---
id: "CVE-2026-114052"
kind: "vulnerability"
aliases: ["CVE-2026-114052"]
title: "vLLM: unbounded processor cache lets unauthenticated requests exhaust memory"
severity: "high"
cvss: 8.7
epss: 0.00392
exploited: false
status: "open"
category: "ai-tooling"
products: []
ecosystems: []
tags: ["ai-tooling", "model-serving", "self-hosted", "vllm"]
not_affected_if: ["managed-hosting", "no-moss-audio-models", "vllm-0.31.0-or-later"]
published_at: "2026-10-21T01:17:27+00:00"
first_seen_at: "2026-10-23T21:57:27+00:00"
updated_at: "2026-10-24T13:15:55+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-114052", "first_seen_at": "2026-10-23T21:57:27+00:00"}, {"source": "epss", "url": "https://api.first.org/data/v1/epss?cve=CVE-2026-114052", "first_seen_at": "2026-10-24T13:15:55+00:00"}]
---

# vLLM: unbounded processor cache lets unauthenticated requests exhaust memory

vLLM caches a processor for every distinct request-controlled mm_processor_kwargs value without limit in MOSS-Audio model serving. An unauthenticated attacker can send many distinct or oversized mel_config values to /tokenize, bypassing --api-key, to exhaust host memory and crash the API process.

_Summary, check and fix written by a model from the sources; labelled as such._

## How to check

Check the installed vLLM version (pip show vllm); affected if it is after 0.24.0 and before 0.31.0. Check whether the server serves MOSS-Audio models and whether /tokenize is reachable from untrusted networks.

## Fix

Upgrade vLLM to 0.31.0 or later.
Until upgraded, block external access to /tokenize at the proxy or firewall, since --api-key does not protect it.

## Timeline

- 2026-10-23T21:57:27+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-114052
- 2026-10-24T13:15:55+00:00 epss https://api.first.org/data/v1/epss?cve=CVE-2026-114052

## From the sources

vLLM after 0.24.0 and before 0.31.0 contains a memory exhaustion vulnerability in MOSS-Audio model serving that caches a processor for every distinct request-controlled mm_processor_kwargs value in an unbounded dictionary. Unauthenticated attackers can send distinct or oversized mel_config values to the /tokenize endpoint, which bypasses --api-key, to exhaust host memory and crash the API process.

---

_Free to read at a human pace. An API with keys and an MCP server are in preparation; early access: hello@infrastack.ai_