Critical by impact (our framing)OpenJDK / Oracle CVSS 3.1 8.1 HighResearchers: CVSS 4.0 9.2 CriticalBase64 decoder, AVX-512 VBMI

Have you checked the Java runtime behind your Jira, Confluence, Jenkins and Keycloak?

CVE-2026-114938 is a vulnerability in the Compiler component of OpenJDK and Oracle Java SE. The OpenJDK Vulnerability Advisory of 17 November 2026 and Oracle's November 2026 Critical Security Patch Update both score it CVSS 3.1 8.1 High, with high attack complexity. Successful attacks, the official CVE record says, can result in takeover of the Java runtime. BleepingComputer reports that security researchers call it Java4Shell, describe it as memory corruption in Java's hardware-accelerated Base64 decoder, and say it can lead to unauthenticated remote code execution in default configurations of widely used applications. Upgrade to the 17 November 2026 OpenJDK and Oracle Java SE releases (17.0.21.1, 21.0.13.1, 25.0.5.1 or 27.0.1.1 in the lines this CVE affects).

Critical by impact (our framing). Vendor score: CVSS 3.1 8.1 High

OpenJDK and Oracle score this CVSS 3.1 8.1 High (vector CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H). BleepingComputer reports that security researchers rate it CVSS 4.0 9.2 Critical. We use Critical by impact as our framing because the official record describes takeover of the Java runtime, and because that report describes unauthenticated code execution in default installs of widely deployed applications. The CVE was not in CISA's Known Exploited Vulnerabilities catalogue when we checked at 18 November 2026, 18:09 GMT (catalogue 2026.11.17, released 17:53 GMT on 17 November 2026). CISA's ADP entry on the CVE record, last updated at 22:02 GMT on 17 November 2026, lists Exploitation as "none". That entry predates some of today's public reporting, and we do not treat absence from KEV, or that ADP label, as proof that a flaw is unused. Recheck KEV. We do not link proof-of-concept or exploit code.

At a Glance

Based on the OpenJDK Vulnerability Advisory of 17 November 2026, Oracle's November 2026 Critical Security Patch Update, the official CVE-2026-114938 record, the NVD copy of that record, BleepingComputer (17 November 2026), and CISA's Known Exploited Vulnerabilities catalogue, checked at 18 November 2026, 18:09 GMT. Last updated 18 November 2026, 18:09 GMT.

