Safeguard
Vulnerability Analysis•

CVE-2026-34361: HAPI FHIR Validator Unauthenticated loadIG Flaw

CVE-2026-34361 is a critical flaw in the HAPI FHIR Validator HTTP service where an open /loadIG endpoint can expose credentials. Upgrade to 6.9.4 or later.

Safeguard Security Team
2 min read

CVE-2026-34361 is a critical (CVSS 9.3) flaw in the HAPI FHIR Validator HTTP service. An unauthenticated /loadIG endpoint makes outbound requests to caller-chosen URLs, and a URL matching bug can then leak configured credentials. The fix is version 6.9.4. Restrict who can reach the service.

What's affected

The project's advisory (GHSA-vr79-8m62-wh98) lists the Maven package ca.uhn.hapi.fhir:org.hl7.fhir.validation below 6.9.4 as affected, and 6.9.4 as patched.

Two problems combine. The /loadIG endpoint accepts unauthenticated POST requests with a JSON field naming an implementation guide, and passes it to the loader with no URL validation or allowlist. Separately, the credential lookup in ManagedWebAccessUtils.getServer() uses a startsWith() prefix match, described in CVE-2026-34359.

Together, an outside caller can steer the service to fetch from a domain that prefix-matches a configured server and collect the tokens it sends.

How to check if you're exposed

Check whether you run the validator as an HTTP service, and which version:

mvn dependency:tree | grep hl7.fhir.validation

For a packaged validator, look at the version string it reports. Then check who can reach its port. A service open only to a build pipeline is far safer than one on a public address.

What to do now

  1. Find any deployment of the FHIR Validator HTTP service.
  2. Upgrade to 6.9.4 or later.
  3. Put the service behind authentication, or limit it to trusted networks.
  4. Rotate tokens and keys configured for FHIR servers if it was reachable from outside.
  5. Limit outbound traffic from the service to the hosts it needs.

Why it matters

The record lists CWE-552, files or directories accessible to external parties. The deeper pattern is a helper service with no authentication that fetches whatever it is told to, which is server-side request forgery in effect, plus a credential store that decides where to send secrets by string prefix.

Each part would be a smaller issue alone. Combined, they let an outsider turn a configured credential into a stolen one. The record does not mark this CVE as exploited, and it is not on the CISA KEV list.

Quick check

Is this the same as CVE-2026-34359?

They are related. That CVE is the URL matching bug. This one is the open endpoint that makes it reachable from outside.

Sources

Last reviewed: 19 October 2026.

Never miss an update

Weekly insights on software supply chain security, delivered to your inbox.

Self-healing security runs on Safeguard.

Your first fix PR is minutes away.

No sales call required, even your agent can complete the purchase over MCP.