Skip to content
THREATWIRE

CVE Tracker

CVE-2026-105759

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the Rust frontend's track…

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the Rust frontend's track_http_metrics middleware records the raw HTTP method token as a Prometheus label for requests reaching registered routes. An unauthenticated attacker can send unique arbitrary method tokens to unguarded routes such as /tokenize, causing Prometheus's Family::get_or_create function to permanently create counter and histogram label sets. Those label sets increase process memory usage and enlarge the /metrics response until the service or monitoring path is exhausted. This issue is fixed in version 0.30.0.

Vendor
vllm
Product
vllm
CVSS 3.1
7.5 (high), scored by NVD (primary)CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
CVSS 3.1
5.9 (medium), scored by security-advisories@github.comCVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H
Published
2026-10-05
NVD modified
2026-10-16
NVD status
Analyzed

Not in the CISA KEV catalogue at the last update.

EPSS

0.36% chance of exploitation in the next 30 days (percentile 28th)

FIRST EPSS, updated 17 Oct 2026. About EPSS

CVSS is severity, EPSS is likelihood, KEV means already exploited. EPSS is a statistical probability, not a certainty: a low score does not make a bug safe, and a high one does not prove it is being exploited.

References

Data last updated 18 Oct 2026. Sources: NVD API 2.0, FIRST EPSS. Links: NVD, CISA KEV. Metadata only. ThreatWire does not host exploit code.