Java4Shell security alert graphic urging admins to patch Java and OpenJDK

Java4Shell Is Fixed: Which Java Servers Are at Risk, Which CPUs Matter and How to Patch Today

Java4Shell is a critical flaw in the Base64 decoder inside OpenJDK, and it can let an attacker run code on some Java servers without logging in. Oracle and the OpenJDK team shipped fixes on November 17, so if you run Java 17, 21, 25 or 27 on an AVX-512 Intel or AMD server, patch and restart this week. Most home PCs aren’t the target, but the check is quick enough that it’s worth doing anyway.

Here are the quick facts for skimmers.

  • The flaw: CVE-2026-114938, a memory bug in Java’s Base64 decoder, disclosed by security firm Calif
  • The fix: Oracle Java 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1, released November 17, 2026
  • Who’s exposed: Java 17 and newer on x86-64 chips with AVX-512 VBMI; ARM64 isn’t affected
  • Severity: 9.2 from Calif, 8.1 from Oracle
  • Attacks so far: Tenable knew of none as of November 17

What Java4Shell is and why a Base64 bug matters

Base64 turns data into plain text, and Java uses it everywhere: login headers, SAML tokens, CSRF tokens and email encoding. Java also has a speed-optimized decoder for AVX-512 chips, and that fast path is where the bug lives, according to The Hacker News.

Feed it a malformed value with one extra “=” at the end and it writes 62 or 63 bytes past the end of its output buffer. Java only rejects the input after the write has already happened. Attackers choose where the overflow lands, though not which bytes get written, and Calif says it chained that into a memory-manipulation technique it calls JavaBleed to reach code execution.

Servers often decode this kind of text before they check who you are, so an attacker may not need an account. That’s what earns the “critical” label.

Is your Java affected? Check the build and the CPU

You need both things to line up: a vulnerable Java build and a processor with the right instructions. Here are the Oracle builds Tenable lists as affected, next to the fixed versions.

Java line Affected Oracle build Fixed build
Java 8 Performance Pack 8u511b33-perf 8u513-perf
Java 17 17.0.21.0.3 17.0.21.1
Java 21 21.0.13.0.3 21.0.13.1
Java 25 25.0.5.0.3 25.0.5.1
Java 27 27.0.1.0.3 27.0.1.1

Standard Java 8 and 11 aren’t in the OpenJDK advisory. Calif says the flaw reaches OpenJDK 17.0.6 and newer, and that other distributions such as Corretto, Temurin and Zulu are affected too.

On the hardware side, the bug needs x86-64 chips with AVX-512 VBMI: Intel Xeon from Ice Lake through Granite Rapids, and AMD Zen 4 or newer. On Linux, look in /proc/cpuinfo for avx512f, avx512bw, avx512vbmi and bmi2, and note that all four must be present. The same container image can be safe on one host and exposed on another, which is the sneaky part.

Which apps could Java4Shell expose?

Calif tested popular software in default settings and named the ones it could exploit. The list isn’t complete.

  • Without logging in: Keycloak, Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins and PingFederate
  • After logging in: Apache Kafka Broker, Apache Solr, Elasticsearch, Gerrit and Spinnaker

If one of these sits on the internet at your company, it goes to the top of the patch queue. Calif is holding back its per-app exploit code until people have had time to upgrade.

How worried should you be? The scores disagree

Calif rates the bug 9.2 on the CVSS 4.0 scale. Oracle scores it 8.1 and calls it hard to exploit, while Red Hat’s preliminary number is 9.8 and Amazon’s is 8.8, according to The Hacker News.

The gap exists because exploitation is fiddly. Each target needs its own memory layout and timing, though failed attempts can crash Java and attackers can often retry. Other labs reproduced crashes and memory corruption but not full code execution, and Canada’s Cyber Centre knew of no attacks.

One wrinkle: the triggering input was posted publicly on September 9 as an OpenJDK pull request that was never merged.

