Amazon Neptune EUVD-2026-102193

Exploitation requires valid authenticated access to the Amazon Neptune engine with standard (non-privileged) database permissions, and the ability to reach the Gremlin graph data import component to submit a malicious serialized payload. Limiting factors: the attacker must already possess working database credentials (PR:L - this is not an unauthenticated flaw), so network reachability alone is insufficient; realistic risk is highest in multi-user, multi-tenant, or lower-trust deployments where untrusted parties hold standard DB accounts. Managed clusters already on the auto-update track may be remediated without action; self-managed-version operators must manually upgrade to 1.5.0.0 or the 1.4.8.x/1.4.7.x backports.

This is a genuine high-priority issue, not a CVSS-inflated one. The vendor-assigned CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N) scoring 9.4 is internally consistent with the description: a network-reachable, low-complexity deserialization flaw (CWE-502) requiring only standard database permissions (PR:L) and no user interaction, yielding full compromise (VC/VI/VA:H) plus subsequent-system impact (SC/SI/SA:H) - the latter matching the stated theft of internal service credentials, which is a meaningful escalation signal. AWS/AMZN self-reported and has shipped patches across three release lines, indicating vendor confirmation and a mature remediation path. Countervailing the raw severity: there is NO CISA KEV listing, NO EPSS score provided, and NO public exploit code referenced, so there is no evidence of active or imminent widespread exploitation. The exploitation barrier is also non-trivial - the attacker must already hold valid, standard-permission credentials to a Neptune cluster, making this primarily an insider/post-compromise privilege-and-lateral-movement risk rather than an internet-wide drive-by. Net: prioritize patching for any Neptune deployment with multi-tenant or lower-trust database users, but recognize exploitation is credential-gated and not yet observed in the wild per available data.

An actor holding standard, authenticated Neptune database credentials (e.g., a low-privilege application user or a malicious tenant) submits a crafted payload through the Gremlin graph data import component. The unsafe deserialization of that payload executes attacker-controlled code within the engine and exposes internal AWS service credentials, enabling lateral movement. No public proof-of-concept is referenced, and no user interaction is required.