ItemDetail
CVECVE-2026-114938 (CNA: Oracle). Press and researchers also use the name Java4Shell.
Vendor scoreCVSS 3.1 8.1 High (OpenJDK and Oracle). Vector: CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H.
Researcher scoreCVSS 4.0 9.2 Critical, as reported by BleepingComputer. This is not an OpenJDK or Oracle score.
OpenJDK linesThe 17 November 2026 risk matrix marks 17, 21, 25 and 27 only. It does not mark 8 or 11 for this CVE, even though the advisory introduction lists 8u512 and 11.0.33 for the release as a whole.
Affected (OpenJDK, this CVE)The advisory says: "The affected versions are 27.0.1.0.3, 25.0.5.0.3, 21.0.13.0.3, 17.0.21.0.3, 11.0.33, 8u512, and earlier." That sentence covers the whole 17 November 2026 advisory. The risk matrix for CVE-2026-114938 marks only the 17, 21, 25 and 27 lines, not 8 or 11.
Affected (Oracle, this CVE)The Java SE risk matrix column is headed "Supported Versions Affected" and lists 8u511b33-perf, 17.0.21.0.3, 21.0.13.0.3, 25.0.5.0.3 and 27.0.1.0.3, plus GraalVM for JDK 17.0.21.0.3 and 21.0.13.0.3 and GraalVM Enterprise Edition 21.3.20.0.3. Oracle's 17.0.21.0.3 release notes (24 October 2026) said not to use that build after the next critical patch update on 17 November 2026. The Java 8 entry is the 8u511b33-perf performance build only, not general Java 8, and Java 11 is not listed for this CVE.
Fixed builds (17 November 2026)Confirmed by the Oracle Java SE release notes dated 17 November 2026, the OpenJDK JDK 27.0.1.1 release notes (release date 2026-11-17) and Oracle's current JRE security baseline: 17.0.21.1 (17.0.21.1+4), 21.0.13.1 (21.0.13.1+4), 25.0.5.1 (25.0.5.1+4) and 27.0.1.1 (27.0.1.1+4). For the listed Java 8 performance build, the matching 17 November 2026 release is 8u513-perf (1.8.0_513-perf-b4). Oracle GraalVM for JDK 21.0.13.1 is a confirmed 17 November 2026 CSPU release: its own release notes say it is built on Oracle JDK 21.0.13.1+4. Oracle GraalVM for JDK 17.0.21.1 and GraalVM Enterprise Edition 21.3.20.1 are under review, because their vendor notes conflict: both are labelled as the 17 November 2026 CSPU, but the notes say they are built on Oracle JDK 17.0.20.1+4, which is below the 17.0.21.1 fix. The Java 17-based Enterprise Edition build is the one given as 17.0.20.1+4; the same notes give 11.0.33.1+4 and 8u513-b04 for the Java 11 and Java 8 based builds. OpenJDK 8 and 11 are not marked for this CVE; 8u513 and 11.0.33.1 are the current security baseline for those families, but we do not list them as the fix for CVE-2026-114938.
Start of the 17 line (press)BleepingComputer reports that the flaw affects OpenJDK 17.0.6 and later, and distributions built from that code, including Oracle Java SE, Amazon Corretto, Eclipse Temurin, Microsoft Build of OpenJDK, Azul Zulu and Red Hat OpenJDK. OpenJDK and Oracle do not give a 17.0.6 start version.
NVDReceived. Published 22:18 GMT on 17 November 2026. The NVD copy repeats Oracle's CVSS 3.1 8.1 score. Status when we checked: NVD status when we checked: Received.

What to Do in the Next 24 Hours

Vendor steps first, then our advice for internet-facing Java applications.

  1. Inventory internet-facing Java first

    Our advice: find the Java runtime behind internet-facing Jira, Confluence, Jenkins, Keycloak, JFrog Artifactory and PingFederate, then other Java services. Record the distribution, the full version string, and whether the host CPU has AVX-512 VBMI (grep avx512vbmi /proc/cpuinfo on Linux).

  2. Upgrade the runtime

    Vendor advice (OpenJDK and Oracle): upgrade as soon as possible to the 17 November 2026 OpenJDK release and the Oracle Java SE / GraalVM patches in the November 2026 CSPU. For the lines this CVE affects, the confirmed fixed builds are 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1. For the Oracle Java SE 8u511b33-perf build listed as affected, use 8u513-perf. For GraalVM, Oracle GraalVM for JDK 21.0.13.1 is confirmed fixed. Treat Oracle GraalVM for JDK 17.0.21.1 and GraalVM Enterprise Edition 21.3.20.1 as under review, because their vendor notes conflict (they say they are built on Oracle JDK 17.0.20.1+4). Our advice: also apply bundled-JRE updates from the application vendors as they ship them, and do not assume a product installer is already on this runtime.

  3. If you cannot patch immediately

    BleepingComputer reports that security researchers recommend a temporary workaround: start the JVM with -XX:-UseBASE64Intrinsics, or with -XX:UseAVX=2 to turn off AVX-512. We did not find those flags in the OpenJDK or Oracle advisories. Oracle's JDK 21 java manual says flags that control intrinsics, including -XX:+UseBASE64Intrinsics, now require -XX:+UnlockDiagnosticVMOptions first. That option unlocks diagnostic JVM options, which the same manual says are disabled by default and are not supported. If you use the Base64 flag, put -XX:+UnlockDiagnosticVMOptions before it, and treat the whole command line as a temporary measure. The same manual describes -XX:UseAVX= as enabling a specified AVX version, defaulting to the highest supported. Setting it to 2 limits the JVM to AVX2, so AVX-512 paths, including the Base64 decoder, are not used. That turns off all AVX-512 paths, not only the Base64 intrinsic, so the performance effect is wider than the Base64 flag alone. Still plan the upgrade.

