CVE-2026-82526

R2R through version 3.6.6 contains a stacked SQL injection flaw (CWE-89) in its vector index creation endpoint. The endpoint takes an index name parameter and interpolates it directly into a PostgreSQL CREATE INDEX statement via string formatting, without identifier quoting or allowlist validation, so an unauthenticated attacker can inject semicolon-separated statements that execute as additional SQL. Successful exploitation grants arbitrary DDL and DML execution under the PostgreSQL superuser account, allowing the attacker to read, modify, or delete data and create or drop database objects (CVSS 4.0 score 9.3, critical, network-exploitable with no privileges or user interaction required and high impact to confidentiality, integrity, and availability). Any deployment running an affected version with the index creation endpoint reachable by untrusted network clients is exposed. As of now there are no known exploits in the wild, no public proof-of-concept, and no CISA KEV listing, with EPSS estimating only a 0.4% probability of exploitation in the next 30 days.

What to do: Upgrade R2R to the latest release above 3.6.6 once a patched version is published (the source data does not specify a fixed version). Until then, restrict network access to the vector index creation endpoint, run the application's PostgreSQL connection under a least-privilege role instead of superuser, and consider WAF rules constraining the index name parameter. Check database logs for unexpected CREATE INDEX statements or semicolon-separated queries that would indicate exploitation attempts.