Java4Shell CVE-2026-114938: Patch Java Now, What to Do
TL;DR · the short version
CVE-2026-114938, named "Java4Shell" by the finder Calif, is a heap-based buffer overflow in the Base64 decoding path of the HotSpot Java virtual machine on x86-64 processors with AVX-512 VBMI. On 17 November 2026 Oracle fixed it in its November Critical Security Patch Update (Java SE 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1), OpenJDK published an advisory marking the 17, 21, 25 and 27 lines affected, and CISA, with the Canadian Centre for Cyber Security, urged immediate patching. Oracle scores it CVSS 3.1 8.1 (high attack complexity, no login needed) and says a successful attack can take over the Java runtime; CISA says crafted Base64 input could crash the JVM and allow remote code execution. No source reports exploitation in the wild and the CVE was not in CISA KEV on 18 November, but the finder has published a limited trigger and says it demonstrated pre-auth code execution on Jira, Confluence, Jenkins and Keycloak in testing, with those exploits held back until people have patched. Find every Java runtime, including those bundled in products, containers and appliances, patch internet-facing ones first, and use the documented stopgaps where you cannot.
CISA is urging organisations to patch Java now. On 17 November 2026 Oracle fixed CVE-2026-114938, a memory-corruption flaw in the Base64 decoder of the HotSpot Java virtual machine (JVM) on x86-64 processors, and CISA said crafted Base64 input could crash the JVM and allow unauthenticated remote code execution. Calif, the research firm whose researcher reported it, named it "Java4Shell" and says it took over Jira, Confluence, Jenkins and Keycloak servers in testing without logging in. Oracle's fixed builds are JDK 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1. No source reports exploitation in the wild yet.
Who this is for: CISOs, vulnerability-management and platform leads, and anyone who runs Java services or buys products that ship their own Java.
Based on CISA, Oracle, OpenJDK, the CVE record and the Canadian Centre for Cyber Security, all read on 18 November 2026. PlayCISO is not affiliated with Oracle or Calif. We publish no exploit details.
Confirmed vs not yet known
Oracle confirms: CVE-2026-114938 in Java SE (component Compiler), remotely exploitable without authentication, CVSS 3.1 8.1 (AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H), and that a successful attack can result in takeover of Java SE, GraalVM for JDK and GraalVM Enterprise Edition. Oracle's note says it is reached by supplying data to the affected APIs, "such as through a web service". The CVE record classes it CWE-122 (heap-based buffer overflow).
OpenJDK confirms: the 17, 21, 25 and 27 release lines are affected; its matrix does not mark 8 or 11.
Not exploited, as far as anyone says: CISA's enrichment of the CVE record says "Exploitation: none" and "Automatable: no" (17 Nov); the Canadian Cyber Centre is "not aware of active exploitation at this time"; the CVE was not in the CISA KEV catalog on 18 November.
Claimed by the finder, not confirmed by the vendors named: Calif says it took over Jira, Confluence, Jenkins and Keycloak servers (and other Java products) without logging in, in its own testing of default configurations, and that its write-up contains a limited proof of concept that triggers the bug, with full exploits held back for now.
Not yet known: fixed version numbers for every third-party JDK build (Temurin, Corretto, Zulu, Microsoft, Red Hat and others), which products that bundle Java are affected and when they will ship updates, and whether exploitation will follow. Severity labels differ: Oracle says 8.1 (High); CISA and the Cyber Centre call it critical; Calif gives 9.2 under CVSS 4.0.
