Java4Shell Flaw Lets Attackers Run Code on Major Java Apps
A critical OpenJDK bug in Base64 decoding can expose memory or allow remote code execution in tools such as Jenkins and Jira. Oracle has already shipped patches.
A critical bug in OpenJDK, the open-source implementation of Java, can let an attacker run code on well-known enterprise tools without logging in. Researchers call it Java4Shell, and patches are already out.
Fix first, then stopgaps
Oracle shipped fixes on November 17 in its November Critical Security Patch Update, which covers 655 vulnerabilities. The company urges customers to install them as soon as possible.
Teams that cannot update yet have two temporary options. One is to start OpenJDK with the -XX:-UseBASE64Intrinsics option, which turns off the vulnerable Base64 function. The other, on x86-64 systems, is -XX:UseAVX=2, which disables AVX-512 instructions. Calif, the firm behind the finding, stresses that both are stopgaps and that developers should move to the latest runtimes.
Who can be hit
The flaw affects OpenJDK 17.0.6 and later. Distributions built from that code are exposed too: Oracle Java SE, Amazon Corretto, Eclipse Temurin, Microsoft Build of OpenJDK, Azul Zulu and Red Hat OpenJDK.
Calif says it got remote code execution against several products running default settings. Keycloak, Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins and PingFederate fell without any login. Apache Kafka Broker, Apache Solr, Elasticsearch, Gerrit and Spinnaker needed an authenticated account. Sample apps built with Apache Struts and Spring Boot were also exploited.
Scores differ. Calif rates the bug 9.2 under CVSS 4.0, which is critical. Oracle assigned 8.1 (High) under CVSS 3.1.
How a short string breaks the decoder
The bug is tracked as CVE-2026-114938. It is a memory-corruption problem in the hardware-accelerated Base64 decoder. Base64 turns binary data into plain text, and applications use it in authentication, file uploads and network traffic. Java also sits under many unrelated products, so a flaw in a core component can reach all of them at once.
According to Calif, a short, deliberately malformed string can make the decoder write over other parts of Java's memory. That can leak sensitive data or give an attacker code execution. The trigger is short: 64 "A" characters followed by a single equals sign. Older OpenJDK 17 builds need 128.
Limits on the attack
The researchers say there is no universal exploit. An attacker must find an application-specific way to push malicious Base64 data into the decoder. The affected function also runs only on some x86-64 processors with AVX-512 VBMI support.
Calif published a proof-of-concept that shows the memory corruption. It has not released the weaponized exploits used against the enterprise products. The flaw was disclosed on Tuesday.
Found with an AI assistant
Researcher Quang Luong made the discovery while auditing OpenJDK internals with OpenAI's Codex coding assistant. By Calif's account, he spotted the flaw in about an hour and had a first working exploit two hours later. The total AI cost was under $200. The firm adds that AI agents also helped build exploitation chains for individual applications.
Source: BleepingComputer