
A critical vulnerability in the OpenJDK Java runtime could allow unauthenticated attackers to execute arbitrary code on vulnerable systems, following the discovery of a memory-corruption flaw in its built-in Base64 decoder. Tracked as CVE-2026-114938 and named Java4Shell, the vulnerability affects OpenJDK 17.0.6 and later on compatible x86-64 processors.
Security research firm Calif disclosed the vulnerability on November 17, reporting successful remote code execution against several widely deployed Java applications, including Keycloak, Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins and PingFederate. The demonstrations reportedly achieved pre-authentication code execution under the applications’ default configurations, meaning attackers did not need valid user credentials to reach the vulnerable functionality in the tested environments.
The flaw resides in OpenJDK HotSpot’s optimized Base64 decoding implementation, which uses AVX-512 Vector Byte Manipulation Instructions (VBMI) to accelerate processing on supported processors. According to Calif’s technical analysis, malformed Base64 input containing incorrectly positioned padding causes the decoder to calculate an invalid output length. This calculation produces an incorrect write mask, allowing a 64-byte CPU instruction to write beyond the allocated heap buffer.
Although the decoder eventually rejects the malformed input and raises an IllegalArgumentException, the memory corruption occurs before the exception is returned. Depending on the objects occupying the affected memory, the resulting corruption can expose sensitive information, crash the Java process or provide an opportunity for further exploitation. Calif demonstrated how the memory overwrite could be developed into application-specific remote code execution chains.
The vulnerability is hardware-dependent. Exploitation requires an x86-64 processor supporting AVX-512 VBMI, including compatible Intel Xeon processors from the Ice Lake generation and AMD Zen 4 or newer processors. Native ARM64 Java environments and x86-64 systems without the required instruction support are not known to be affected. Applications must also process attacker-controlled data through the vulnerable Base64 decoding path.
Independent analysis by Datadog Security Labs confirmed the out-of-bounds write and demonstrated that it could corrupt neighbouring heap objects and cause the JVM to crash. Its researchers cautioned that reliable remote code execution requires additional exploitation techniques tailored to the targeted application and its JVM environment. There is no universal exploit that works against every Java application.
Oracle released security updates addressing the vulnerability on November 17. The fixed Java releases include 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1. The Red Hat security bulletin also identifies affected Red Hat Enterprise Linux environments and says the company is expediting fixes for its OpenJDK packages.
Calif has published a proof of concept demonstrating the underlying memory corruption but is withholding its application-specific exploitation chains. The company also reported successful exploitation of Apache Kafka Broker, Apache Solr, Elasticsearch, Gerrit and Netflix Spinnaker after authentication, as well as sample applications built with Spring Boot and Apache Struts.
The Canadian Centre for Cyber Security advised organisations to identify affected Java installations, including runtimes bundled with third-party applications and container images, and apply available vendor updates. The agency said it was not aware of active exploitation of CVE-2026-114938 at the time of its November 17 alert. Where immediate patching is not possible, administrators can temporarily disable the vulnerable hardware-accelerated Base64 implementation using the JVM option -XX:-UseBASE64Intrinsics.
Discover more from Aree Blog
Subscribe now to keep reading and get access to the full archive.


