Apache Kylin CVE-2026-108764

Exploitation requires the Apache Kylin job service and its job-driver UI proxy to be directly reachable by the attacker - the advisory states impact is greatest when the job service is 'directly exposed to public network, which is not common,' and that in-house (firewalled/segmented) deployments are 'mostly not impacted.' The core limiting factor is therefore network exposure of the job/UI proxy component; deployments that keep Kylin internal-only sharply reduce or eliminate realistic exploitability. Affected versions are 4.0.0 through 5.0.4; upgrading to 5.0.5 removes the condition. Authentication requirements are not confirmed from available data, as no CVSS vector was published.

No CVSS vector, EPSS score, KEV listing, or POC was provided in any source, so quantitative exploitation signals are entirely absent - this assessment rests on the vendor advisory text and the CWE class (SSRF) alone. The ASF rates this 'moderate', and the advisory explicitly scopes real-world risk: impact is greatest only for Kylin job services 'directly exposed to public network, which is not common,' and typical in-house deployments are 'mostly not impacted.' This is a self-limiting, deployment-dependent SSRF rather than a broadly exploitable flaw. No public exploit identified at time of analysis, and there is no evidence of active exploitation (not in CISA KEV). Treat this as a routine-priority patch for most operators, escalating only where the Kylin job/UI proxy is internet-facing. Key missing data: an authoritative CVSS vector and any EPSS/KEV signal - exploitation probability cannot be numerically estimated from the inputs, so risk is inferred from the vulnerability class and the vendor's own exposure caveat.

An attacker who can reach an internet-exposed Kylin job service sends a request to the job-driver UI proxy containing an attacker-controlled target URL, causing the server to forward the request to internal or otherwise unreachable network resources (e.g., cloud metadata endpoints or internal admin services). Because the proxy issues the request on the server's behalf, the attacker can probe or interact with internal systems the server can reach. No public proof-of-concept was identified at time of analysis.