A critical Jenkins vulnerability, tracked as SECURITY-3911 and CVE-2026-70426, could allow malicious Jenkins agents or attackers with Agent/Connect permission to execute code on a Jenkins controller.
The issue affects Jenkins deployments using vulnerable versions of the Remoting library, which handles serialized object communication between controllers and build agents.
The vulnerability carries a Critical CVSS 3.1 score of 9.0, with a network attack vector, no privileges required in some exposure scenarios, scope change, and high impacts to confidentiality, integrity, and availability.
Critical Jenkins Vulnerability
Successful exploitation could give an attacker control over the Jenkins controller, placing build pipelines, stored credentials, source code, and connected deployment environments at risk.
Jenkins relies on its Remoting library, commonly distributed as agent.jar or remoting.jar, to enable communication between the controller and agents.
This communication uses serialized Java objects, a mechanism historically associated with high-impact security risks when untrusted objects are deserialized.
To reduce this risk, Jenkins introduced the JEP-200 class filter. The filter restricts which classes can be deserialized when the controller receives objects across a Remoting channel.
However, researchers found that vulnerable Remoting releases do not consistently apply this protection. In Remoting 3384.v60d89463d9e0 and earlier, except version 3355.3357.v931d3c992987, classes resolved through a fallback deserialization path bypass the JEP-200 filter.
An attacker could exploit this gap by supplying specially crafted serialized objects that trigger the fallback path and load eligible classes from the Jenkins core classpath.
The flaw can be exploited by agent processes, code already executing on Jenkins agents, or users and attackers that possess Agent/Connect permission.
If successfully leveraged, the issue may permit arbitrary code execution on the Jenkins controller rather than remaining confined to the agent environment.
This distinction is significant because Jenkins controllers frequently hold powerful credentials, including source-control tokens, cloud access keys, artifact repository credentials, signing secrets, and deployment permissions.
A compromised controller can also alter build jobs, tamper with release artifacts, inject malicious code into pipelines, or pivot into other corporate systems.
The exposure is limited to classes available on the Jenkins core classpath. This includes classes bundled with Jenkins or supplied by the Java platform, provided they are not included in the older pre-JEP-200 deserialization denylist.
Affected and Fixed Versions
The issue affects Jenkins 2.575 and earlier, as well as Jenkins LTS 2.568.1 and earlier. Jenkins has addressed the vulnerability in Jenkins 2.576 and LTS 2.568.2.
The fixed releases update Remoting to ensure that the JEP-200 class filter is also enforced when classes are resolved via the fallback code path.
Organizations operating Jenkins controllers should prioritize upgrading to a patched release, particularly where build agents are shared, dynamically provisioned, externally managed, or capable of running untrusted workloads.
Mitigation
Administrators who are unable to upgrade immediately can use the vendor-provided workaround available in the Jenkins SECURITY-3911-3930 GitHub repository.
Teams should also review Agent/Connect permissions, restrict agent connectivity to trusted systems, isolate build agents, and audit controller logs for unexpected Remoting activity.
Detect, investigate, and respond faster with in-browser data inspection from ANY.RUN. Gain complete phishing visibility to strengthen your SOC and reduce MTTR