Affected and fixed versions
| Product | Affected (Oracle's listing) | Fixed build |
|---|---|---|
| Oracle Java SE 17 | 17.0.21.0.3 | 17.0.21.1 (17 Nov 2026) |
| Oracle Java SE 21 | 21.0.13.0.3 | 21.0.13.1 (17 Nov 2026) |
| Oracle Java SE 25 | 25.0.5.0.3 | 25.0.5.1 (17 Nov 2026) |
| Oracle Java SE 27 | 27.0.1.0.3 | 27.0.1.1 (17 Nov 2026) |
| Oracle Java SE 8 (Performance Pack) | 8u511b33-perf | See Oracle's patch availability document |
| Oracle GraalVM for JDK | 17.0.21.0.3, 21.0.13.0.3 | See Oracle's patch availability document |
| Oracle GraalVM Enterprise Edition | 21.3.20.0.3 | See Oracle's patch availability document |
| OpenJDK and builds made from it | 17, 21, 25 and 27 lines (OpenJDK); from 17.0.6 per Calif | Your vendor's November 2026 security build |
Oracle lists only the versions it still supports and says older releases are likely affected too. The standard Java 8 and Java 11 lines are not marked affected by Oracle or OpenJDK for this CVE, although both received fixes for other CVEs in the same release. When we checked on 18 November, the latest Eclipse Temurin builds in Adoptium's release API were still the October releases, so do not assume a third-party build is fixed until its release notes name CVE-2026-114938.
Hardware matters. The Canadian Cyber Centre says the flaw affects Java processes on x86-64 processors with AVX-512 Vector Byte Manipulation Instructions (VBMI). Calif names recent Intel Xeon and AMD Zen 4 or newer chips and says ARM64 is not affected. Cloud instances move between CPU generations, so use this to set priority, not to skip the patch.
What Calif says it found
Calif, the security firm whose researcher Quang Luong reported the bug (Oracle credits him alongside Albert Yang), published its own write-up on 17 November. Everything in this section is Calif's account, not confirmed by Oracle or by the product vendors named.
- Takeovers without logging in, in testing. Calif says its "initial testing demonstrated remote code execution" without authentication against Keycloak, Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins and PingFederate in their default configurations, and after login against several more Java products. It calls the list non-exhaustive. This is a researcher's demonstration, not exploitation in the wild.
- Where the bug is. Calif says the flaw sits in the Base64 decoding code, where malformed input makes the decoder write past the end of its memory buffer. We do not reproduce the input format.
- Found with AI, quickly and cheaply. Calif says Codex, OpenAI's coding agent, brought its researcher the finding, that discovery took about an hour and the first working exploit two more hours, and that the total token cost was under $200. These figures are the researcher's own; nobody else has checked them.
- Exploits held back. Calif says it is withholding its application-specific exploits "until everyone has had a chance to upgrade" and will publish them later, with no date given. Its write-up does include a limited trigger for the bug itself. Plan on working exploits becoming public.
Vendor advisories: when we checked on 18 November, the Jenkins security advisories, the Keycloak security page and Atlassian's security advisories did not mention CVE-2026-114938. Until they do, patching the Java runtime under each product is on you.
Useful next, if this was relevant to you
Timeline
| Date | Event | Source |
|---|---|---|
| 22 Oct 2026 | Oracle reserves CVE-2026-114938 | CVE record |
| 17 Nov 2026 | Oracle's November Critical Security Patch Update fixes it (credits Albert Yang, and Quang Luong of Calif); JDK 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1 released | Oracle |
| 17 Nov 2026 | OpenJDK Vulnerability Advisory marks the 17, 21, 25 and 27 lines affected | OpenJDK |
| 17 Nov 2026 | Calif publishes its "Java4Shell" write-up: says it demonstrated pre-auth code execution on Jira, Confluence, Jenkins, Keycloak and others in testing, and that it is holding back those exploits | Calif |
| Before 17 Nov 2026 (undated) | Calif says its researcher found the bug with Codex in about an hour and built a working exploit in two more, for under $200 in AI costs | Calif |
| 17 Nov 2026 | CVE record published; CISA enrichment: exploitation none, not automatable, technical impact total | CVE record |
| 17 Nov 2026 | Canadian Centre for Cyber Security alert AL26-029 | Cyber Centre |
| 17 Nov 2026 | CISA alert "Critical Java Vulnerability Requires Immediate Patching", with the Cyber Centre | CISA |
| 18 Nov 2026 | No Atlassian, Jenkins or Keycloak advisory mentions the CVE yet | Atlassian, Jenkins, Keycloak |
| 18 Nov 2026 | Not in the CISA KEV catalog (catalog version 2026.11.17); NVD record still awaiting analysis | CISA KEV, NVD |
What security leaders should do this week
- Find every Java runtime, not just the ones in your inventory. CISA warns that JDKs bundled in third-party applications, container images and appliances may not appear as separate entries. Search hosts for the JVM library (
libjvm.soon Linux,jvm.dllon Windows) rather than only thejavacommand, scan container images and registries for JDK and JRE layers, read the SBOMs your suppliers give you for a bundled runtime, and ask every appliance and software vendor in writing whether their product ships Java and when they will ship a fixed build. Record vendor and full build version for each. - Patch in order of exposure. Internet-facing Java applications first, starting with the kinds Calif says it took over in testing: Jira, Confluence, Jenkins and Keycloak, plus other identity providers, CI servers, wikis and artifact repositories that accept data without a login. Many of these ship or pin their own Java, so update the runtime the product actually uses and watch for each vendor's advisory, then internal services, then developer machines and build agents. A runtime update needs a restart and, for containers, a rebuilt and redeployed image: confirm the version that is actually running.
