PostgreSQL Anonymizer contains a vulnerability that allows a user to gain superuser privileges by creating a custom operator in the public schema and place malicious code in that operator. This operator will later be executed with superuser privileges when the extension is created. The risk is higher with PostgreSQL 14 or with instances upgraded from PostgreSQL 14 or a prior version. With PostgreSQL 15 and later, the creation permission on the public schema is revoked by default and this exploit can only be achieved if a superuser adds a new schema in her/his own search_path and grants the CREATE privilege on that schema to untrusted users, both actions being clearly discouraged by the PostgreSQL documentation. The problem is resolved in PostgreSQL Anonymizer 3.0.1 and further versions

PostgreSQL Anonymizer v1.2 contains a SQL injection vulnerability that allows a user who owns a table to elevate to superuser when dynamic masking is enabled. PostgreSQL Anonymizer enables users to set security labels on tables to mask specified columns. There is a flaw that allows complex expressions to be provided as a value. This expression is then later used as it to create the masked views leading to SQL Injection. If dynamic masking is enabled, this will lead to privilege escalation to superuser after the label is created. Users that don't own a table, especially masked users cannot exploit this vulnerability. The problem is resolved in v1.3.

PostgreSQL Anonymizer v1.2 contains a vulnerability that allows a user who owns a table to elevate to superuser. A user can define a masking function for a column and place malicious code in that function. When a privileged user applies the masking rules using the static masking or the anonymous dump method, the malicious code is executed and can grant escalated privileges to the malicious user. PostgreSQL Anonymizer v1.2 does provide a protection against this risk with the restrict_to_trusted_schemas option, but that protection is incomplete. Users that don't own a table, especially masked users cannot exploit this vulnerability. The problem is resolved in v1.3.