Skip to main content
CybersecurityVulnerability Management

Java Flaw Exposes Major Apps to Remote Code Execution Attacks

Laptop screen with blurred code on a cluttered office desk with daylight in the background.

CVE-2026-114938 is a memory‑corruption flaw in OpenJDK’s hardware‑accelerated Base64 decoder that can allow unauthenticated remote code execution (RCE) against major Java applications, researchers at Calif disclosed on Tuesday.

CVE-2026-114938: the Base64 decoder bug and how it breaks Java memory

Calif says the flaw — dubbed Java4Shell — is a memory‑corruption bug in Java’s hardware‑accelerated Base64 decoding functionality. Specially crafted Base64 input can cause the decoder to overwrite portions of Java process memory, enabling attackers to either disclose sensitive information from memory or achieve RCE. The researchers demonstrated that the memory corruption can be triggered with just 64 "A" characters followed by a single equals sign (=):

AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=

Older OpenJDK 17 versions, Calif reports, require 128 "A" characters followed by the equals sign. The vulnerable code path is only used when Java’s hardware‑accelerated decoder runs on certain x86‑64 processors that support AVX‑512 instructions (specifically AVX‑512 VBMI capabilities), so exploitation depends on both software and supported CPU features.

Verified targets: Keycloak, Jenkins, Atlassian products and others

Calif says it successfully exploited Java4Shell to perform remote code execution against a number of widely used Java applications running with default configurations. The company separates confirmed cases into unauthenticated and authenticated RCE:

  • Unauthenticated RCE: Keycloak, Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins, and PingFederate.
  • Authenticated RCE: Apache Kafka Broker, Apache Solr, Elasticsearch, Gerrit, and Spinnaker.

Researchers also reported successful exploitation of sample applications built with the Apache Struts and Spring Boot frameworks. Calif notes that Java4Shell does not provide a single, universal exploit; attackers must find an application‑specific way to deliver the malicious Base64 payload to the vulnerable decoder.

Affected Java builds and distributions

Calif reports that Java4Shell affects OpenJDK 17.0.6 and later and any Java distributions built from the affected OpenJDK code. The source lists impacted distributions by name: Oracle Java SE, Amazon Corretto, Eclipse Temurin, Microsoft Build of OpenJDK, Azul Zulu, and Red Hat OpenJDK. Because many applications share the same Java runtime, a flaw in OpenJDK’s internals can ripple across different, unrelated products.

Discovery and exploit development: AI‑assisted research

Calif says researcher Quang Luong discovered the flaw using OpenAI’s Codex AI coding assistant while auditing OpenJDK internals. According to Calif, Luong identified the bug in roughly one hour and built an initial working exploit in another two hours, at a combined AI usage cost of less than $200. Calif also says AI agents helped develop additional application‑specific exploitation chains. The company published a proof‑of‑concept that demonstrates the memory corruption but has not released the weaponized exploits used against the affected enterprise applications.

Patches, temporary mitigations, and vendor guidance

Oracle released patches for CVE-2026-114938 on November 17 as part of its November Critical Security Patch Update, which addresses 655 security vulnerabilities across Oracle’s product portfolio. Oracle recommends customers apply the available security patches as soon as possible. Calif rates the flaw as critical with a CVSS 4.0 score of 9.2; Oracle assigned it a lower severity of 8.1 (High) using CVSS 3.1.

For environments that cannot immediately update Java, Calif recommends launching OpenJDK with the JVM option -XX:-UseBASE64Intrinsics to disable the vulnerable hardware‑accelerated Base64 functionality. Administrators running Java on x86‑64 systems can alternatively use -XX:UseAVX=2 to disable AVX‑512 instructions. Calif emphasizes these steps as temporary mitigations and advises that developers update Java runtimes to patched versions.

What this means for security teams, enterprises, and adversaries

  • Technologists and security teams: prioritize patching Java runtimes listed as affected (OpenJDK 17.0.6+ and named distributions) or apply the temporary JVM flags Calif recommends; inventory efforts must identify both Java versions and the underlying CPU capabilities (AVX‑512 VBMI).
  • Affected enterprises and procurement leaders: expect that a single OpenJDK flaw can affect multiple vendor products — vendors named in Calif’s tests (Atlassian, JFrog, Jenkins, Keycloak, Kafka, Solr, Elasticsearch, Gerrit, Spinnaker) should be treated as potential exposure points until their specific deployments are patched and validated.
  • Adversaries and threat actors: Calif’s tests show that exploitation requires an application‑dependent delivery vector plus eligible AVX‑512 hardware — meaning there is no one‑size‑fits‑all exploit, but successful chains do exist for common enterprise products and default configurations.

Oracle’s patches are available and Calif’s mitigations provide stopgap options, but the combination of a core runtime flaw, proof‑of‑concept code, and AI‑accelerated exploit development underlines the urgency for administrators to inventory Java versions and CPU capabilities, apply updates, or enable the recommended JVM mitigations without delay.

Original story