A critical memory-corruption vulnerability in OpenJDK, designated Java4Shell (CVE-2026-114938), was disclosed on November 17, 2026. This flaw allows remote, unauthenticated attackers to induce heap memory corruption within Java servers by sending a single malformed message. The vulnerability specifically resides within the optimized assembly code that handles AVX-512 Base64 decoding.
The vulnerability stems from an optimization implemented within the Java Virtual Machine (JVM) designed to accelerate Base64 decoding on newer server processors from Intel and AMD. When the decoder processes a malformed string, the vulnerable code writes beyond its allocated buffer. This out-of-bounds write corrupts nearby memory structures before the Java runtime has a chance to reject the input, creating a pathway for exploitation.
Security researchers have demonstrated that exploiting Java4Shell can lead to pre-authentication remote code execution across various enterprise applications. Initial testing confirmed this capability against systems such as Keycloak, Atlassian Jira, Confluence, JFrog Artifactory, Jenkins, and PingFederate, all when running with default settings. Furthermore, researchers identified a related memory leak capability, which they named JavaBleed.
This critical flaw affects several versions of OpenJDK, including OpenJDK 17, 21, 25, and 27. To address this, fixes have been released in OpenJDK versions 17.0.21.1, 21.0.13.1, 25.0.5.1, and 27.0.1.1, and corresponding patches are available from major Java vendors. In response to the discovery, authorities issued a notice regarding this vulnerability, urging organizations to review their software inventories and prioritize internet-facing systems.
The severity of Java4Shell is significant; Oracle assigned it a high-severity score of 8.1 using CVSS v3.1, while Calif rated it with a critical severity score of 9.2 using CVSS v4.0. Beyond causing crashes, the vulnerability allows attackers to overwrite adjacent object metadata in controlled environments, which can lead to code execution. The attack vector is broad, as the decoder can be reached through common application functions, including file uploads and authentication processes.
While Web Application Firewalls can offer some mitigation against emerging exploits, the vulnerability is inherent to OpenJDK itself. Because the flaw is rooted in the core library, many possible code paths utilizing user inputs may lead into the vulnerable code path, making comprehensive protection challenging. The best defense remains upgrading to the latest, patched OpenJDK version.
Sources: