MySQL Server EUVD-2026-46569

Exploitation requires (1) the MySQL Server: Group Replication Plugin to be active - i.e. the instance is deployed as part of a Group Replication / InnoDB Cluster or MySQL Cluster topology, not a standalone default server - and (2) the attacker to already possess a local logon to the host or infrastructure where MySQL executes (AV:L), per Oracle's description. No MySQL account or database authentication is required (PR:N), no user interaction is needed (UI:N), and attack complexity is low (AC:L). The key limiting factors are the need for prior host access and a non-default clustered configuration; a hardened, single-tenant database host with no untrusted local accounts and without Group Replication enabled is not exploitable.

The CVSS vector (AV:L/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H, 8.4) is the strongest signal available and describes an easily exploitable, local-vector flaw requiring no privileges but yielding full compromise. Oracle's own phrasing ('unauthenticated attacker with logon to the infrastructure where MySQL executes') reconciles the AV:L / PR:N combination: no database authentication is needed, but the attacker must already be able to log on to the host or its infrastructure, which materially limits remote mass exploitation. No EPSS score, no KEV listing, and no POC were provided, so there is currently no evidence of active or automated exploitation; the practical risk is therefore high for multi-tenant or shared hosts running Group Replication but lower for well-isolated database servers. Note a data conflict to verify with the vendor: the 'Information Disclosure' tag understates the rated C:H/I:H/A:H 'takeover' impact - treat this as a full-compromise issue, not merely disclosure.

An attacker who has obtained an ordinary (non-database) shell on a server running MySQL with Group Replication - for example via a compromised co-located application account - interacts locally with the Group Replication Plugin to trigger the flaw and take over the MySQL instance, gaining full read/write control and the ability to disrupt the cluster. No database credentials and no user interaction are needed, and because attack complexity is low the action is reliable, though no public exploit code has been identified.