×

Book a Demo

*First Name

*Last Name

*Work Email

*Company

Tell Us How We Can Be Successful Together

Submit →

Thank you. The form was submitted successfully. You can now close this modal.

BLOG

Java4Shell: The Java Flaw that Strikes Before Login

A Java Base64 bug lets attackers run code on Jira, Confluence, Jenkins and Keycloak without logging in. Here’s how to check exposure, patch, and virtual patch.

‍

November 18, 2026

The Threat

Java4Shell is a newly disclosed flaw that can let an attacker run code on Jira, Confluence, Jenkins, Keycloak and other enterprise applications without logging in. There are fixes from Oracle, Amazon, and Azul, with more Java distributions to follow. However, if your organization runs Java, and nearly every large organization does, this one goes to the top of the queue.

CVE-2026-114938

What Java4Shell Is

Java constantly decodes Base64, a standard way of packaging binary data so it can travel across the web. To do it faster, OpenJDK uses an optimized decoder built on specialized x86-64 CPU instructions that process many bytes at once. Researchers at the security firm Calif used AI to find and test exploits to discover a memory-corruption bug in that fast path.

A malformed string ending in an extra “=” makes the decoder write 62 or 63 bytes past the end of its output buffer. Java does reject the string, but only after the write, but by then the memory is already corrupted.

Why It Hits Before Login

Servers often decode Base64 data before they check who is asking. HTTP Basic authentication headers, SAML artifacts and CSRF tokens are common examples, and that ordering turns a memory bug into a pre-authentication attack. An attacker needs no stolen password and no account, only network access to the login page.

Warning

Calif’s attacks worked without a login against default configurations of Keycloak, Jira, Confluence, JFrog Artifactory, Jenkins and PingFederate, and with a login against Kafka, Solr, Elasticsearch, Gerrit and Spinnaker. The researchers say the list is incomplete.

Exposure

Who Is Exposed

Two conditions must both be true: the vulnerable code has to be in your Java build, and the server’s CPU has to support the instructions that run it.

ExposedNot exposed
Java 17.0.6 and later on the 17 lineJava 17 releases before 17.0.6
Java 21, 25 and 27Standard Java 8 and 11 (one Java 8 Performance Pack build is affected)
x86-64 CPUs with AVX-512 VBMI: Intel Xeon Ice Lake and newer, AMD Zen 4 and newerARM64 servers, and x86-64 CPUs without AVX-512 VBMI
Both conditions trueEither condition false

The hardware condition is the one teams seem to miss. The same container image can run the vulnerable code on one server and not on the server next to it, depending on the CPU underneath. An SBOM alone can’t tell you whether you’re exposed, you also need to know which CPUs your workloads land on.

Fix availability depends on your Java distribution. Status at the time of writing:

Java distributionFix statusAction now
Oracle Java SEFixed in 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1Patch and restart
Amazon Corretto, Azul ZuluFixed builds availablePatch and restart
Red Hat (RHEL 8, 9, 10 and its OpenJDK builds)Fix not yet releasedVirtual patch until it ships
Eclipse Temurin, Microsoft Build of OpenJDK, BellSoft LibericaNo fixed builds listedVirtual patch until they ship

Risk Assessment

How Worried to Be

Worried enough to act today, but not worried enough to panic.

9.8

Red Hat’s preliminary CVSS score. Oracle rates the flaw 8.1 under CVSS 3.1.

20 sec

Time Blackpoint Cyber needed to crash a test Spring Boot app with about 28,300 malformed requests, no login required.

Sept 9

Date an input that triggers the overflow went public in a proposed OpenJDK fix, giving attackers a possible head start.

Several factors limit the immediate risk. No single exploit works everywhere; each target requires its own memory layout and timing. Other labs have reproduced crashes and a memory leak but not code execution, and Canada’s Cyber Centre said it was not aware of any attacks. The numbers above are why that is no reason to wait: even a failed attempt causes an outage, and a working trigger has been public since September.

The Bigger Shift

A New Pace of Discovery

Java4Shell arrived roughly a month after Calif and Google’s Wiz division independently used AI to find two critical PHP flaws. Malwarebytes’ Chris Fistonich told Cybersecurity Dive where this is heading:

“…very likely to be the new normal.”

Chris Fistonich - Malwarebytes

Fistonich also noted that monthly patch cycles weren’t designed for critical flaws found this quickly. When AI shrinks discovery from months to days, the gap between disclosure and exploitation shrinks with it. Teams that can deploy a compensating control within hours, then patch on a verified schedule, absorb events like this one. Teams that wait for the monthly window carry the risk until it opens.