The discovery story matters too. Calif’s write-up says researcher Quang Luong pointed Codex at the HotSpot runtime, found the bug in about an hour, and had a working exploit two hours later for under $200 in tokens. That fits what we covered in our look at AI tools that turn security patches into attacks, and it’s why a slow patch cycle is riskier than it used to be.

How to fix Java4Shell today

Patching is the real fix, and the order matters. Start with this checklist, which mirrors the advice from SentinelOne and Calif.

  • Inventory every Java runtime, including containers, bundled installs and appliances, and note the vendor, build and CPU type
  • Update to your vendor’s fixed build, then restart Java services and rebuild container images
  • Where no fix exists yet, add a temporary JVM flag and remove it after patching
  • Check your logs for signs of probing

For the temporary flag, -XX:-UseBASE64Intrinsics turns off only the vulnerable fast path, and Upwind found it stopped the overflow in its tests, per The Hacker News. The alternative, -XX:UseAVX=2, disables AVX-512 entirely but works only on x86-64. On ARM64 the JVM won’t start with it.

When you hunt through logs, look for long Base64 tokens ending in “=” whose length isn’t divisible by 4, especially in Authorization headers. Also watch for bursts of IllegalArgumentException errors and unexplained hs_err crash logs. Clean logs don’t prove you’re safe, so pair this with our five-check list after the FBI’s Artex warning.

Vendors are patching on their own schedules, and this status comes from the The Hacker News roundup dated November 18. Check your vendor’s advisory for anything newer.

Distribution Status
Oracle Java SE and GraalVM Fixed builds released November 17
Amazon Corretto Fixed: 17.0.21.1.2, 21.0.13.1.2, 25.0.5.1.2, 27.0.1.1.4
Azul Zulu (free builds) Fixed: 17.70.201, 21.54.201, 25.38.201, 27.30.201
SapMachine New builds out, but the release notes don’t say what they fix
Red Hat OpenJDK (RHEL 8, 9, 10) No fix yet, per bulletin RHSB-2026-018
Eclipse Temurin, Microsoft Build of OpenJDK, BellSoft Liberica No fixed builds listed when checked

Software that bundles its own Java needs an update from its maker, not from you.

Frequently Asked Questions

What is Java4Shell?

It’s the nickname for CVE-2026-114938, a memory-corruption flaw in OpenJDK’s Base64 decoder that can lead to remote code execution on some AVX-512 servers. Calif disclosed it, and fixes landed on November 17, 2026.

Does Java4Shell affect Java 8 and Java 11?

Standard Java 8 and 11 aren’t listed in the OpenJDK advisory. The exception is the Java SE 8 Performance Pack build 8u511b33-perf, which should move to 8u513-perf.

Is Java4Shell being exploited in the wild?

Not as far as Tenable knew on November 17, and CISA’s known-exploited catalog didn’t list it when The Hacker News checked. That can change, so recheck before you relax.

Does Java4Shell hit ARM servers?

No. The vulnerable code path needs x86-64 chips with AVX-512 VBMI, and ARM64 isn’t affected. Don’t add the -XX:UseAVX=2 flag there, because the JVM won’t start.

How do I check if my CPU is at risk?

On Linux, check /proc/cpuinfo for avx512f, avx512bw, avx512vbmi and bmi2. All four must be present for the vulnerable path to run.

Our take: patch servers first, laptops second

Treat this as a patch-this-week job for anything Java that faces the internet, starting with login systems like Jenkins, Jira, Keycloak and Artifactory, and don’t panic about your laptop. The severity debate matters less than the economics: if one researcher with an AI coding agent can find and weaponize a bug like this in a few hours for under $200, per Calif, a month-long patch cycle is the real risk. Use the JVM flag as a bridge where your vendor is slow, and keep a live inventory of your Java runtimes so the next one takes minutes instead of days.


Comments

Leave a Reply

Your email address will not be published. Required fields are marked *