# Java4Shell (CVE-2026-114938): affected Java 17, 21, 25, 27 builds and fixes

> Oracle shipped fixed Java 17, 21, 25 and 27 builds on November 17, 2026 for a reported unauthenticated RCE rated 9.2 under CVSS 4.0. No workaround is published, and the trigger is not yet public.

- Published: 2026-11-17
- By: Samir Sengupta (https://www.samcodeman.com/about)
- Canonical: https://www.samcodeman.com/writing/java4shell-pre-auth-rce-java-17-to-27-patch-builds
- How this was made: Drafted with Claude, Anthropic's model from 1 cited source, then checked against it in a second editing pass. Editorial policy: https://www.samcodeman.com/editorial-policy

## The short version

- Oracle's fixed builds are 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1. Anything earlier on those lines is in scope.
- Amazon Corretto has fixed builds. The source lists no confirmed fix for Eclipse Temurin and no confirmed affected range for GraalVM.
- No workaround has been published, so patching is the only mitigation the source names.
- The vulnerable component and the exploitation conditions are not public, so teams cannot yet write targeted detection or filtering rules.

Java4Shell, tracked as CVE-2026-114938, is a reported unauthenticated remote code execution flaw rated 9.2 under CVSS 4.0. Oracle released patched builds on November 17, 2026: 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1. Treat every Oracle Java SE 17, 21, 25 or 27 runtime older than those builds as exposed. Java SE 8 and 11 are not affected by this CVE. The trigger, the vulnerable component and the server configurations involved are not public, and no workaround has been published, so upgrading the runtime is the only documented fix.

The main source is an Aikido post by Sooraj Shah, published on November 17, 2026, the same day as Oracle's release. It states that the vulnerable component and the conditions for exploitation are not yet public. That limits what engineers can do. You cannot scope exposure by feature, endpoint or configuration, so the practical unit of exposure is the runtime version.

## Java4Shell affected and fixed builds by vendor

Aikido's table lists the Java builds Oracle marks as affected, with the patched builds from Oracle's November 17, 2026 release notes, plus rows for other vendors. Oracle's own Java SE lines and Amazon Corretto have fixed builds. The Eclipse Temurin and GraalVM rows are incomplete.

- Oracle Java SE 17: releases before 17.0.21.1 are affected; fixed in 17.0.21.1.
- Oracle Java SE 21: releases before 21.0.13.1 are affected; fixed in 21.0.13.1.
- Oracle Java SE 25: releases before 25.0.5.1 are affected; fixed in 25.0.5.1.
- Oracle Java SE 27: releases before 27.0.1.1 are affected; fixed in 27.0.1.1.
- Amazon Corretto: affected through 17.0.21.0.11, 21.0.13.0.11, 25.0.5.0.10 and 27.0.1.0.11, including earlier releases in each branch; fixed in 17.0.21.1.2, 21.0.13.1.2, 25.0.5.1.2 and 27.0.1.1.4.
- Eclipse Temurin: affected through 17.0.21+9, 21.0.13 and 25.0.5+8, including earlier releases in each branch; no confirmed fix.
- Oracle GraalVM for JDK and GraalVM Enterprise Edition: CVE-specific affected range not confirmed, fix not confirmed.
- Oracle Java SE 8 and 11: not affected by this CVE.
- Oracle JavaFX: not affected by this CVE; listed separately under CVE-2026-86144.

## Scope and severity reports for Java4Shell disagree

The reports do not agree on scope. Early reports tie the CVE to the Java 17 line only, covering all versions before 17.0.21.1. Oracle's affected list covers its whole November update, which includes Java 21, 25 and 27. Aikido's advice is to treat every Java line you run as in scope, and that is the safer reading while the two accounts differ.

The severity figures also differ. The reported 9.2 is a CVSS 4.0 score. Oracle scores Java flaws under CVSS 3.1, and its highest Java SE score this month is 8.1. Aikido says the two numbers may describe different issues. Do not use either score to lower the priority of a JVM fleet while it is unclear which issue each figure describes.

## The trigger behind the pre-auth RCE is not public

The source does not explain how the flaw is triggered. Aikido says further technical details are expected after Oracle's patch release, and that the vulnerable component and exploitation conditions are not public. The only attack property reported is that no authentication is required. The source names no API, protocol, deserialization path or server configuration. A claim that a specific framework setting or JVM flag closes the hole goes beyond the published material.

> None has been published that we are aware of. Patching is the fix.
>
> - Aikido, answering whether a workaround exists for CVE-2026-114938

There is no configuration-level mitigation to apply today. Network controls that reduce reachability are general hygiene, and the source does not describe them as a fix for this flaw. Upgrading to a fixed build is the only documented remediation.