  4. Reduce exposure

    Our advice: keep these applications behind SSO or VPN, and off the public internet where you can, until the runtime is patched or the temporary flag is in place.

  5. Look for signs of misuse

    Our pattern, not vendor indicators: unexpected JVM crashes or new hs_err_pid*.log files; abnormally long Base64 strings in authentication headers, upload or API logs; shells or curl/wget spawned by Java service accounts; new outbound connections from application servers. If you find any, isolate, collect hs_err and request logs, rotate application and service secrets, and patch.

Proactive: Vendor Steps, Then Ours

Vendor advice is labelled as such. Everything else is our advice.

Vendor advice

  • OpenJDK and Oracle: upgrade as soon as possible to the 17 November 2026 release. For this CVE, the confirmed fixed builds in the marked lines are 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1. The builds listed as affected, including 17.0.21.0.3, 21.0.13.0.3, 25.0.5.0.3 and 27.0.1.0.3, should be replaced.
  • Oracle: apply the November 2026 Critical Security Patch Update patches for Oracle Java SE and the listed GraalVM products, using the Patch Availability Documents linked from Oracle's advisory. Oracle GraalVM for JDK 21.0.13.1 is confirmed: its release notes say it is built on Oracle JDK 21.0.13.1+4. Treat Oracle GraalVM for JDK 17.0.21.1 and GraalVM Enterprise Edition 21.3.20.1 as under review, because their notes say they are built on Oracle JDK 17.0.20.1+4.
  • Amazon Corretto: the 17 November 2026 changelogs name CVE-2026-114938 in Corretto 17.0.21.1.2, 21.0.13.1.2, 25.0.5.1.2 and 27.0.1.1.4.
  • Azul Zulu: the November 2026 release notes include CVE-2026-114938 in the risk matrix for lines 17, 21, 25 and 27, and list these November CSPU PSU builds: Zulu 17.70.201 (CA) and 17.70.202 (SA), based on JDK 17.0.21.1+2; 21.54.201 and 21.54.202, based on 21.0.13.1+2; 25.38.201 and 25.38.202, based on 25.0.5.1+2; and 27.30.201 and 27.30.202, based on 27.0.1.1+4.
  • The official OpenJDK and Oracle advisories do not publish a workaround flag, indicators of compromise, or a list of affected applications.

Our advice

  • Inventory Java runtimes on internet-facing Jira, Confluence, Jenkins, Keycloak, Artifactory and PingFederate first, then other Java services, including those that ship a bundled JRE.
  • Check whether the host CPU exposes AVX-512 VBMI. BleepingComputer reports the hardware-accelerated decoder is only used on those x86-64 CPUs. Absence of the flag lowers the likelihood, but do not use that as a reason to skip the upgrade.
  • Patch the runtime, or apply the temporary JVM flags reported by BleepingComputer if you cannot patch immediately, and keep the applications behind SSO or VPN.
  • Apply application-vendor bundled-JRE updates as they ship. Amazon Corretto and Azul Zulu have published builds that name this CVE, as listed under Vendor advice. When we checked on 18 November 2026, Eclipse Temurin's latest GA releases in the Adoptium API were jdk-17.0.21+9, jdk-21.0.13+9, jdk-25.0.5+8 and jdk-27.0.1.0.2+1, not the 17 November 2026 .1 builds. Microsoft's OpenJDK release notes, last updated 4 November 2026, did not list a .1 build or this CVE; the latest versions named there were 17.0.21.0.2, 21.0.13.0.2 and 25.0.5.0.2. Recheck those vendors, and do not assume a product installer already includes a fixed runtime.

