CVE-2026-100647 — Vllm-project
MEDIUMCVE-2026-100647 is a medium-severity (CVSS 5.3) Improper Input Validation vulnerability. No vendor fix is recorded yet; mitigation options are listed below.
vLLM versions before 0.29.0 contain a denial-of-service vulnerability in the cache_salt parameter accepted on OpenAI-compatible and Anthropic API endpoints, which lacks maximum length…
Exploitation Status
No confirmed exploitation observed yet
- CISA assesses this as automatable — exploitation doesn’t require manual, per-target effort, which raises the odds of mass scanning and opportunistic attacks.
- CISA’s own triage has not observed active exploitation or public proof-of-concept code for this CVE as of its last assessment.
Exploitation and automatability from CISA’s SSVC triage for CVE-2026-100647.
EPSS Exploitation Probability
Probability of exploitation in the next 30 days, from FIRST.org EPSS.
How urgent is this, really
CVE-2026-100647 by exploitation likelihood (EPSS) against impact (CVSS). Outside the shaded patch-first corner.
Where this sits among everything scored
Of 383,485 CVEs with a current EPSS score, this one falls in the < 10% band (highlighted). Counts from FIRST.org, log-scaled.
Description
vLLM versions before 0.29.0 contain a denial-of-service vulnerability in the cache_salt parameter accepted on OpenAI-compatible and Anthropic API endpoints, which lacks maximum length validation and is processed on the single EngineCore scheduler thread. Unauthenticated attackers can send HTTP requests with multi-hundred-megabyte salt values that trigger expensive pickle serialization and SHA-256 hashing, stalling the scheduler thread and denying service to all concurrent requests.
Affected Products
vllmvllmDetection & mitigation playbook
VulnerabilityDetect
Identify every host running the affected component and compare the installed build against the fixed version below — for source-built or distro-packaged software the version string, not a lockfile, is the source of truth (`dpkg -l`, `rpm -q`, or the binary's own `--version`).
Remediation status
No fixed release is recorded for CVE-2026-100647 yet. Track the upstream advisory, and apply the exposure-reduction steps below in the meantime.
Mitigate without a patch
If you can't upgrade right away: gate or disable the affected feature, validate untrusted input at the boundary, and avoid passing attacker-controlled data into the vulnerable path. O3's runtime protection blocks exploitation in production as an interim safeguard until the upgrade lands.
Fixing This On Your OS
If you run this on a Linux distribution, patch through your package manager against the distro's own security advisory below — it tracks the exact backported fix for your release, which can ship on a different timeline (and sometimes a different severity) than the upstream project.
This vulnerability is rated as Moderate severity because, while it can be triggered remotely without authentication to degrade service availability, it causes a transient scheduling delay without leading to host compromise, data modification, or information disclosure. In Red Hat OpenShift AI and Red Hat AI inference…
There is no vLLM env-var toggle that bounds cache_salt on vulnerable builds. Try to strip or deny cache_salt at the gateway or put Auth + rate limits on OpenAI/Anthropic-compatible endpoints.Source: Red Hat security advisory for CVE-2026-100647 (CC BY 4.0)
Frequently Asked Questions
Is CVE-2026-100647 in your dependencies?
Find it across , including transitive dependencies.