November 19, 2026
Java4Shell (CVE-2026โ114938): the Oracle Java / OpenJDK pre-auth RCE, and how to hunt for it
What happened, who is affected, the fixed versions, and how to hunt, as of the morning of November 18, 2026.

By Mario Velasco
4 min read
Written with AI assistance and checked against the cited sources. The always-updated version lives on SockAnalyst.
What happened
On November 17, 2026, Oracle and the OpenJDK Vulnerability Group fixed CVE-2026โ114938, nicknamed Java4Shell, and CISA issued an alert the same day urging organizations to patch immediately. The bug is a heap out-of-bounds write in the hand-written AVX-512 VBMI HotSpot intrinsic that speeds up java.util.Base64 decoding on x86โ64 CPUs. A malformed Base64 value whose final fragment is only padding makes the routine compute a negative output length, and a 64-byte masked store then writes 62 bytes past the end of the destination array. Calif's trigger is 64 'A' characters followed by a single '='; on older 17.x builds it is 128 'A's plus '='. Java throws an exception afterwards, but neighbouring heap objects are already corrupted.
Calif, which found the flaw with AI-assisted auditing, says it reached unauthenticated remote code execution in default configurations of Keycloak, Atlassian Jira and Confluence, JFrog Artifactory, Jenkins and PingFederate, plus post-auth RCE in Kafka, Solr, Elasticsearch, Gerrit and Spinnaker, a Heartbleed-style memory leak, and a client-side exploit against Java programs that connect to a malicious TLS server. Datadog Security Labs cautions that crashing the JVM is the easier outcome and that reliable RCE has to be tailored to each application and JVM; there is no universal exploit. Oracle and OpenJDK score it CVSS 3.1 8.1 (AC:H). Calif scores it CVSS 4.0 9.2, and Red Hat's preliminary CVSS 3.0 score is 9.8. Oracle's November 2026 Critical Security Patch Update fixes it.
Who's affected
Any Java service running OpenJDK-based Java 17.0.6 or later, 21, 25 or 27 on an x86โ64 CPU with AVX-512 VBMI is in scope. That includes Intel Ice Lake, Sapphire Rapids, Emerald Rapids and Granite Rapids, AMD Zen 4, Zen 5 and Zen 6, and cloud families on Red Hat's list such as AWS C6i/M6i/R6i and C7i/M7i/R7a, GCP C3/C3D/C4/C4D/N4/N4D/M3/M4 and Azure D-series v5 to v7. A cloud resize can silently move a workload onto an affected CPU.
Calif names these affected distributions: Oracle Java SE, Oracle OpenJDK, Oracle GraalVM for JDK, Amazon Corretto, Eclipse Temurin, Microsoft Build of OpenJDK, Azul Zulu, BellSoft Liberica, SAP SapMachine and Red Hat OpenJDK (RHEL 8, 9, 10 and anything built on RHEL/UBI images). Oracle also lists its Java SE 8u511b33-perf build and Oracle GraalVM for JDK 17.0.21.0.3 / 21.0.13.0.3. The CPU must expose avx512f, avx512bw, avx512vbmi and bmi2 (Calif). Not affected: standard Java 8 and 11, native ARM64 JVMs, and x86โ64 CPUs without AVX-512 VBMI. The app must also pass attacker-controlled Base64 to Base64.getDecoder() or getUrlDecoder(). HTTP Basic credentials, SAML messages, CSRF tokens and MIME encoded words are common paths and are often decoded before login.
The fix and versions
Inventory first. On each Linux host run java -version, and check the runtime bundled inside each product and container image, not just the one on your PATH. Red Hat's check is grep -o avx512vbmi /proc/cpuinfo | head -1; empty output means the CPU is not known to be affected. Calif's stricter check must print all four of those CPU flags. For a running process, jcmd VM.version shows the runtime actually in use.
Oracle fixed builds are 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1 (only Oracle builds ending in .1 are patched), plus 8u513 for the 8u511b33-perf build (per Sysdig). As of the morning of November 18, 2026 (CT): Amazon Corretto 17.0.21.1.2 and 21.0.13.1.2 shipped November 17; Debian has fixed packages only in unstable (sid), with bookworm and trixie still listed vulnerable; Ubuntu lists every openjdk package as "Needs evaluation"; Red Hat errata are still pending (RHSB-2026โ018: fixes being expedited); Eclipse Temurin's latest builds are still 17.0.21+9 / 21.0.13+9 / 25.0.5+8, with no fix yet; Check Point lists Azul Zulu as shipped and Microsoft and Liberica as unconfirmed, so check each vendor's page. Rebuild and redeploy container images after updating base images, then restart every Java process.
If you cannot patch now, start x86โ64 JVMs with -XX:-UseBASE64Intrinsics (Red Hat's preferred, narrower workaround) or -XX:UseAVX=2. Apply UseAVX=2 only on x86โ64 hosts; native ARM64 JVMs do not recognise it and will fail to start. Do not use -XX:DisableIntrinsic=_base64_decodeBlock on its own; without -XX:+UnlockDiagnosticVMOptions the JVM will not start (Sysdig, Check Point). Confirm with jcmd VM.flags -all that UseBASE64Intrinsics = false or UseAVX = 2, and remove the flag once patched.