BleepingComputer reports unauthenticated remote code execution, in default configurations, in Keycloak, Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins and PingFederate, and authenticated remote code execution in Apache Kafka, Apache Solr, Elasticsearch, Gerrit and Spinnaker. The same report says security researchers also triggered the flaw in sample applications built with Apache Struts and Spring Boot, and that there is no universal exploit: an attacker needs an application-specific way to feed Base64 to the decoder. Product names are from that report, not from OpenJDK or Oracle.

Free Detection Pack

Built by the CybaVerse threat hunt team for businesses, internal IT teams and MSPs. Check exposure offline and hunt for signs of misuse. Everything is read-only. Pack version 1.0.0 is the download on this page. The KQL and Sigma have not been run against live data.

Verdicts

  • NOT AFFECTEDThe Java version is a fixed build, or a release line OpenJDK and Oracle do not list for this CVE (exit 0).
  • AFFECTEDThe version is affected and no workaround flag was seen, with AVX-512 VBMI present or unknown (exit 2).
  • UNKNOWNNo usable version, a line the advisories do not cover, or a build the sources disagree on (REVIEW) (exit 3).
  • MITIGATEDThe version is affected, but the JVM arguments carry an effective temporary workaround flag (exit 4).
  • AFFECTED, CPU PRECONDITION ABSENTThe version is affected, no workaround was seen, and AVX-512 VBMI is absent on this CPU (exit 6).

Exit codes

CodeMeaning
0NOT AFFECTED
1Not used
2AFFECTED
3UNKNOWN or REVIEW
4MITIGATED
5Usage or read error
6AFFECTED, CPU precondition absent

The full pack zip contains 62 files (pack version 1.0.0). The checker marks GraalVM for JDK 17.0.21.1 and GraalVM Enterprise Edition 21.3.20.1 as REVIEW, because the vendor notes conflict (see the versions section above).

Know the limits

  • Both checkers are read-only and offline. They open no sockets, write no files, never run Java and contain no exploit payloads.
  • OpenJDK and Oracle publish no indicators of compromise. Every hunt and Sigma rule in this pack is OUR PATTERN, not a vendor IOC.
  • A hunt hit is a lead to investigate. It does not prove that code ran.
  • The hunts look back from 17 November 2026. That start date is our choice.
  • None of the KQL or Sigma has been run against live data.

What is in the pack

FileWhat it does
scripts/check-java4shell.shOffline version, CPU-flag and JVM-argument checker for Linux and macOS (bash)
scripts/Check-Java4Shell.ps1The same checker in PowerShell 7
KQL (6 hunts)Microsoft Sentinel and Defender hunts Q1 to Q6, all OUR PATTERN, from 17 November 2026
Sigma (6 rules)All OUR PATTERN, plus Kusto conversions
tests/Fixtures, bash and Pester tests and a sandboxed no-network run
README.mdVerified facts, vendor vs our advice, checker usage, limits and sources

Reactive: How to Hunt (Our Advice)

All of the following is OUR PATTERN. OpenJDK and Oracle have published no indicators of compromise for this CVE. A hunt hit is a lead to investigate, not proof that code ran.

  • Unexpected JVM crashes, or new hs_err_pid*.log files, on Java application servers.
  • Abnormally long Base64 strings in authentication headers, file-upload fields or API logs. We do not describe a crafted input pattern.
  • Shells (sh, bash, cmd, PowerShell) or curl/wget spawned by Java service accounts.
  • New outbound connections from application servers that do not normally make them.

If you find any of these: isolate the host, collect hs_err logs and relevant request logs, rotate application and service secrets, and patch the Java runtime (and the application's bundled JRE, if it has one) before returning the service to use. Rebuild from known-good media if you have evidence of a compromise.

Java4Shell CVE-2026-114938 FAQ

What is CVE-2026-114938, and why is it called Java4Shell?

