OpenSearch SQL CVE-2026-63393

Requires a valid OpenSearch account with basic read/search permissions and network reachability to the plugins/sql endpoint of the OpenSearch SQL plugin (which ships enabled by default in OpenSearch distributions); the attacker then supplies a crafted cursor parameter that reaches the unsafe Java deserialization path (pre-patch PlanSerializer / expression serializers lacking an ObjectInputFilter allowlist). Limiting factors: authentication is required (PR:L - not reachable by fully anonymous attackers), and the deployment must run an affected SQL plugin version (2.8.0.0 through <2.19.6.0, or 3.0.0 through <3.7.0.0). Successful code execution additionally depends on a usable gadget chain being present on the server classpath. No non-default configuration is otherwise needed.

This is a genuine high priority, not an inflated-CVSS issue. CWE-502 unsafe deserialization reaching arbitrary code execution is one of the most reliably weaponizable bug classes, and the vendor CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N) with VC:H/VI:H/VA:H confirms network-reachable, low-complexity exploitation requiring only low privileges and no user interaction - consistent with the description's 'remote authenticated user with basic read/search permissions.' The single meaningful barrier is authentication (PR:L): the attacker needs any account able to read/search, which in many OpenSearch deployments is a low bar (shared read-only credentials, broad org-wide access, or any compromised low-privilege user). No EPSS score and no KEV listing were provided, so there is no evidence of active exploitation; the vuln.today decision label is 'this_week' rather than act_now/emergency, indicating urgent-but-not-in-the-wild. No working exploit reference was supplied - the GitHub links are the advisory, fix PRs, patch commits, and release tags, not proof-of-concept code. Net: a serious, straightforwardly exploitable post-authentication RCE that should be patched promptly, with real-world risk gated primarily by how loosely read/search credentials are distributed in a given environment.

An attacker holding (or having phished/compromised) an OpenSearch account with only basic read/search permissions sends a request to the plugins/sql endpoint containing a maliciously crafted cursor parameter. The server deserializes the untrusted cursor without a safe allowlist, instantiating attacker-chosen classes that lead to arbitrary code execution on the OpenSearch node. No working public exploit was provided in the sources, but given the low attack complexity (AC:L) and no required user interaction, exploitation is practical once valid read credentials are obtained.