## OpenJDK vendor builds need the vendor's fixed version

Aikido tells teams on Temurin, Corretto, Zulu or a distribution package to wait for that vendor's fixed version, which may carry a different build number from Oracle's. Corretto's fixed builds already show this. Its Java 27 fix is 27.0.1.1.4, not Oracle's 27.0.1.1. Matching Oracle's numbers against a non-Oracle runtime can therefore give the wrong answer.

According to Aikido, you are in scope if you run Java 17, 21, 25 or 27 before the fixed version, and that includes bundled JDKs and container base images. Aikido lists these Java builds to check:

- Oracle Java SE, both JDK and JRE.
- Oracle GraalVM for JDK and GraalVM Enterprise Edition.
- Oracle JavaFX.
- OpenJDK builds from other vendors, such as Eclipse Temurin, Amazon Corretto, Azul Zulu, the Microsoft Build of OpenJDK and the Red Hat build of OpenJDK.
- Linux distribution packages such as openjdk-17.
- JDKs bundled inside applications and container base images.

The source gives no affected ranges or fixed versions for Azul Zulu, the Microsoft Build of OpenJDK, the Red Hat build of OpenJDK or Linux distribution packages. If you run any of those, check the vendor's advisory directly. Do not assume Oracle's build numbers cover you.

## Patch internet-facing JVM services, including inference backends, first

Aikido recommends starting with internet-facing services, because anything reachable from outside is the most likely target once details are public. For AI teams, that includes any JVM service that is reachable from outside on the path to model serving. The source has no AI-specific exposure data. These services are in scope because they run an affected Java line, not because of anything particular to inference workloads.

1. Inventory every runtime, including JDKs bundled in applications and in container base images. A service can be patched at the host and still ship an old JDK inside its image.
2. Upgrade internet-facing services on Java 17, 21, 25 or 27 first. Use 17.0.21.1, 21.0.13.1, 25.0.5.1 or 27.0.1.1 on Oracle, or the vendor equivalent such as Corretto 17.0.21.1.2.
3. Rebuild and redeploy images rather than patching running containers, so the fixed JDK persists through the next rollout.
4. Track Temurin, GraalVM, Zulu, the Microsoft and Red Hat builds, and distribution packages, none of which has a confirmed fix in the source. Schedule them as soon as their vendors publish one.
5. Apply Oracle's November update to Java SE 8 and 11 too. Those lines are not affected by this CVE, but the same update fixes other Java flaws in them.

## No exploitation reported as of November 17, 2026

As of November 17, Aikido had seen no reports of exploitation in the wild. It also said technical details were due later that day, and that exploitation can follow quickly once details are public. The source does not say whether those details have been published. Use the time between the patch release and a public trigger to finish upgrades, starting with exposed services. The source does not predict how long that window will be.

## Questions this raises

### Which Java versions are affected by Java4Shell?

Oracle Java SE 17, 21, 25 and 27 releases before 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1 are affected. Java SE 8 and 11 are not affected by this CVE. Early reports tied the flaw to Java 17 only, but Aikido advises treating every Java line you run as in scope.

### Is there a workaround for Java4Shell?

No workaround has been published, and the trigger, the vulnerable component and the exploitation conditions are not public. Upgrading the runtime to a fixed build is the only documented remediation.

### Do Oracle's fixed build numbers apply to Corretto, Temurin or other OpenJDK builds?

No. Each vendor ships its own fixed version, which can carry a different build number. Corretto's Java 27 fix, for example, is 27.0.1.1.4 rather than Oracle's 27.0.1.1. Temurin, Zulu, Microsoft, Red Hat and distribution packages have no confirmed fix in the source, so check each vendor's advisory.

## Key facts

- CVE-2026-114938 is a reported unauthenticated RCE rated 9.2 under CVSS 4.0; Oracle's highest Java SE CVSS 3.1 score this month is 8.1.
- Oracle fixed builds released November 17, 2026: 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1.
- Amazon Corretto fixes: 17.0.21.1.2, 21.0.13.1.2, 25.0.5.1.2 and 27.0.1.1.4; Eclipse Temurin and GraalVM have no confirmed fix.
- Oracle Java SE 8 and 11 are not affected by this CVE; JavaFX is listed separately under CVE-2026-86144.
- As of November 17, 2026, Aikido had seen no reports of exploitation in the wild.

## Sources

1. [Java4Shell: pre-auth RCE flaw in Java 17 puts servers at risk (CVE-2026-114938)](https://www.aikido.dev/blog/java-17-rce-vulnerability-cve-2026-114938) - aikido.dev
