Tuskira AdvisoryNot observedLast updatedNovember 17, 2026

CVE-2026-114938

How to tell if CVE-2026-114938 (Java4Shell, OpenJDK Base64 decoder) is reachable in your environment

Latest change
First published. Oracle and OpenJDK fixed it November 17; Calif showed pre-auth RCE on default installs of six products.

Reachable if an x86-64 host with AVX-512 VBMI runs an affected Java 17, 21, 25 or 27 build, and a service on it decodes Base64 it received from someone else, or a Java program on it connects out to a server you don't control. In the products Calif tested, no login was needed

Affected
OpenJDK 17, 21, 25 and 27 runtimes, including Oracle Java SE and GraalVM, on x86-64 CPUs with AVX-512 VBMI. ARM is not affected
Exploitation
No exploitation reported. Not on CISA KEV. Calif showed pre-auth RCE on default installs of six products and is holding its exploits back.
Fix
Oracle Java SE 17.0.21.1, 21.0.13.1, 25.0.5.1 or 27.0.1.1. Other distributions, check your vendor's build.
Can't patch today
Start Java with -XX:-UseBASE64Intrinsics and restart, per Calif. Oracle has not endorsed it

Patching your apps won't fix this one. CVE-2026-114938 is in the Java runtime itself, so every Java service on an affected runtime and CPU carries it until the runtime is updated or the service is restarted with the stopgap flag. Calif, which found it, says it got remote code execution with no login against default installs of Keycloak, Jira, Confluence, Artifactory, Jenkins and PingFederate.

Oracle, which issued the CVE, and OpenJDK rate it 8.1 High and call the attack hard to pull off. Calif rates it 9.2 Critical on CVSS 4.0. It only affects x86-64 servers whose CPUs support AVX-512 VBMI, which Calif says means Intel Xeon Ice Lake and newer and AMD Zen 4 and newer. ARM servers are not affected. Oracle and OpenJDK published fixes on November 17, 2026.

No exploitation has been reported, and it is not on CISA's Known Exploited Vulnerabilities catalog. Calif published a safe trigger that only proves the bug, and says it is holding back its product-specific exploits until people have had time to upgrade. Calif also says there is no universal exploit, because each application needs its own attack.

Is it reachable for you?

What it is. Java has a faster way to decode Base64 text on newer Intel and AMD chips. Calif found that a malformed Base64 string makes that fast path write past the end of its memory buffer. With careful work, Calif turned that into reading server memory and running code. Base64 shows up in login headers, tokens and file uploads, so many applications decode it from untrusted input before anyone logs in.

Three conditions, all needed. The runtime is an affected build, the CPU supports AVX-512 VBMI, and the application decodes Base64 it received from someone else. Calif also reports a client-side case. A Java program that opens a TLS connection to a malicious server can be compromised even if it exposes no service.

Checks, in three tiers. The first decides exposure. The second and third decide priority. Run these only on systems your organization owns.

  1. Which Java runtimes are on x86-64 hosts with AVX-512 VBMI? List every Java runtime version, including the ones bundled inside products. On Linux, Calif's check is that the CPU flags include avx512f, avx512bw, avx512vbmi and bmi2; all four must be present. ARM hosts and older x86 chips drop out here.
  2. Which of those services take Base64 from the network? Start with the products Calif named. Then look at anything that handles Basic authentication, tokens or file uploads from users who haven't logged in. Services reachable from the internet come first.
  3. Which Java programs connect out to servers you don't control? Calif's client-side case means outbound TLS connections count too, so include tools and agents that call external services.
What Tuskira checks for this flaw in your environment
Java services on affected runtimes running on AVX-512 VBMI hosts and reachable from the internet
What those Java services can reach once compromised
Across every system you run.
See how Tuskira finds exposed systems

Compensating controls (test them, don't assume them)

Oracle and OpenJDK say to update. Calif gives two startup flags for runtimes that can't be updated yet. Oracle has not endorsed them, and each needs a restart.

  • Turn off the fast Base64 path. Start Java with -XX:-UseBASE64Intrinsics, per Calif. Base64-heavy applications may slow down.
  • Or turn off AVX-512. Start Java with -XX:UseAVX=2, per Calif. Use it only on x86-64 hosts. Calif warns that Java on ARM will not start with this flag.
  • Don't count on a proxy. Calif says request buffering in a proxy breaks some of its attacks but is not a general mitigation.

What to look for in logs

Neither Oracle nor OpenJDK has published indicators of compromise. Calif notes that a failed attempt can crash the Java process, so unexplained JVM crashes on affected hosts are worth a look, but no vendor has published a detection rule.

FAQ

Am I exposed to CVE-2026-114938?

You are if a Java service on an affected runtime runs on an x86-64 CPU with AVX-512 VBMI and decodes Base64 it receives from someone else. ARM servers and older x86 chips are not affected.

Is CVE-2026-114938 being exploited?

No exploitation has been reported, and it is not on CISA KEV. Calif has held back its product-specific exploits for now.

Why do the scores differ?

Oracle and OpenJDK rate it 8.1 on CVSS 3.1 and call the attack hard to pull off. Calif rates it 9.2 on CVSS 4.0. Calif's results on default installs of named products are why we cover it.

Is Java 8 or 11 affected?

OpenJDK lists 17, 21, 25 and 27. Oracle also lists one Java 8 build, the Java SE 8 Enterprise Performance Pack (8u511b33-perf), fixed in 8u513-perf.

I use Temurin, Corretto or Zulu. Which build do I need?

Each distribution ships its own builds. Check your vendor's advisory and confirm the build lists CVE-2026-114938. Until then, the startup flag is the stopgap.

Vendor
Oracle and the OpenJDK project
Product
OpenJDK and the Java runtimes built from it, including Oracle Java SE
Affected versions
OpenJDK 17, 21, 25 and 27 (OpenJDK advisory), from 17.0.6 onward per Calif. Oracle Java SE 17.0.21.0.3, 21.0.13.0.3, 25.0.5.0.3, 27.0.1.0.3 and 8u511b33-perf and earlier, and Oracle GraalVM, per Oracle. Only on x86-64 CPUs with AVX-512 VBMI
Vulnerable component
HotSpot's x86-64 Base64 decoder fast path (hotspot/compiler), used only on CPUs with AVX-512 VBMI
CVSS
8.1 High (Oracle as CNA and OpenJDK, CVSS 3.1, attack complexity High). Calif rates it 9.2 Critical on CVSS 4.0
CVSS vector
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H (Oracle)
CWE
Out-of-bounds write (heap), per Calif's write-up
Exploitation
Not observed
CISA KEV added
Patch
Oracle Java SE 17.0.21.1, 21.0.13.1, 25.0.5.1, 27.0.1.1 and 8u513-perf (Oracle's November security baselines). OpenJDK published the fix on November 17, 2026. Other distributions such as Temurin, Corretto, Zulu and Red Hat ship on their own schedules, so confirm your vendor's build names CVE-2026-114938

Updates

Nov 17, 2026. First published. Oracle and OpenJDK fixed CVE-2026-114938 on November 17; Calif published its write-up the same day. No exploitation reported; not on CISA KEV.

Sources

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

We publish an advisory when a vulnerability is Critical or on CISA KEV, has one identifiable component to check, and has a public proof of concept or reported exploitation. Advisories are updated as exploitation status changes.