CVE-2026-114938 is the official identifier. OpenJDK lists it under hotspot/compiler, and Oracle under the Compiler component of Oracle Java SE and GraalVM. The official CVE record describes a difficult-to-exploit, unauthenticated, network-reachable vulnerability that can result in takeover of the Java runtime, by supplying data to APIs in that component, for example through a web service. BleepingComputer reports that security researchers call it Java4Shell and describe it as memory corruption in the hardware-accelerated Base64 decoder. OpenJDK and Oracle do not use the Java4Shell name in their advisories.

How serious is it, and whose score should I use?

Use both, and keep them labelled. OpenJDK and Oracle score it CVSS 3.1 8.1 High, with high attack complexity. BleepingComputer reports that security researchers score it CVSS 4.0 9.2 Critical. The vendor score is not Critical. We frame the page as Critical by impact because of the official takeover wording and the reported effect on default installs of widely used applications.

Which versions are affected, and what is the fixed build?

The OpenJDK advisory says, of the 17 November 2026 release as a whole: "The affected versions are 27.0.1.0.3, 25.0.5.0.3, 21.0.13.0.3, 17.0.21.0.3, 11.0.33, 8u512, and earlier." Its risk matrix marks CVE-2026-114938 only on the 17, 21, 25 and 27 lines, not 8 or 11. Oracle's Java SE risk matrix, under the heading "Supported Versions Affected", lists 8u511b33-perf, 17.0.21.0.3, 21.0.13.0.3, 25.0.5.0.3 and 27.0.1.0.3, plus GraalVM for JDK 17.0.21.0.3 and 21.0.13.0.3 and GraalVM Enterprise Edition 21.3.20.0.3. Those are the latest supported versions listed as affected, not the fixed builds. Oracle's own 17.0.21.0.3 release notes (24 October 2026) said not to use that build after the next critical patch update on 17 November 2026. The confirmed 17 November 2026 fixed builds are 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1 (full version strings 17.0.21.1+4, 21.0.13.1+4, 25.0.5.1+4 and 27.0.1.1+4), plus 8u513-perf for the listed Java 8 performance build, and Oracle GraalVM for JDK 21.0.13.1. Oracle GraalVM for JDK 17.0.21.1 and GraalVM Enterprise Edition 21.3.20.1 are under review, because their vendor notes conflict (they say they are built on Oracle JDK 17.0.20.1+4, which is below the 17.0.21.1 fix). OpenJDK and Oracle agree that 17.0.21.0.3 and the matching 21, 25 and 27 builds are affected. BleepingComputer reports a start version of OpenJDK 17.0.6 and later; that start version is from the press report, not from the official advisories.

Has this been exploited, and is it in KEV?

No in-the-wild exploitation has been reported in the sources we checked. The CVE was not in CISA's Known Exploited Vulnerabilities catalogue when we checked at 18 November 2026, 18:09 GMT (catalogue 2026.11.17, released 17:53 GMT on 17 November 2026). CISA's ADP entry on the CVE record lists Exploitation as "none" (updated 22:02 GMT on 17 November 2026). We do not treat absence from KEV, or that ADP label, as proof that a flaw is unused. BleepingComputer says a memory-corruption proof of concept is public, and that application-specific exploits have not been released. We do not claim a public weaponised exploit, and we do not link exploit code. Recheck KEV.

Is there a workaround if I cannot patch today?

The official OpenJDK and Oracle advisories do not publish a workaround. BleepingComputer reports that security researchers recommend starting the JVM with -XX:-UseBASE64Intrinsics, or -XX:UseAVX=2 to turn off AVX-512, as a temporary measure. Oracle's JDK 21 java manual says intrinsic flags now require -XX:+UnlockDiagnosticVMOptions first, and that -XX:UseAVX=2 selects AVX version 2, which turns off all AVX-512 paths, not only the Base64 intrinsic. Still plan the upgrade, and apply application-vendor JRE updates as they ship.

What is the vendor's official advice?

Vendor advice: upgrade as soon as possible to the 17 November 2026 OpenJDK release, and apply the Oracle Java SE and GraalVM patches in the November 2026 CSPU. Until those patches are applied, the official advisories do not offer a documented workaround.

Official Sources