The Response

Step One

Patch First

CISA has urged organizations to apply the patch immediately. Where a fixed build exists for your Java distribution, deploy it now. There are three steps teams commonly miss:

Restart Java services after updating

A running process keeps the old code until it restarts.

Rebuild and redeploy container images

Updating the base image does nothing for containers already running.

Chase your vendors

Products that bundle their own Java runtime need an update from the product’s maker. Ask each vendor when theirs will ship.

Step Two

Virtual Patch What You Can’t Fix Yet

The truth is patching won’t cover everything this week. Red Hat and several other distributions haven’t shipped fixed builds, and vendor-bundled runtimes will lag further.

Virtual patching closes that gap. Instead of fixing the vulnerable code, you block the path an attacker would use to reach it, at runtime or at the edge. For Java4Shell, apply it in three layers.

At the JVM

Start Java with one of two flags. The first turns off only the optimized Base64 code. The second turns off AVX-512 entirely, so use it only on x86-64: Java on ARM64 won’t start with it. In Upwind’s lab, both flags stopped the overflow on every affected version tested.

// Option 1: turn off only the optimized Base64 path
java -XX:-UseBASE64Intrinsics -jar app.jar

// Option 2: turn off AVX-512 entirely (x86-64 only)
java -XX:UseAVX=2 -jar app.jar

// Remove the flag once the fixed build is in place

At the edge

Inspect Base64 in authentication headers and tokens. Flag or drop tokens of 32 characters or more that end in “=” and whose length is not a multiple of 4. But don’t rely on your reverse proxy alone: a proxy that waits for complete requests, as nginx does by default, blocks some attacks but not all.

In monitoring

Watch for bursts of IllegalArgumentException errors from java.util.Base64$Decoder, unexplained hs_err_pid crash logs and restart bursts. These don’t prove a compromise, and their absence doesn’t prove you’re safe.

Virtual patching doesn’t replace the real fix, but it buys you time when every second counts.

FAQ

Frequently Asked Questions

What is Java4Shell?

+

Java4Shell (CVE-2026-114938) is a memory-corruption vulnerability in the optimized Base64 decoder OpenJDK uses on newer Intel and AMD processors. A malformed string makes Java write past the end of its output buffer. Researchers at Calif found it with help from AI and showed it can let an attacker run code on common enterprise applications without logging in.

Which Java versions are affected by Java4Shell?

+

Java 17.0.6 and later on the 17 line, plus Java 21, 25 and 27, when running on x86-64 processors with AVX-512 VBMI (Intel Xeon Ice Lake and newer, AMD Zen 4 and newer). Standard Java 8 and 11 are not affected, apart from one Java 8 Performance Pack build. ARM64 servers are not affected. Because hardware matters, check both the Java version and the CPU each workload runs on.

Which applications are vulnerable to Java4Shell?

+

Calif showed pre-authentication attacks against default configurations of Keycloak, Jira, Confluence, JFrog Artifactory, Jenkins and PingFederate, and authenticated attacks against Kafka, Solr, Elasticsearch, Gerrit and Spinnaker. The researchers say the list is incomplete. Any Java application that decodes Base64 data, such as login headers or security tokens, before authenticating the user could be exposed.

Can Java4Shell be exploited without logging in?

+

Yes. Many applications decode HTTP Basic headers, SAML artifacts and CSRF tokens before they authenticate the user, so an attacker only needs to reach the login page. Even a failed attempt can cause an outage: Blackpoint Cyber crashed a test application in about 20 seconds without credentials.

Is Java4Shell being exploited in the wild?

+

No attacks have been reported, and other labs have reproduced crashes and a memory leak but not code execution. However, an input that triggers the overflow was public on the OpenJDK code site from September 9, so attackers may have had a head start. Oracle rates the flaw 8.1 under CVSS 3.1, and Red Hat gives a preliminary 9.8.

How do I fix Java4Shell?

+

Patch now. CISA has urged immediate patching, and fixed builds are available for Oracle Java SE, Amazon Corretto and Azul Zulu. After updating, restart Java services, rebuild and redeploy running containers, and ask vendors whose products bundle their own Java runtime when their update will ship.

Questions about Java4Shell

Talk to the Mimic team.

A man and woman working at a desk

See why the world's most targeted organizations trust Mimic to protect what matters most.