# CVE-2026-114050: vLLM multimodal UUID cache poisoning remote engine crash

Source: https://junglewise.ai/threats/cve-2026-114050-vllm-before-0-31-0-uses-a-client-supplied-multimodal-uuid  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-10-21. Updated: 2026-10-21  
License: CC BY 4.0 with a link back

CVE-2026-114050 · Severity: high · CVSS 7.5 · Published 2026-10-21

vLLM is an inference serving engine for large language models with multimodal support. An unauthenticated attacker can submit specially crafted requests with forged image UUIDs that desynchronize vLLM's internal processor caches, causing a fatal crash that terminates the entire serving instance. The attacker needs only network access and does not require authentication.

Technologies: [vllm (PyPI)](https://junglewise.ai/threats/technologies/pypi-vllm), [vLLM Project vllm](https://junglewise.ai/threats/technologies/vllm-project-vllm). Vendors: [PyPI](https://junglewise.ai/threats/vendors/pypi), [vLLM Project](https://junglewise.ai/threats/vendors/vllm-project).

## Quick answers

**Is CVE-2026-114050 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-114050 being exploited in the wild.

**How severe is CVE-2026-114050?** High severity, with a CVSS score of 7.5, published 21 October 2026.

**Which products and versions are affected?** vllm (PyPI) < 0.31.0, ECOSYSTEM:introduced=0,fixed=0.31.0; vLLM Project vllm before 0.31.0.

**Is there a fix for CVE-2026-114050?** Yes. vllm (PyPI): upgrade to 0.31.0 or later. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.

**How is CVE-2026-114050 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-114050?** Check whether you run vllm (PyPI) or vLLM Project vllm at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## How to fix

- vllm (PyPI): upgrade to 0.31.0 or later

  ```sh
  pip install "vllm>=0.31.0"
  ```


## Technical details

vLLM uses client-supplied UUID strings verbatim as cache keys for multimodal content without validating consistency between the frontend (P0) and engine-core (P1) processor caches. These independent LRU caches can evict items asynchronously, and the P1 cache-hit path does not validate that cached tensor shapes or token counts match the current request's placeholders. An attacker exploits this by seeding a cache entry with a small image under UUID "U", flooding P0 with oversized images to evict "U" from P0 only, then reusing "U" with a large image to trigger a shape mismatch (N_A tokens cached vs. N_B placeholders expected), causing an uncaught ValueError in the worker that kills the EngineCore process and terminates all API servers.

## Affected products

- vLLM Project vLLM before 0.31.0

## Identifiers

- CVE-2026-114050
- GHSA-p92p-rxj5-7p2x

## Timeline

- 2026-10-06: disclosed
- 2026-10-21: advisory

## References

- [https://github.com/vllm-project/vllm/security/advisories/GHSA-p92p-rxj5-7p2x](https://github.com/vllm-project/vllm/security/advisories/GHSA-p92p-rxj5-7p2x)
- [https://www.vulncheck.com/advisories/vllm-before-0.31.0-remote-engine-crash-via-forged-multimodal-uuid](https://www.vulncheck.com/advisories/vllm-before-0.31.0-remote-engine-crash-via-forged-multimodal-uuid)

## Related threats

- [vLLM chat template resource exhaustion DoS](https://junglewise.ai/threats/cve-2026-115918-vllm-is-an-inference-and-serving-engine-for-large-language)
- [vLLM: Permissive default CORS origins allow cross-origin requests to the vLLM HTTP server](https://junglewise.ai/threats/cve-2026-115917-vllm-is-an-inference-and-serving-engine-for-large-language)
- [vLLM off-by-one error in tokenizer max_token_id validation](https://junglewise.ai/threats/cve-2026-115916-vllm-is-an-inference-and-serving-engine-for-large-language)
- [vLLM image format validation bypass in media processing](https://junglewise.ai/threats/cve-2026-115915-vllm-is-an-inference-and-serving-engine-for-large-language)
- [vLLM denial of service in DeepSeek-OCR image processing](https://junglewise.ai/threats/cve-2026-115914-vllm-is-an-inference-and-serving-engine-for-large-language)