The vulnerability stems from a failure to neutralize quoting syntax in the Apache Iceberg write component of the AWS SDK for pandas. This flaw allows an authenticated remote user who has the right to alter AWS Glue Data Catalog metadata to supply a specially crafted column name. The crafted name can bypass normal quoting rules, giving the user the ability to read any table that a more privileged principal can access and to write to that same table as the principal is updating. The result is unauthorized data read and write, effectively elevating the user’s privileges on targeted tables.

Amazon Web Services’ SDK for pandas (also known as awswrangler) versions 3.5.2 through 3.17.0 are affected. The product is branded as aws-sdk-pandas by AWS and is distributed under an open‑source license.

The CVSS score of 5.0 places this issue in the medium severity range. No EPSS value is currently available, and the vulnerability is not listed in the CISA KEV catalog, indicating no widespread exploitation has been documented. However, the attack requires an authenticated user with permission to modify Glue Data Catalog metadata—typically an account with administrative or data‑engineering roles. Because the flaw leverages a subtle quoting issue rather than a blatant injection, exploitation may be non‑trivial but is achievable for skilled attackers with the necessary permissions. The overall risk is therefore moderate: it is not widely exploitable from the public Internet, but insiders or compromised privileged accounts could read or tamper with data they should not access.