Improper Validation of Array Index Affecting vllm package, versions [,0.28.0)


Severity

Recommended
0.0
high
0
10

CVSS assessment by Snyk's Security Team. Learn more

Threat Intelligence

Exploit Maturity
Proof of Concept
EPSS
0.51% (41st percentile)

Do your applications use this vulnerable package?

In a few clicks we can analyze your entire application and see what components are vulnerable in your application, and suggest you quick fixes.

Test your applications
  • Snyk IDSNYK-PYTHON-VLLM-20044428
  • published22 Sept 2026
  • disclosed18 Sept 2026
  • creditECMGit

Introduced: 18 Sep 2026

NewCVE-2026-93592  (opens in a new tab)
CWE-129  (opens in a new tab)

How to fix?

Upgrade vllm to version 0.28.0 or higher.

Overview

vllm is an A high-throughput and memory-efficient inference and serving engine for LLMs

Affected versions of this package are vulnerable to Improper Validation of Array Index via InputProcessor._validate_model_input in the /v1/embeddings and /pooling endpoints, which validates only the upper bound of caller-supplied token IDs but not the lower bound. An unauthenticated attacker can submit a single request containing a negative token ID (e.g. -1), which bypasses validation and is used directly as an index into the GPU embedding table, triggering a CUDA device-side assertion inside PyTorch's index kernel. This assertion poisons the CUDA context, causing all subsequent CUDA operations in the process to fail, crashing the engine and rendering every subsequent request from every client unserviceable until the process is restarted.

Workaround

Users that are not able to upgrade to the fixed version are advised to:

  1. Reject request bodies containing negative integers in input at a reverse proxy or API gateway in front of vLLM.

  2. Do not expose the token-id form of the input; accept only string input from untrusted callers.

  3. Restrict network access to /v1/embeddings and /pooling.

CVSS Base Scores

version 4.0
version 3.1