AWS patches Gremlin import flaw in Amazon Neptune graph database
EditorialBy TrustList Editorial
Bulletin 2026-136-AWS covers every engine from 1.2.0.0. Clusters with automatic minor upgrades get the fix in their next maintenance window; the rest must move to 1.4.7.1, 1.4.8.1 or 1.5.0.0.
About AWS patches Gremlin import flaw in Amazon Neptune graph database
AWS patches Gremlin import flaw in Amazon Neptune graph database
15 October 2026: Amazon Web Services has fixed CVE-2026-110812, a flaw in the Gremlin graph-import path of Amazon Neptune that lets a user with query permissions make the engine process untrusted input unsafely. AWS classes its bulletin 2026-136-AWS as important, says every Neptune engine version from 1.2.0.0 is affected, and has shipped fixed engines 1.4.7.1, 1.4.8.1 and 1.5.0.0.
Not yet independently verified. Single source: the AWS bulletin, which states no CVSS score and says nothing about exploitation. We found no Apache TinkerPop advisory for the same CVE yet and will add one when it appears. We will update this when it can be confirmed, and remove this note.
What the flaw allows
Neptune is AWS's managed graph database, queried with Gremlin, openCypher or SPARQL. The flaw is in Apache TinkerPop, the open-source project that defines Gremlin and whose components Neptune uses. AWS describes it as deserialization of untrusted data: under certain conditions an authenticated principal with Neptune data-plane permissions can submit a request that leads to "unintended activity within the query engine environment".
That wording stops short of saying what an attacker gains, and the bulletin carries no severity score. What it does make clear is the precondition. Nobody can reach the flaw anonymously; the request has to come from an identity that already holds neptune-db query permissions through IAM, or from an application running under one.
How the fix reaches each cluster
The route depends on how a cluster is configured:
- Clusters with automatic minor version upgrade turned on receive the fix during their next maintenance window, with no action beyond letting that window run.
- Clusters whose engine version is managed by hand need an upgrade to the patched release in their line: 1.4.7.1, 1.4.8.1 or 1.5.0.0.
Neptune engine upgrades restart the cluster, so teams that pin versions usually schedule them, and the maintenance window setting in the console shows when an automatic upgrade will land.
Who should look first
Until a cluster is upgraded, AWS's advice is least privilege: grant neptune-db query permissions only to trusted principals. In practice that means reviewing which roles, Lambda functions and applications hold those permissions, and in particular any service that passes user-supplied Gremlin or import requests through to the database. A graph that backs a recommendation engine or a fraud model queried only by internal services carries less exposure than one behind a multi-tenant API.
The bulletin was published at 09:30 Pacific time on 15 October and asks for questions to go to aws-security@amazon.com.
Our NoSQL databases software category lists 27 products for teams weighing managed graph and document stores.
Sources
Categories & features
- United States
- Washington
- Seattle
- Cybersecurity
- Cloud Security
- Vulnerability Management
- AWS
- Graph Database
- DataBase
TrustList Weekly
The week in software and IT, in one email
The news that matters to buyers, new rankings and our own research. Every Thursday, free, and easy to leave.
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- More United StatesThe ranking for this subject
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- AI ModelsBenchmark scores and community ratings for every major model.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.