Vulnerabilities and exploitation

Java4Shell: attackers begin probing OpenJDK flaw as vendor advisories pile up

Researchers report scanning and blind exploitation attempts against CVE-2026-114938, which affects OpenJDK 17.0.6 and later. Cisco, Red Hat, Elastic, Siemens, Atlassian and Jenkins have all published guidance.

By Junglewise News DeskPublished 6 min read
A monument sign reading "Cisco" with the company's logo, set on a lawn with trees in the background.
Prayitno, CC BY 2.0 via Wikimedia Commons

A memory corruption flaw in the way OpenJDK handles Base64 data is now drawing live scanning and blind exploitation attempts across the internet, days after patches appeared. The bug, tracked as CVE-2026-114938 and nicknamed Java4Shell, affects OpenJDK 17.0.6 and newer, according to Cisco's advisory. Because the vulnerable code path sits in a core part of the Java runtime, the exposure follows wherever Java goes: enterprise web applications, network management appliances, identity servers and build systems.

The researchers who found the flaw demonstrated unauthenticated remote code execution against default configurations of Jira, Confluence, Jenkins, Keycloak and other enterprise applications, though SecurityWeek reported that each target needs its own tailored attack chain, so no single exploit works everywhere. There is one meaningful limit: exploitation requires an x86-64 processor supporting AVX-512 VBMI, an instruction set extension found in newer Intel and AMD chips. ARM-based systems are not affected.

What to do now

  1. Patch the Java runtime first, not just the applications on top of it. Oracle and the OpenJDK project released fixes on 17 November 2026. Amazon, Microsoft and Azul have published patched builds of their OpenJDK distributions. Red Hat rates the issue Critical and is fast-tracking patches for Red Hat Enterprise Linux 8, 9 and 10 and for its own OpenJDK 17, 21 and 25 builds.
  2. Work out which of your hosts can actually be attacked. Check whether the CPUs running your Java workloads report AVX-512 VBMI support. If they do, assume the systems are exploitable in their default state: Red Hat said "No special configuration or non-default features are required. The accelerated code path is selected automatically whenever the JVM runs on a CPU reporting AVX-512 VBMI support".
  3. If you run Atlassian Data Center products, disable AVX-512 support as a workaround for Bamboo, Bitbucket, Confluence, Crowd, Jira and Jira Service Management until patches arrive. Atlassian says cloud customers need take no action while it addresses the issue in its cloud products.
  4. For Jenkins, update to the November Java releases and restart the controller. The Jenkins project says controllers are only at risk when running a vulnerable Java 17, 21 or 25 release on an AVX-512 VBMI-capable CPU, and notes workarounds are available.
  5. If you use Keycloak container images 26.8.2-1 or 26.8.2-2, replace them. Those images bundled a vulnerable OpenJDK 21 build, and the project published updated images with a patched Java runtime on 19 November 2026, including release 26.8.2-3.
  6. For Cisco products, there is nothing to install yet. Cisco has not released fixes and states plainly that no workarounds address the vulnerability, so monitor the advisory and restrict network access to affected management appliances in the meantime.
  7. If you run Siemens Industrial Edge Management, note that Siemens recommends taking affected instances offline while it prepares patches.
  8. Watch logs on internet-facing Java applications for oddly shaped Base64 input and for requests carrying a java4shell-recon User-Agent string, though scanners have already begun disguising themselves.

What the scanning looks like

GreyNoise had logged more than 150 requests matching the shape of a Java4Shell attack from five IP addresses as of early on 19 November 2026, with no successful intrusions observed on its sensors, SecurityWeek reported.

WatchTowr saw similar traffic. "We are already seeing attackers spray the internet and hit our global honeypot network with requests decorated in 'java4shell-recon' User-Agents", said Jake Knott, head of threat intelligence at the firm, alongside blind exploitation attempts using correctly shaped requests. Knott added: "While what we're seeing so far isn't particularly functional or scary yet, it shows the clear signal that attackers want to leverage this vulnerability." Knott said WatchTowr expects attackers to refine target-specific exploits over the coming days.

Cloudflare's Cloudforce One reported that scanners have started concealing their tooling. One IP address probed with a user agent referencing Java4Shell, then came back roughly 17 hours later carrying the exploit trigger while impersonating Chrome. On 18 November 2026, Cloudforce One also saw a single IP used for both Java4Shell activity and Artex, an open source artificial intelligence penetration testing framework that threat actors have been abusing.

Threat intelligence firm Previdian logged nine exploitation attempts from six IP addresses on 18 and 19 November 2026, but cautioned that its sensor data cannot confirm whether any target was compromised. None of the reported activity amounts to a confirmed intrusion. Cisco's product security team said it is aware of proof-of-concept exploit code but has no knowledge of malicious use, and the Canadian Centre for Cyber Security said on 17 November 2026 that it was not aware of active exploitation while warning the picture could change quickly.

Technical details

CVE-2026-114938 is a memory corruption bug in Base64 handling in OpenJDK 17.0.6 and later. Exploitation requires an x86-64 CPU with AVX-512 VBMI support. The flaw was disclosed on 17 November 2026, the same day Oracle and the OpenJDK project shipped patches, and Cisco credited the researchers who reported it. Cisco's advisory, rated High severity, was first published 18 November 2026 at 16:00 GMT and last updated 19 November 2026 at 20:22 GMT. It references Oracle's Critical Patch Update advisory for November 2026, the OpenJDK vulnerability advisory of 17 November 2026 and a Talos threat advisory.

Cisco has confirmed more than a dozen affected products. The list runs from management platforms such as Catalyst Center, Intersight Virtual Appliance and Smart Software Manager On-Prem, through security tooling including Secure Network Analytics and Secure Workload, to the Splunk AppDynamics On-Premises Virtual Appliance. Also named are Smart PHY, Connected Mobile Experiences, Cisco IQ Virtual Appliance, Crosswork Assurance User Experience, Crosswork Workflow Manager, Native Broadband Network Gateway, Optical Network Planner, Services as Code Common Automation Framework, Splunk AppDynamics On-Prem Enterprise Console, NextFusion Platform, Ultra Cloud Core products, Finesse, Unified Contact Center Enterprise, Unified Customer Voice Portal, Unified Intelligence Center and Virtualized Voice Browser. IOS XE Software and UCS Central Software are listed as not affected, although SecurityWeek reported that Cisco was still investigating both.

Elastic says every Elasticsearch release from 7.17.3 onward is affected. Remote code execution there requires valid credentials with read permissions or higher, but a denial-of-service condition can be triggered without authentication. Logstash 8.6.2 and later are affected only under a specific configuration and only to the extent of a crash, with fixes planned for upcoming releases. On its own infrastructure, Elastic has pushed a hotfix to the Serverless offering and altered the configuration of Elastic Cloud Hosted, noting that only a small share of hosted deployments run on vulnerable CPUs.

Red Hat rates the flaw Critical across RHEL 8, 9 and 10 and its OpenJDK 17, 21 and 25 builds. Siemens says three Industrial Edge Management products are affected and recommends taking them offline pending patches.

The public proof of concept only triggers the memory corruption. VulnCheck has made a full remote code execution exploit available to its customers, and Kevin Breen of Immersive reported developing a working exploit. How quickly a reliable public exploit emerges remains the open question.

Share this article

On Junglewise Threat Intelligence

Related stories

Scan your domain free: see what attackers can see from outside, in about 30 seconds.

This article as Markdown. Atom feed.