New Java Flaw Exposes Jira, Jenkins, Keycloak to Pre-Auth Attacks
A newly disclosed Java vulnerability lets attackers without an account run code on Jira, Confluence, Jenkins, Keycloak, JFrog Artifactory and PingFederate under default settings, according to security research firm Calif.

· Originally published by ontime+ · Last verified: 18 Nov 2026

Key Points
- Security firm Calif disclosed Java4Shell, a Base64 decoder flaw allowing unauthenticated code execution on six widely deployed applications.
- Oracle and OpenJDK shipped fixes this week; Red Hat has released none for its Enterprise Linux and OpenJDK builds.
- The vulnerable code runs only on x86-64 chips supporting AVX-512 VBMI, so identical container images are not uniformly exposed.
The latest:
A newly disclosed Java vulnerability lets attackers without an account run code on Jira, Confluence, Jenkins, Keycloak, JFrog Artifactory and PingFederate under default settings, according to security research firm Calif. Tracked as CVE-2026-114938 and nicknamed Java4Shell, it was patched this week by Oracle and the OpenJDK project. Calif rates it 9.2 out of 10 under CVSS 4.0.
Details:
- The mechanism: The flaw sits in a faster Base64 decoder Java switches to on chips with AVX-512 VBMI, Calif said. A malformed string ending in an extra = causes the decoder to write 62 or 63 bytes past the end of its output buffer. Java rejects the string, but only after the write has already happened.
- Why pre-auth: Servers routinely decode Base64 data — HTTP Basic login headers, SAML artifacts, CSRF tokens — before verifying whether a user is logged in, which is how an attacker with no account reaches the vulnerable code path.
- The escalation step: Because the overflowing bytes are fixed, early readings of the source code pointed only to a crash. Calif says it shapes server memory so the overflow expands a neighboring array, then uses a second array to read memory back — a step it calls JavaBleed — reaching the parts of Java that execute code.
- Affected products: Calif lists unauthenticated attacks against Keycloak, Jira and Confluence, JFrog Artifactory, Jenkins and PingFederate, and authenticated attacks on Apache Kafka Broker, Apache Solr, Elasticsearch, Gerrit and Spinnaker. The firm says its lists are incomplete.
- Competing severity scores: Oracle rates it 8.1 under CVSS 3.1, hard to exploit but requiring no login, and says in its CVE record that successful attacks “can result in takeover of Oracle Java SE.” Red Hat gives a preliminary 9.8; Amazon scores it 8.8.
- Independent reproduction: Upwind reproduced memory corruption on an AWS server with an Intel Ice Lake chip, damaging the adjacent object before Java crashed. Blackpoint Cyber’s Adversary Pursuit Group crashed a test Spring Boot app with roughly 28,300 malformed login requests over about 20 seconds, without logging in. Neither reproduced remote code execution.
- Who is exposed: The code is present in Java 17.0.6 and later on the 17 line, and in Java 21, 25 and 27, plus the Java SE 8 Performance Pack build 8u511b33-perf and three GraalVM builds. Standard Java 8 and 11 are clean, and ARM64 chips do not run the vulnerable code.
- Patch status: Oracle, Amazon Corretto, Azul Zulu and SAP’s SapMachine shipped new builds this week, though SapMachine’s release pages do not specify what they fix. Red Hat has released nothing yet for RHEL 8, 9 and 10 or its OpenJDK 17, 21 and 25 builds. Eclipse Temurin, Microsoft’s OpenJDK build and BellSoft Liberica had not listed fixed builds.
- Interim mitigations: Calif and Upwind point to two JVM flags: -XX:-UseBASE64Intrinsics, which disables only the faster Base64 path, and -XX:UseAVX=2, which disables AVX-512 entirely and must not be used on ARM64. Upwind says both stopped the overflow on every affected version it tested.
- The earlier warning: A malformed input triggering the overflow was posted publicly on September 9 in a proposed OpenJDK fix describing an overflow in the same AVX-512 decoder code. Its author closed it without merging in early November, after an automated notice flagged more than eight weeks of inactivity.
Between the lines:
The gap between Calif’s 9.2 rating and the crashes other labs reproduced reflects a disclosure where attack code is being withheld. Oracle’s own takeover language and Red Hat’s 9.8 preliminary score suggest vendors are treating code execution as credible even without public proof. The September pull request means the overflow sat described in a public repository for roughly two months.
What’s next
Watch for Red Hat’s fixes for RHEL 8, 9 and 10 and its OpenJDK builds, patch listings from Eclipse Temurin, Microsoft and BellSoft, any CISA KEV catalog addition, and whether Calif releases the withheld exploit code.