Skip to content
Critical Java4Shell Flaw in OpenJDK Base64 Decoder Enables Remote Code Execution

Critical Java4Shell Flaw in OpenJDK Base64 Decoder Enables Remote Code Execution

Gbhackers •Divya • November 18, 2026

A critical memory-corruption vulnerability in OpenJDK’s Base64 decoder lets unauthenticated remote attackers overwrite heap memory, leak sensitive data, and execute code remotely.

Tracked as CVE-2026-114938 and nicknamed Java4Shell, the flaw carries a CVSS 4.0 score of 9.2. Researchers at Calif disclosed it on November 17, 2026, after the startup shared proof-of-concept details.

Calif says AI helped find and weaponize the bug. Researcher Quang Luong found it in an hour, built the first working exploit in two more hours, and the total token cost was under $200.

Critical Java4Shell Flaw

The bug is in HotSpot’s x86-64 Base64 intrinsic, a hand-written assembly routine that uses AVX-512 VBMI instructions for faster data decoding. When the input ends with padding characters, this routine miscalculates the output length, resulting in a negative value.

This invalid length leads to a write operation that exceeds the allocated buffer by 64 bytes. For an input of 65 characters consisting of 64 “A” characters followed by “=”, OpenJDK allocates a 50-byte array, causing 62 bytes to overflow beyond the end of the array.

Although Java ultimately rejects the input, the neighboring objects in memory have already been corrupted. In older OpenJDK 17 builds, it takes 128 “A” characters followed by “=” to trigger this issue.

Researchers used heap grooming to turn that fixed overwrite into a controlled primitive. They corrupt an array length to build a “writer,” then widen a “reader” to leak heap contents such as keys, tokens, and personal data.

Calif demonstrated pre-authentication RCE in Keycloak, Atlassian Jira and Confluence, JFrog Artifactory, Jenkins and PingFederate. Post-authentication RCE worked against Kafka, Solr, Elasticsearch, Gerrit and Spinnaker.

Teams already tracking critical Jenkins vulnerabilities that put CI/CD servers at risk of RCE should treat this as another reason to prioritize their build infrastructure. A client-side variant also lets a malicious TLS server compromise a Java program that connects to it.

Exposure is wide because Java runs deep inside enterprise stacks. Affected products include OpenJDK from version 17.0.6 onward and downstream builds such as Amazon Corretto, Eclipse Temurin, Azul Zulu, Red Hat OpenJDK and Oracle Java SE.

Many vendors bundle the runtime in their own products, as seen when Dell’s PowerProtect update patched flaws in bundled Java OpenJDK components . Only x86-64 CPUs with AVX-512 VBMI are vulnerable, including Intel Ice Lake and newer Xeons and AMD Zen 4 or newer. ARM64 systems are not affected.

Calif stresses that no universal exploit exists. Each target needs a custom profile covering the decoder route, heap layout, and feedback path, and the attack is sensitive to network latency.

Some chains also depend on streamed request bodies, so proxies that buffer them, such as nginx by default, disrupt certain profiles. That is not a general fix, and Elasticsearch attacks still work. Calif is withholding product-specific exploit chains until users have had time to patch.

Oracle shipped a fix on November 17 as part of its November Critical Security Patch Update , which contains 655 patches. Its vulnerable Java SE versions include 8u511b33-perf, 17.0.21.0.3, 21.0.13.0.3, 25.0.5.0.3 and 27.0.1.0.3.

Administrators who have followed Oracle’s large quarterly patch releases know how quickly delays create exposure. Where patching must wait, start the JVM with -XX:-UseBASE64Intrinsics to turn off the hardware-accelerated decoder. Alternatively, use -XX:UseAVX=2 on x86-64 hosts only, because ARM builds fail to start with that flag.

Defenders should move fast and inventory every JVM-based service, including appliances and embedded runtimes. Attackers already target Java-based tooling, as shown when Jenkins RCE flaws were exploited in the wild . Calif’s researchers expect Java4Shell to remain a threat for years on systems that never update.

Give your SOC the alert context on the latest threats to stop incidents quickly -> Integrate Threat Intel from 16K SOCs :

Artificial Intelligence

Cyber security Course

Cyber Security Resources

Cybersecurity

Information Gathering

Information Security Risks

New TheMailer AI Phishing Kit Uses ChatGPT and Claude to Bypass Email Spam Filters

Microsoft’s AI Agent Framework Package Compromised on PyPI to Steal Cloud Credentials

Hackers Target Indian Military Personnel With GravityRAT Spyware Hidden in Fake Job Recruitment Apps

OPTIMJACK Attack Injects Malicious JavaScript Into CNN, Zoom and Other Major Websites to Steal Data

Hackers Deploy PerfectPioneer RAT and KUKURAT Backdoor to Steal Sensitive Government Data

McDonald’s iOS App Flooded With Suspicious Push Notifications in Possible Cyberattack