- Use the stopgaps where you cannot patch this week. CISA: validate and restrict Base64 input at the application boundary, cap input size and limit unauthenticated access. The Cyber Centre: isolate internet-facing applications and, if you cannot patch, start the JVM with
-XX:UseAVX=2to turn off AVX-512 (x86-64 only; test the performance cost). Calif also names-XX:-UseBASE64Intrinsics. Neither flag fixes the bug; log them as temporary exceptions with an end date. - Watch for signs of trouble. Unexpected JVM crashes are the first thing CISA names: collect HotSpot fatal error logs (
hs_err_pid<pid>.log, written in the working directory or wherever-XX:ErrorFilepoints) and alert on repeated restarts of a Java service. Alert on Java processes starting shells or scripting tools, unusual outbound connections from Java services, and Java runtimes appearing where none were approved. Review web, application and authentication logs for the internet-facing services you could not patch. - Prioritise with data, and re-check daily. Check the CVE in the KEV checker and EPSS lookup (no EPSS score had been published on 18 November), then set the deadline with the SSVC calculator or the vulnerability prioritizer. If it enters KEV, federal agencies get a fixed clock; the BOD 26-04 calculator shows it.
- Record it in your risk register with an owner and a date, including the vendor products you cannot patch yourself.
Severity: why the numbers differ
Oracle's 8.1 rests on "high" attack complexity: exploitation is not trivial and depends on the application and environment. CISA's enrichment agrees it is not automatable but rates the technical impact "total". CISA's alert and the Cyber Centre both call it critical, and Calif's own CVSS 4.0 score is 9.2. For prioritisation, the facts that matter are that it needs no login, Java is everywhere, and a trigger is already public.
For Oracle's view of Oracle's other incidents and vulnerabilities, see Oracle in the Breach Index. Java4Shell is a product vulnerability, not a breach of Oracle.
Frequently asked questions
What is Java4Shell (CVE-2026-114938)?
A heap-based buffer overflow (CWE-122) in the Base64 decoding path of the HotSpot Java virtual machine on x86-64 processors with AVX-512 VBMI instructions. Oracle lists it under Java SE, component Compiler, and says an unauthenticated attacker with network access can use it to take over Java SE, GraalVM for JDK and GraalVM Enterprise Edition. CISA says crafted Base64 input could crash the JVM and allow remote code execution. The name "Java4Shell" was given by Calif, whose researcher reported the flaw.
Which Java versions are affected by CVE-2026-114938?
Oracle lists Java SE 8u511b33-perf, 17.0.21.0.3, 21.0.13.0.3, 25.0.5.0.3 and 27.0.1.0.3, GraalVM for JDK 17.0.21.0.3 and 21.0.13.0.3, and GraalVM Enterprise Edition 21.3.20.0.3. The OpenJDK advisory marks the 17, 21, 25 and 27 lines as affected and not 8 or 11. Calif says affected OpenJDK builds start at 17.0.6, and that builds from other vendors (Temurin, Corretto, Zulu, Microsoft, Red Hat and others) share the bug because they are built from OpenJDK.
