Jenkins EUVD-2026-112177

Affects Jenkins 2.587 and earlier and LTS 2.580.1 and earlier, where the custom XStream converter fails to apply the JEP-200 deserialization filter to classes named by the resolves-to attribute. The attacker must have Overall/Read permission AND at least one of: a non-anonymous user account, or a permission allowing them to POST config.xml (Item/Configure, View/Configure, or Agent/Configure). This authenticated, permission-gated prerequisite is the primary limiting factor - anonymous or read-only-without-account access is insufficient. The only complete fix stated by the vendor is upgrading; administrators may disable the newly added XStream type-permission enforcement via the hudson.util.XStream2.SUPPRESS_TYPE_PERMISSION system property, but that is a compatibility toggle, not a mitigation.

This is a genuinely serious issue that warrants prompt patching, though several decision signals are missing. The Jenkins CNA rates it High severity, and the mechanism - bypassing Jenkins' JEP-200 deserialization filter via the unchecked resolves-to attribute to instantiate arbitrary types - is a classic path to remote code execution in a build/CI controller, a high-value target. However, no CVSS vector was published (advisory lists only 'High'), so the precise metrics are not vendor-confirmed. Exploitation is NOT fully unauthenticated: the attacker must hold Overall/Read plus either a real user account or a */Configure permission that allows POST config.xml, which meaningfully narrows the attacker population to authenticated or semi-trusted users and argues against mass, internet-wide exploitation. No EPSS score, no CISA KEV listing, and no public exploit were provided, and the vuln.today Decision label is 'awaiting_data' - so this is a high-severity, pre-triage item rather than a confirmed actively-exploited emergency. Net: prioritize patching for any Jenkins instance with non-trivial user populations, but treat active-exploitation claims as unverified for now.

An attacker who already holds a low-privilege authenticated Jenkins account with Overall/Read (or the ability to POST a config.xml) submits a crafted XML document whose resolves-to attribute references a dangerous class that Jenkins' deserialization filter would normally block. During deserialization the filter is skipped for that attribute, allowing instantiation of attacker-chosen types and execution of their readResolve behavior, potentially achieving remote code execution on the Jenkins controller. No public exploit was identified at the time of analysis.