Innovation & Tech

Calif Reveals Java4Shell; OpenJDK and Oracle Ship Fixes

Calif found Java4Shell with AI assistance, rated it 9.2 under CVSS 4.0, and disclosed target-specific remote code execution as OpenJDK and Oracle shipped fixes.

By Amara Osei

5 min read

Calif Uses AI to Find Java4Shell as OpenJDK and Oracle Ship Fixes
Calif Uses AI to Find Java4Shell as OpenJDK and Oracle Ship FixesStockSnap / pixabay

What's News

  • Calif disclosed Java4Shell on November 17, 2026; OpenJDK and Oracle shipped fixes the same day.
  • CVE-2026-114938 scores 9.2 under CVSS 4.0 and 8.1 under CVSS 3.1.
  • AI-assisted discovery took one hour; a working exploit took two more, with token costs below $200.
  • The trigger can write 62 bytes beyond a 50-byte allocation in affected OpenJDK builds.
  • Java4Shell affects OpenJDK 17.0.6 and later on x86-64 systems with AVX-512 VBMI; ARM64 is unaffected.

By Amara Osei

Calif disclosed Java4Shell on November 17, 2026, rating the remote code-execution flaw 9.2 under CVSS 4.0; OpenJDK and Oracle released fixes that day. The vulnerability, tracked as CVE-2026-114938, can let an unauthenticated attacker corrupt memory and execute code remotely, according to Calif.

How did AI-assisted research find the bug?

Calif researcher Quang Luong directed Codex to examine OpenJDK's JIT compiler, intrinsics, widely accessible APIs and cryptography. Calif said the model surfaced the Base64 finding in one hour; Luong produced a working exploit in another two hours, with total token costs below $200.

Luong chose the targets and supplied exploitation expertise. Calif built pipelines plus a heap tracer and visualizer that helped AI agents develop later exploit chains.

The flaw sits in hand-written x86-64 assembly that OpenJDK HotSpot uses to accelerate Base64 decoding with AVX-512 VBMI instructions. Calif said the routine's location undercuts a familiar Java safeguard: managed memory and memory-access checks do not eliminate corruption risk in the assembly code.

What does the exploit chain demonstrate?

Calif's trigger uses a 65-character string with 64 “A” characters and one “=” to force a 50-byte output allocation. An invalid negative output length enables a 64-byte write, including 62 bytes beyond the allocation; Java rejects the malformed input only after nearby objects may have been corrupted.

For OpenJDK 17, Calif also published a 128-“A”-plus-“=” trigger because that version reaches the affected path at a higher threshold. Calif's chain corrupts a buffer length to upgrade the fixed overwrite into chosen-value writes, while another corrupted array can disclose heap contents, including private keys and tokens.

Calif demonstrated pre-authentication RCE in default configurations of Keycloak, Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins and PingFederate. Calif also demonstrated post-authentication RCE in Apache Kafka Broker, Apache Solr, Elasticsearch, Gerrit and Netflix Spinnaker, plus sample Apache Struts and Spring Boot applications. Calif said the published product lists are non-exhaustive.

Researcher samczsun separately shared a client-side exploit with Calif. It showed that a Java program opening a TLS connection to a malicious server could be compromised even without exposing a service of its own.

Which runtimes and processors are exposed?

Calif says Java4Shell affects OpenJDK 17.0.6 and later, along with distributions built from OpenJDK: Amazon Corretto, Eclipse Temurin, Microsoft Build of OpenJDK, Azul Zulu, Red Hat OpenJDK, Oracle Java SE and Oracle GraalVM for JDK.

Calif says exposure requires an x86-64 processor with AVX-512 VBMI support. Calif says ARM64 processors are not affected. Calif lists Intel Xeon Ice Lake, Sapphire Rapids, Emerald Rapids and Granite Rapids, plus AMD EPYC and Ryzen chips based on Zen 4 or newer, as qualifying systems.

The OpenJDK advisory marks Java 17, 21, 25 and 27 as affected and categorizes the issue under hotspot/compiler. Oracle assesses the flaw as remotely exploitable over a network without authentication, with high attack complexity and high impacts on confidentiality, integrity and availability.

Calif's 9.2 critical score uses CVSS 4.0, while the OpenJDK advisory and Oracle's November update assign 8.1 under CVSS 3.1. The difference follows the scoring-system version, not a change in the described flaw.

Why isn't there one universal payload?

Calif says “each target needs a bespoke profile” with a decoder route, a live writer, a reader or another feedback path, a suitable heap layout and an application action. Calif's requirement for target-specific construction means the published work does not include one universal payload.

Calif says remote heap grooming remains sensitive to round-trip time and jitter, while misplaced objects can crash the JVM. Calif said retries often work after a crash.

Calif says buffering complete request bodies can defeat some grooming profiles, but not all. The Elasticsearch chain works from one complete request, while configured WebSocket tunnels change the buffering conditions.

What fixes and mitigations are available?

OpenJDK's November 17, 2026 advisory recommends upgrading as soon as possible. Oracle's JDK 17.0.21.1 release notes carry the same date and link to a November advisory containing 11 new Java SE patches; they credit Quang Luong of Calif.io and Albert Yang for CVE-2026-114938.

OpenJDK's mainline change adds a four-line guard. It sends a padded Base64 tail shorter than one complete block to the Java handler instead of the affected optimized routine.

Until updates are installed, Calif lists -XX:-UseBASE64Intrinsics to disable hardware-accelerated Base64 encoding and decoding. Calif also lists -XX:UseAVX=2 to disable AVX-512 on x86-64 hosts; OpenJDK on ARM does not recognize that option and will fail to start.

Calif published the vulnerability mechanics and a bounded standalone trigger but withheld application-specific exploit chains while users upgrade. Calif's publication provides the underlying flaw and general exploitation technique rather than ready-made attacks against individual products.

What does Java4Shell mean for AI-assisted security?

Calif's broader argument is that AI can help maintainers identify and remove unnecessary features and optimizations. Calif describes the Base64 optimization as “intended for applications decoding enormous amounts of data” yet shipped to every affected runtime, exposing applications that gained little from it.

Calif says the same AI-assisted workflow can help reduce attack surface, not only discover and weaponize flaws. The practical test for maintainers now is whether AI-assisted review can remove low-value optimizations before they create another assembly-level exposure. Enterprises still face the immediate task of applying the November 17 fixes across affected deployments.

Source: unite.ai

Share this article:

More from Amara Osei

Amara Osei

Show full bio

Senior reporter covering companies, deals and strategy in entrepreneurship, startups and venture funding at Business Bearings.

2,248 articles

Related articles

« Previous articleNext article »