Prioritise internet-facing SSO/IdP and auth services (Keycloak, PingFederate), Atlassian, Jenkins, Artifactory and anything that decodes HTTP Basic or SAML before authentication. Products that bundle their own JVM should be fixed or flagged through the product vendor's supported path. CISA's interim controls: validate and size-limit Base64 input at the application boundary and limit unauthenticated access until patched.
KEV status
As of the morning of November 18, 2026 (CT), CVE-2026โ114938 is still not in CISA's KEV catalog (latest catalog 2026.11.17). No in-the-wild exploitation has been reported. Check Point says a triggering input had been publicly visible since September 9, 2026, and Immersive's Kevin Breen says he built a working (unreleased) exploit.
How to hunt
No IOCs (IPs, domains or hashes) have been published, so hunt for behaviour. The published trigger is a Base64 value made of a long run of 'A's ending in a lone '='. Sysdig suggests flagging Base64-alphabet tokens of 32 or more characters that end in '=' with a length that is not a multiple of 4. Search back to September 9, 2026.
Look for JVM crashes (new hs_err_pid*.log files, SIGSEGV, "A fatal error has been detected by the Java Runtime Environment", crashes during garbage collection, or bursts of worker restarts), application-log bursts of "IllegalArgumentException: Input byte array has wrong 4-byte ending unit" from one source (Sysdig saw it on every attempt in tests), a java process spawning a shell, curl/wget, python or a reverse-shell tool, and malformed Base64 in Authorization: Basic headers or SAMLResponse values where a WAF, reverse proxy or API gateway logs them. Most access logs do not record the Authorization header. MITRE ATT&CK: T1190, T1203, T1059.004, T1499.004.
Keyword hunt (crash and decode-exception strings; search from September 9, 2026):
"A fatal error has been detected by the Java Runtime Environment"
hs_err_pid
"Input byte array has wrong 4-byte ending unit""A fatal error has been detected by the Java Runtime Environment"
hs_err_pid
"Input byte array has wrong 4-byte ending unit"Full Splunk and Microsoft Sentinel (KQL) hunt queries are in the SockAnalyst playbook: https://sockanalyst.com/patch-now/oracle-java-openjdk-cve-2026-114938-java4shell-base64-avx512-rce?utm_source=medium&utm_medium=referral&utm_campaign=2026-w47
Primary sources
Oracle Critical Security Patch Update Advisory, November 2026: https://www.oracle.com/security-alerts/cspunov2026.html
OpenJDK Vulnerability Advisory: 2026/11/17: https://openjdk.org/groups/vulnerability/advisories/2026-11-17
CISA: Critical Java Vulnerability Requires Immediate Patching (Nov 17, 2026): https://www.cisa.gov/news-events/alerts/2026/11/17/critical-java-vulnerability-requires-immediate-patching
Calif: Java4Shell (CVE-2026โ114938) technical write-up: https://calif.io/java4shell/
Red Hat RHSB-2026โ018: OpenJDK CVE-2026โ114938 "java4shell": https://access.redhat.com/security/vulnerabilities/RHSB-2026-018
Amazon Corretto 21 changelog (21.0.13.1.2, Nov 17, 2026): https://github.com/corretto/corretto-21/blob/develop/CHANGELOG.md
Debian security tracker: CVE-2026โ114938: https://security-tracker.debian.org/tracker/CVE-2026-114938
Ubuntu: CVE-2026โ114938: https://ubuntu.com/security/CVE-2026-114938
Datadog Security Labs: Java4Shell: Remote Code Execution in OpenJDK (CVE-2026โ114938): https://securitylabs.datadoghq.com/articles/java4shell-remote-code-execution-in-openjdk-cve-2026-114938/
Sysdig: Java4Shell (CVE-2026โ114938): Pre-auth RCE in OpenJDK: https://www.sysdig.com/blog/java4shell-cve-2026-114938-pre-auth-rce-in-openjdk
Check Point: Java4Shell (CVE-2026โ114938): Check Your Java Services Now (Nov 18, 2026): https://blog.checkpoint.com/security/java4shell-cve-2026-114938-check-your-java-services-now/
Tenable: Java4Shell (CVE-2026โ114938) FAQ: https://www.tenable.com/blog/java4shell-cve-2026-114938-frequently-asked-questions-about-openjdk-rce-vulnerability
BleepingComputer: Critical Java4Shell flaw allows RCE attacks on major Java applications: https://www.bleepingcomputer.com/news/security/critical-java4shell-flaw-allows-rce-attacks-on-major-java-applications/
Originally published at SockAnalyst: https://sockanalyst.com/patch-now/oracle-java-openjdk-cve-2026-114938-java4shell-base64-avx512-rce
Hunt queries are starting points; test them in your own environment.