

Defending Against Java4Shell: Vulnerability in OpenJDK Enables RCE on AVX-512 CPUs
On November 17, 2026, security researchers at Calif disclosed Java4Shell (CVE-2026-114938, CVSS 4.0: 9.2 Critical), a memory-corruption vulnerability in OpenJDK’s Base64 decoder that allows malformed Base64 input to trigger remote code execution (RCE).
Calif demonstrated pre-authentication remote code execution in default configurations of Apache Kafka Broker, Atlassian Confluence and Jira, JFrog Artifactory, Jenkins, Keycloak, Gerrit, PingFederate, and Spring Boot applications, and post-authentication execution against Elasticsearch.
Oracle released patches addressing this flaw in its November 2026 Critical Security Patch Update. Organizations running OpenJDK 17.0.6 or newer on x86-64 CPUs with AVX-512 VBMI support are affected. Additionally, affected Oracle Java SE versions include 8u511b33-perf, 17.0.21.0.3, 21.0.13.0.3, 25.0.5.0.3, and 27.0.1.0.3.
What Is Java4Shell
The flaw sits in the Java standard library, so exploit traffic is specific to each application. Any Base64 value an application passes to java.util.Base64.Decoder can carry the payload, including HTTP Basic credentials, SAML artifacts, CSRF tokens, and MIME encoded words. In default configurations of the affected applications listed above, attackers can reach this code path prior to authentication.
Recommended Next Steps
Inventory every Java runtime. Record the vendor, full build version, OS, CPU architecture, and owner across your environment—including containers, bundled installations, and vendor appliances. Prioritize internet-facing applications, authentication services, and business-critical systems.
Mitigate immediately by disabling AVX-512 (if unable to patch right away). Start x86-64 JVMs with the flag -XX:UseAVX=2. Note: Apply this flag on x86-64 hosts only; ARM JVMs do not recognize it and will fail to start.
Apply Oracle’s security updates once available. Monitor Oracle and your specific OpenJDK distribution vendor for patched builds addressing CVE-2026-114938, and upgrade affected runtimes as soon as fixes are released. For products that bundle their own Java runtime, follow the vendor’s supported update path.
Hunt and audit logs. Where your WAF, reverse proxy, or application logs capture Base64 values (HTTP Basic credentials, SAML artifacts, CSRF tokens), inspect them after normal application parsing, and flag any that meet all of the following conditions:
Total length L (including padding) is 65 or more
$L \pmod{64}$is 1, 2, or 3Ends with one or two = characters (not exceeding
$L \pmod{64}$)All other bytes belong to the standard Base64 alphabet
SentinelOne’s Defense Against Java4Shell
The exploit arrives as a malformed Base64 string that corrupts memory inside the running JVM, so there is no file to scan at that stage. Instead of relying on CVE-specific signatures, SentinelOne protects against what an attacker does next.
Real-Time Behavioral & Static Protection
Across both Windows and Linux, the SentinelOne Agent’s behavioral engines continuously monitor JVM-hosted services and web servers in real time to detect malicious post-exploitation activity – regardless of the initial exploit vector. If an attacker achieves RCE and acts on the host, the Agent monitors for common post-exploitation behaviors, including a JVM process spawning a shell to run a remote or obfuscated one-liner (such as downloading a payload and piping it to a shell), suspicious child processes, credential access, persistence, and payload execution. Files written to disk during an attack, including JSP web shells and second-stage binaries, are evaluated by the Agent’s static engines across both operating systems.
Linux: In addition to general JVM behavioral monitoring, the Linux Agent applies deeper behavioral checks across web and application servers. These monitor for web shell drops, server-initiated file-transfer utilities, user enumeration, access to sensitive credential stores (such as shadow files and SSH keys), password changes, and use of AWS IMDS tokens– a common indicator of cloud credential theft following exploitation.
Windows: Alongside JVM and post-exploitation behavioral monitoring (such as suspicious child processes, credential access, and persistence), the Windows Agent specifically monitors web server processes for spawning suspicious child-processes as well as unauthorized file drops and execution.
Customers across all OS platforms with Live Security Updates enabled benefit from our most up to date detection capabilities.
Platform Detection Rules
For an added layer of visibility on top of Agent telemetry, SentinelOne provides Platform Detection Rules that can alert on post-exploitation activity from Java application servers and general Java-based exploitation. SentinelOne customers have access to these rules and can enable them today.
Threat Hunting
Our Wayfinder Threat Hunting teams provide an additional 24/7 layer of protection using AI-powered hunting systems that continuously look for post-exploit activity, persistence, and lateral movement across web servers and other potentially vulnerable applications. They add human review and agentic technology to find attacker activity that may originate from this or other application vulnerabilities.



