---
id: "GHSA-5r3c-5x4v-8gjh"
page: "https://infrastack.ai/live/GHSA-5r3c-5x4v-8gjh"
kind: "vulnerability"
aliases: ["GHSA-5r3c-5x4v-8gjh"]
title: "Missing request-wide media budget allows single-request frontend resource exhaustion"
severity: "high"
cvss: 7.5
epss: null
exploited: false
status: "open"
category: "ai-tooling"
products: [{"name": "vllm", "fixed": ["0.33.0"], "affected": ">= 0.6.0, < 0.33.0", "ecosystem": "pypi"}]
ecosystems: ["pypi"]
tags: ["ai-tooling", "cwe-400", "cwe-770", "llm-inference", "multimodal", "self-hosted", "vllm"]
not_affected_if: ["auth-required", "internal-only-network", "text-only-models"]
published_at: "2026-11-02T19:48:58+00:00"
first_seen_at: "2026-11-09T22:53:23+00:00"
updated_at: "2026-11-09T22:54:24+00:00"
sources: [{"source": "ghsa-repo", "url": "https://github.com/vllm-project/vllm/security/advisories/GHSA-5r3c-5x4v-8gjh", "first_seen_at": "2026-11-09T22:53:23+00:00"}]
exploit_sources: []
---

# Missing request-wide media budget allows single-request frontend resource exhaustion

Link for people: https://infrastack.ai/live/GHSA-5r3c-5x4v-8gjh (this .md is for agents)

vLLM downloads and decodes every image in a request without a request-wide size or memory budget, so one compact request with many remote image URLs can consume API-frontend memory, decoder CPU and outbound connections before prompt length is checked. An attacker can degrade or crash the API process for other clients.

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

## Affected

- pypi:vllm affected >= 0.6.0, < 0.33.0 fixed 0.33.0

## How to check

- Check the installed version with `pip show vllm` and confirm whether it is at least `0.33.0`.
- Check whether the server is reachable from untrusted clients and whether API key authentication is enabled, since authentication is optional by default.
- Check whether the served model accepts multiple images per request.

## Fix

- Upgrade vLLM to `0.33.0` or later.

## Timeline

- 2026-11-09T22:53:23+00:00 ghsa-repo https://github.com/vllm-project/vllm/security/advisories/GHSA-5r3c-5x4v-8gjh

## Documents

- [advisory] Maintainer's advisory: github.com https://github.com/vllm-project/vllm/security/advisories/GHSA-5r3c-5x4v-8gjh

## From the sources

### Summary

  A remote inference client can submit one request containing many individually valid remote images. vLLM validates only
  the image count and per-item limits before scheduling every image for download and decoding. It does not enforce a
  request-wide encoded-byte, decoded-byte, or memory budget.

  For models permitting multiple images, one compact request can therefore consume outbound connections, decoder CPU,
  and API-frontend memory before prompt-length admission. API authentication is optional by default; when enabled,
  ordinary inference access is sufficient.

  Affected entry points include /v1/messages, non-Harmony /v1/responses, and equivalent chat-rendering paths.

  ### Details

  The verified flow is:

  1. parse_image() registers a deferred fetch after count validation.
  2. AsyncMultiModalItemTracker.resolve_items() starts all registered media operations in one asyncio.gather() and
     materializes every result.

  3. MediaConnector.load_from_url_async() downloads and decodes each response independently.
  4. The HTTP size limit is per response, not per request.
  5. ImageMediaIO.load_bytes() eagerly decodes each image into a MediaWithBytes, retaining both the decoded raster and
     original bytes.

  6. _resolve_items() retains the complete result list.
  7. Fetching precedes downstream template/token-length admission in the HF renderer and Responses renderer.

  MultiModalConfig defaults an unspecified modality count to 999. Representative LLaVA processing information supplies
  no lower model-specific limit.

  The aiohttp connection-pool limit of 100 and vLLM’s eight-worker decode pool constrain instantaneous concurrency, but
  do not constrain aggregate request memory. Completed images accumulate until gathering and later preprocessing finish.

  A complete fix should enforce a request-wide media resource budget, including aggregate encoded/decoded-byte
  accounting, bounded request/process concurrency, and cancellation/release after validation failure or budget
  exhaustion.

  ### Impact

  A remote client can cause disproportionate frontend memory, CPU, network, and connection-pool consumption with one
  syntactically small request containing many compressed-image URLs. This work occurs before prompt-length rejection, so
  even an over-context request can incur the full media cost.

  Possible consequences include API-process termination, decoder saturation, outbound-connection occupation, and
  degraded availability for other clients.

  The bounded PoC demonstrates the resource-allocation and retention primitive, not an actual out-of-memory crash.

A fix for this was introduced in https://github.com/vllm-project/vllm/pull/61226.

---

_Free to read, with fair use of about 60 pages a minute. For higher limits or API access: sales@infrastack.ai_