← All updates

Security / News explained

AWS Go Wrapper fix closes a database privilege shortcut

A November 12 advisory explains how an affected Go database wrapper could let one database user gain another user’s privileges. The fix dates to July, and schema permissions still deserve attention.

By Zyel · Published · 4 min read

AWS disclosed CVE-2026-126793 on November 12, 2026: a privilege-escalation issue in the Aurora PostgreSQL detection logic of AWS Advanced Go Wrapper. Under the affected conditions, connecting to a database could give an existing database user an opportunity to act with another user’s authority. The relevant boundary is inside the database, where shared schemas and differently privileged users meet. [2]

The affected awssql/v2 module versions are 2.0.0 through versions before 2.0.3. AWS identifies the affected release window as April 6 through before July 29, 2026. The July 29 release contains the fix. This is a new public disclosure about an older available patch; it does not mean the correction first arrived in November. [1] [3]

The conditions matter

The advisory describes an authenticated database user who can create functions or other objects in a schema searched by a second user. That second user must connect through an affected wrapper. The first user may then obtain the second user’s privileges. Those conditions are more specific than simply having an Aurora cluster or running a Go application. [2]

The potential victim can include a member of rds_superuser. Aurora documents that role as carrying substantial database administration powers, including managing roles and databases. It is distinct from unrestricted PostgreSQL superuser or host operating-system access. Keeping that distinction clear makes the consequence serious without turning it into an unsupported claim about control of the entire AWS account. [5]

A search path carries trust

PostgreSQL schemas organize named objects, including functions. A search path determines where the database looks for names that are not explicitly qualified. PostgreSQL’s documentation warns that putting a schema on that path effectively trusts people allowed to create objects there. The convenience of a short name therefore comes with an authority decision, even when the application never calls it one. [4]

Consider a fictional reporting system where an analyst can create helper functions in a shared schema and a more privileged service account searches that schema. The analyst’s ability to add objects and the service account’s authority are separate permissions. The dangerous connection is allowing the former to influence what the latter resolves and executes. This illustrates the trust relationship; it is not a reported customer incident.

Our reading is that connection setup belongs in the security model alongside application queries. Teams often scrutinize the SQL they wrote while treating a driver’s startup queries as plumbing. A library can still execute those queries using a powerful identity. Plumbing, unfortunately, does not become harmless because it lives behind a useful interface.

Check the module that actually ships

The July release notes identify a change to fully qualify PostgreSQL queries. They also list separate versions for the wrapper’s modules. The security advisory gives the corresponding affected and patched ranges, so a single repository-level date should not substitute for checking the modules an application actually includes. [2] [3]

For example, the patched awssql/v2 version is 2.0.3, while pgx-driver is patched at 1.1.3 and aws-secrets-manager at 1.1.4. Other modules have their own entries in the advisory. These are package-specific minimum fixes, not interchangeable version numbers or a statement that every module provides the same database behavior. [2]

For an application owner, the useful evidence is the dependency revision in the deployed application, including copied or forked code. A corrected development branch is encouraging, but the operational question is whether the running service uses that correction. AWS explicitly includes forks and derivatives in its patch recommendation. [1]

Permissions are a separate part of the response

AWS’s interim advice is to remove broadly available creation rights from the public schema and allow creation in other schemas on these users’ search paths only for trusted database users. That is a workaround for teams unable to upgrade immediately, rather than a reason to leave the affected wrapper indefinitely in place. [1]

Database history deserves attention here. PostgreSQL documents different public-schema defaults for newer databases and databases upgraded from PostgreSQL 14 or earlier. A recent engine version alone is therefore poor evidence that inherited grants are safe. The actual permissions and search paths tell the more useful story. [4]

Our interpretation is that a permissions review should account for legitimate application ownership and migration needs. Removing a grant without understanding who relies on it can interrupt work; retaining it without understanding who can create objects preserves a trust relationship. The decision needs an owner and a reason, rather than a comforting assumption about defaults.

Close the finding with evidence

The cited advisories establish a possible escalation path and a fix. They do not establish that a particular organization was attacked. Where the affected combination existed, a proportionate review would consider unexpected functions, object ownership and role changes using the evidence available for that environment. Missing history should remain an explicit uncertainty.

For release review, we would keep two conclusions separate: the application now contains corrected wrapper modules, and the database’s shared-schema permissions match its intended trust boundaries. Either can be checked without pretending the other follows automatically. That makes the final record useful to the next person who inherits the service.

The broader engineering lesson is modest but durable: names resolved on another user’s behalf are part of authorization. The package fix removes this particular shortcut. Deliberate schema ownership and appropriately limited service identities help keep the surrounding database design understandable when the next library advisory arrives.

Sources + further reading

Source dates describe the original material. Publication above is the date this Zyel article went live.

  1. CVE-2026-126793: AWS Advanced Go Wrapper privilege escalationAWS Security · Checked Nov 13, 2026
  2. GHSA-fhr2-6r5v-gm9f: affected Go modules and search-path impactAWS Advanced Go Wrapper on GitHub · Checked Nov 13, 2026
  3. AWS Advanced Go Wrapper release, July 29, 2026AWS Advanced Go Wrapper on GitHub · Checked Nov 13, 2026
  4. PostgreSQL 18 documentation: schemas, search paths and privilegesPostgreSQL Global Development Group · Checked Nov 13, 2026
  5. Understanding the rds_superuser roleAmazon Aurora documentation · Checked Nov 13, 2026

Reader discussion

What would you ask next?

Questions, useful corrections and thoughtful disagreement are welcome. Comments are reviewed before they appear.

No reader comments yet.

10–2,000 characters. Plain text. Leave out passwords, private project data and personal contact details. A blank display name appears as Anonymous.

Complete the CAPTCHA before sending. Verification runs on your device.