Which Java version fixes Java4Shell?
Oracle released JDK 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1 on 17 November 2026 with the November Critical Security Patch Update. If your Java comes from another vendor, install that vendor's November 2026 security build; we could not confirm every vendor's fixed version when this was written, so check their release notes for CVE-2026-114938.
Is CVE-2026-114938 critical?
It depends who you ask. Oracle scores it CVSS 3.1 8.1 (High), because exploitation is rated "high" complexity. CISA and the Canadian Centre for Cyber Security call it critical, and Calif scores it 9.2 under CVSS 4.0. Treat it as urgent either way: it needs no login and Java sits behind many internet-facing services.
Is Java4Shell being exploited?
No source we read reports exploitation in the wild. CISA's enrichment of the CVE record says "Exploitation: none" as of 17 November, the Canadian Cyber Centre says it is not aware of active exploitation, and the CVE was not in the CISA KEV catalog on 18 November. Calif says its write-up includes a limited proof of concept that triggers the bug and that it is holding back application-specific exploits for now, so expect that to change.
Are Jira, Confluence, Jenkins and Keycloak affected by Java4Shell?
They run on Java, so they are exposed if they run on an affected JVM and accept Base64 data an attacker controls. Calif says its initial testing showed remote code execution without a login against Keycloak, Jira, Confluence, Jenkins and other Java products in their default configurations. That is the researcher's demonstration, not reported exploitation in the wild, and Atlassian, Jenkins and Keycloak had not published advisories mentioning CVE-2026-114938 when we checked on 18 November. Patch the Java runtime these products use (many ship or pin their own), follow each vendor's advisory when it appears, and put these servers first if they face the internet.
Did AI find Java4Shell?
Calif says so. Its write-up says its researcher Quang Luong pointed Codex, OpenAI's coding agent, at the area, that Codex brought back the Base64 decoder finding, that finding it took about an hour and building the first working exploit two more, and that the total token cost was under $200. That is Calif's own account; no other source we read confirms the figures.
How do I mitigate Java4Shell if I cannot patch yet?
CISA advises validating and restricting Base64 input at the application boundary, size limits on input, and limiting unauthenticated access. The Canadian Cyber Centre suggests isolating internet-facing applications and, if you cannot patch, starting the JVM with -XX:UseAVX=2 to turn off AVX-512. Calif also names -XX:-UseBASE64Intrinsics. Use the AVX flag only on x86-64 hosts (Calif says JVMs on ARM will not start with it), test performance first, and treat both as stopgaps until you patch.
Am I affected if my servers run on ARM?
Calif says ARM64 is not affected, and the Canadian Cyber Centre describes the flaw as affecting x86-64 processors with AVX-512 VBMI. Oracle's advisory does not mention CPU type, so patch anyway: the same images often run on both kinds of hardware.
Sources
- CISA: Critical Java Vulnerability Requires Immediate Patching (17 November 2026)
- Oracle: Critical Security Patch Update Advisory, November 2026 (Java SE risk matrix, Note 1, credit statement)
- OpenJDK Vulnerability Advisory: 2026/11/17
- CVE.org: CVE-2026-114938 (CNA Oracle; CISA ADP enrichment)
- NVD: CVE-2026-114938 (status Received on 18 November 2026)
- Canadian Centre for Cyber Security: AL26-029, Critical vulnerability impacting Oracle Java (17 November 2026)
- CISA Known Exploited Vulnerabilities catalog (checked 18 November 2026, version 2026.11.17: not listed)
- Oracle JDK release notes: 17.0.21.1, 21.0.13.1, 25.0.5.1, 27.0.1.1
- Calif: Java4Shell (the finder's write-up; source of the name, CPU scope and the mitigation flags; its application claims are the firm's own)
- Checked for vendor advisories on 18 November 2026 (none mentioned the CVE): Atlassian, Jenkins, Keycloak
- oss-security mailing list: CVE-2026-114938 summary (17 November 2026)