Java4Shell: Critical OpenJDK Base64 Decoder Flaw Enables Pre-Auth RCE as Oracle, Red Hat and Its Discoverer Score It Differently
A heap-corruption bug in OpenJDK's hardware-accelerated Base64 decoder, dubbed Java4Shell, enables unauthenticated RCE; Oracle, Red Hat and its discoverer rate its severity differently.
Editor's Note ·
- Correction:
- The article states that Red Hat "labels it 'Important' in its own severity scale." Red Hat's CVE page for CVE-2026-114938 does not use the word "Important" as a severity rating anywhere (the only occurrence of that word on the page is the heading of an unrelated FAQ note about CVSS score variance); the page's own severity field for this CVE records it as "CRITICAL," consistent with Red Hat's four-tier scale, where its 9.8 base score falls in the Critical band.
Overview
A critical vulnerability in OpenJDK’s Base64 decoder, nicknamed “Java4Shell” and tracked as CVE-2026-114938, can let an unauthenticated attacker corrupt heap memory and achieve remote code execution against a wide range of enterprise Java applications, according to BleepingComputer and SecurityWeek. The flaw was disclosed on November 17, 2026, after researcher Quang Luong at the security firm Calif found it using OpenAI’s Codex AI, according to BleepingComputer.
What We Know
The bug is “a memory safety flaw in the way OpenJDK processes Base64-encoded data,” according to SecurityWeek, which reported that invalid Base64 input can cause the decoding routine to write past the end of its output buffer, corrupting nearby heap data. The flaw’s own entry in the National Vulnerability Database classifies it under CWE-122, a heap-based buffer overflow.
The vulnerability affects OpenJDK 17.0.6 and later, along with Java distributions built on it, including Oracle Java SE, Amazon Corretto, Eclipse Temurin, Microsoft Build of OpenJDK, Azul Zulu and Red Hat OpenJDK, according to BleepingComputer and SecurityWeek. A posting on the oss-security mailing list hosted at openwall.com likewise describes OpenJDK 17.0.6 and newer as affected across multiple distributions, including Amazon Corretto, Eclipse Temurin and Oracle’s own builds. Oracle’s CVE record, published through the National Vulnerability Database, additionally lists Oracle GraalVM for JDK and Oracle GraalVM Enterprise Edition as affected products.
Calif demonstrated pre-authentication remote code execution against default configurations of Keycloak, Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins and PingFederate, and authenticated RCE against Apache Kafka Broker, Apache Solr, Elasticsearch, Gerrit and Spinnaker, according to BleepingComputer and SecurityWeek. The same oss-security posting lists the identical pre-authentication targets — Keycloak, Jira, Confluence, Artifactory, Jenkins and PingFederate — according to openwall.com.
The flaw lives in hand-written assembly code that the HotSpot JVM uses to accelerate Base64 decoding on x86-64 processors with AVX-512 VBMI instructions. The affected processor families span Intel Xeon chips from Ice Lake through Granite Rapids and AMD EPYC and Ryzen chips from Zen 4 onward, while ARM64 systems are not affected, according to openwall.com. The trigger input is simple: repeated “A” characters with ”=” padding, the same posting says.
Oracle released fixes for CVE-2026-114938 on November 17 as part of its November 2026 Critical Security Patch Update, according to BleepingComputer and SecurityWeek. Administrators who cannot patch immediately can start the JVM with either the -XX:-UseBASE64Intrinsics flag, which disables the hardware-accelerated decoder, or -XX:UseAVX=2, which disables AVX-512 use on x86-64 systems, according to openwall.com.
SecurityWeek reported researchers cautioning against assuming a single exploitation path: “One thing to note about this vulnerability is that there is no universal exploit. Depending on how Base64 is used and how the application works, different steps may be required.”
Security vendors disagree sharply on how severe the bug is. Oracle’s own CVE record, scored under CVSS 3.1 and published via the National Vulnerability Database, rates it 8.1 out of 10 — “High” severity — with an attack-complexity rating of “High.” Red Hat, which ships its own OpenJDK builds, rates the same flaw 9.8 under CVSS v3, with an attack-complexity rating of “Low,” and labels it “Important” in its own severity scale. Red Hat’s CVE page itself surfaces the gap, stating in an FAQ that vendor CVSS scores can vary depending on the version a vendor ships, how it is shipped, the platform, and how the software is compiled. Calif, the firm that found the bug, arrived at a third figure: 9.2 under the newer CVSS 4.0 standard, according to BleepingComputer and SecurityWeek. None of the three figures is reconciled with the others in the sources reviewed.
What We Don’t Know
None of the sources reviewed for this article report that CVE-2026-114938 has been exploited outside of the researchers’ own proof-of-concept work; the available reporting describes demonstrated exploitation built by Calif, not an observed attack in the wild.
It is also not clear from the sources reviewed which specific downstream JVM builds — such as Eclipse Temurin, Microsoft Build of OpenJDK, and Azul Zulu — have already shipped their own patched releases, as distinct from simply appearing on the list of affected distributions.