Java4Shell: Critical OpenJDK Flaw Lets Attackers Run Code Without Logging In as Oracle and CISA Urge Immediate Patching

A critical memory-corruption flaw in OpenJDK's Base64 decoder, dubbed Java4Shell and tracked as CVE-2026-114938, can let unauthenticated attackers crash Java applications or run code on widely used enterprise software. Oracle patched it on November 17, 2026, and CISA is urging organisations to update immediately.

Nov 19, 2026 - 18:38
 0  4
Java4Shell: Critical OpenJDK Flaw Lets Attackers Run Code Without Logging In as Oracle and CISA Urge Immediate Patching
A desktop computer monitor in a bright office displaying Java source code in green text on a dark screen.

Java4Shell: Critical OpenJDK Flaw Lets Attackers Run Code Without Logging In as Oracle and CISA Urge Immediate Patching

A serious security flaw in Java itself, not in any single application, has put a large slice of enterprise software on urgent patching duty. The vulnerability, nicknamed Java4Shell and tracked as CVE-2026-114938, sits in the part of OpenJDK that decodes Base64 data — a routine so common that almost every Java web application touches it. Researchers say it can be abused to crash a Java application or, in the worst case, to run an attacker's code without any login at all. Oracle shipped a fix on November 17, 2026, the same day the flaw was disclosed, and the US Cybersecurity and Infrastructure Security Agency (CISA) has told organisations to patch immediately.

What Java4Shell is

Base64 is a standard way of turning binary data into text, and Java programs use it constantly — in login credentials, file uploads, security tokens and network messages. To make that decoding faster, OpenJDK uses hand-written machine code that takes advantage of a feature called AVX-512 VBMI, available on newer x86-64 processors, to process many bytes at once.

The flaw is a memory-corruption bug in that optimised code path. Security firm Calif, which found and disclosed the vulnerability, says specially crafted Base64 input can make the decoder write past the end of its buffer and overwrite neighbouring memory. The trigger is strikingly small: a string of 64 letter "A" characters followed by a single equals sign is enough to set off the corruption. From there, careful exploitation can leak sensitive data out of memory, in a manner reminiscent of the Heartbleed bug, or take control of the Java process entirely.

Calif rates the flaw 9.2 out of 10 under the CVSS 4.0 scoring system, which puts it firmly in the critical range. Oracle, using the older CVSS 3.1 scale and judging the attack difficult to carry out, scores it 8.1, still high. Red Hat has given it a preliminary score of 9.8. The difference is about how hard the attack is to pull off, not about whether it matters: every assessment agrees an attacker needs no credentials.

Which systems are affected

The bug affects OpenJDK 17.0.6 and later, covering the OpenJDK 17, 21, 25 and 27 release lines. That reaches every major Java distribution built from the affected code, including Oracle Java SE, Amazon Corretto, Eclipse Temurin, Microsoft Build of OpenJDK, Azul Zulu, BellSoft Liberica, SAP SapMachine and Red Hat OpenJDK, as well as Oracle GraalVM for JDK. Oracle's advisory also lists its Java 8 Enterprise Performance Pack build as affected.

Three conditions have to line up. The system must run a vulnerable Java version, it must run on an x86-64 processor with AVX-512 VBMI support, and the application must feed attacker-controlled data into the Base64 decoder. Systems running on ARM64 processors are not affected, because they never execute the vulnerable code path, and neither are older x86-64 chips that lack the instruction set. In practice, though, that still leaves a very large number of modern Intel Xeon and AMD servers exposed.

Popular enterprise apps shown at risk

What lifted Java4Shell from a theoretical memory bug to a top-priority patch is what Calif demonstrated against real software. In default configurations, the researchers say they achieved unauthenticated remote code execution against Keycloak, Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins and PingFederate. With a login, they report the same result against Apache Kafka, Apache Solr, Elasticsearch, Gerrit and Spinnaker.

That list matters because these are the workhorses of corporate IT — identity systems, issue trackers, build servers and search platforms that usually sit facing internal networks or, in many cases, the internet. The researchers stress there is no single universal exploit: each application needs its own tailored attack chain, because the malicious Base64 data has to reach the decoder through that application's own inputs, whether HTTP Basic credentials, SAML artifacts, CSRF tokens or encoded file uploads. But the breadth of the demonstration suggests the attack surface is wide, and Cisco's Talos team notes that more targets are likely to be confirmed as testing continues.

Found with an AI coding assistant

Part of why the story has travelled so far, so fast, is how the flaw was found. Calif researcher Quang Luong discovered it in about an hour while auditing OpenJDK's internals with OpenAI's Codex coding assistant, and built a first working exploit in about two more hours. The total AI usage cost was less than 200 dollars, according to the researchers. It is the same team that previously disclosed PHP2Shell, another memory flaw found with AI assistance.

Calif has published a proof of concept showing the memory corruption, but has not released the weaponised exploits it used against the enterprise applications. The firm also says news of the flaw leaked before public disclosure, a claim that gains some support from Talos, which says it expects probing attempts to pick up quickly now that the details are public.

The patch and what to do now

Oracle released fixes for CVE-2026-114938 on November 17 as part of its November Critical Security Patch Update, a mammoth release covering 655 vulnerabilities across its product portfolio, 62 of them rated critical. The Java portion alone contains 11 newly reported flaws, and Oracle says all 11 may be remotely exploitable without authentication.

The fixed Java releases are 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1, with 8u513-perf for the affected Java 8 performance build. Vendors including Amazon and Red Hat have released their own fixed builds on the same day. Organisations that cannot patch immediately have a stopgap: starting the JVM with the -XX:UseAVX=2 flag disables the AVX-512 code path and prevents the out-of-bounds writes, though the flag must only be used on x86 systems, as ARM64 JVMs do not recognise it and will fail to start.

CISA issued its alert on November 17 urging organisations to find affected Java installations and patch them as quickly as possible, starting with internet-facing systems. Europe's CERT-EU published its own warning the following day. As of November 17 and 18, no exploitation in the wild had been confirmed, and the flaw had not been added to CISA's Known Exploited Vulnerabilities list — but security firms warn that window is likely to close fast. Censys reported on November 18 that exploit code was already circulating in public repositories, though unverified, and VulnCheck said a full exploit was available to its customers.

For anyone running Java in production, the message from Oracle, CISA and the researchers is the same: check your Java versions, prioritise the servers the internet can reach, and update now. A flaw in a library this fundamental rarely stays unexploited for long once the details are public.

Sources

What's Your Reaction?

Like Like 0
Dislike Dislike 0
Love Love 0
Funny Funny 0
Angry Angry 0
Sad Sad 0
Wow Wow 0
Nahid Joy I’m Nahid Hasan Joy, a technology writer, web developer, and digital enthusiast with a strong interest in the ever-changing world of technology.