high privilege level and user interaction (PR:H, UI:R). The network-accessible vector (AV:N) indicates remote authentication is sufficient, but the confluence of authentication, superuser involvement, and the need to trick a superuser into importing attacker JSON keeps the score in the medium range. For data-sensitive environments, the business impact may justify treating this as higher priority than the numeric score alone suggests.

Frequently asked questions