A lot has already been written about Java4Shell. For those of you who follow our blog, I wanted to bring together what it means for the technologies I work with every day, including Elasticsearch, Kafka and Java Application Servers.

Two nodes can run the same version of Elasticsearch and still have different exposure to Java4Shell. One might use a processor that activates the vulnerable Java optimization, while the other never runs that code. On Kubernetes, a pod moving between those nodes can change its exposure without a single change to the container image.
Java4Shell is a good example of why an application inventory only answers part of a security question. To understand the risk, you need to know which Java runtime the service actually uses, which hardware it runs on, and what data it processes.
What is Java4Shell?
Java4Shell (CVE-2026-114938), disclosed on 17 November 2026, is a memory-corruption vulnerability in the Base64 decoder used by the HotSpot Java Virtual Machine (JVM).
Base64 is a common way to represent binary data as text. Java applications may decode it when processing an authentication header, a sign-in token or a field in a data pipeline. On certain x86-64 processors, the JVM can accelerate that work using a CPU feature called AVX-512 VBMI.
A deployment is exposed when a vulnerable Java build uses that optimization to decode data an attacker can control. The immediate consequence can be a JVM crash, taking the service offline. In some applications, the flaw can also allow an attacker to execute code with the permissions of the Java process.
Questions about your own setup? A dbi services consultant can look at your case.
Talk to an expert (opens in a new tab)Which platforms can be affected?
Elasticsearch and Logstash
Elasticsearch. Elastic identifies 7.17.3 and later as potentially affected when running with a vulnerable JDK on the relevant hardware. This includes the 8.17 branch. An unauthenticated request can crash an affected node. An authenticated user with as little as read access may be able to execute code remotely. Read more about Elastic impact
Logstash. For 8.6.2 and later, the pipeline configuration matters. Elastic confirmed a denial-of-service risk when the elastic_integration filter runs an ingest pipeline that decodes Base64 fields from an untrusted source. In that configuration, someone who can send an event can potentially crash the Logstash process and interrupt ingestion.
Apache Kafka and Confluent Platform
Apache Kafka. Exposure follows the JDK used by each process. A Kafka 3.9 deployment, for example, can include brokers, controllers, Connect workers and Java clients running different Java builds. Updating every broker can still leave the Connect worker that processes incoming records on a vulnerable runtime. Apache identifies the JDK update as the fix. Read more about Apache Kafka impact
Confluent Platform. Confluent lists 7.6.x through 8.4.x for package and archive installations that use an affected Java runtime, and 7.7.x through 8.4.x for published container images that bundle one. Both ranges include 7.9.0. The assessment should cover the Java services in use, including Connect, Schema Registry and Control Center where deployed. Read more about Confluent impact
Tomcat, WildFly, JBoss EAP and Oracle WebLogic Server
For these application servers, the relevant version to check first is the JDK that starts the server. On the Java 17 line, for example, the affected upstream range is 17.0.6 through 17.0.21, with a fix in 17.0.21.1. Check the runtime vendor’s advisory for the corresponding build.
The next question is what reaches that JVM. An application that decodes an externally supplied sign-in token may expose the affected routine if the hardware and runtime conditions also apply. This is why two applications hosted on the same middleware release can have different exposure.
How should you address it?
Start with the runtime used by each running service. Elasticsearch may use a bundled JDK, a WebLogic domain may point to a dedicated Java installation, and a Kafka container carries its own runtime. Updating the host operating system can leave all three running the same Java build as before.
The lasting fix is to deploy a vendor-supported JDK that contains the correction, or a product release or container image that includes it. Prioritize services that accept data from external or untrusted sources. Where an upgrade needs more preparation, vendors document a temporary mitigation that disables the affected Base64 optimization.
After the restart or redeployment, verify the runtime or mitigation on each instance. In a cluster, include replicas and standby nodes. On Kubernetes, check that a replacement pod will inherit the fix the next time it starts.
If you’d like help assessing your exposure to Java4Shell, planning the remediation or verifying the fix across your environments, feel free to reach out. I’d be happy to help.