Read the docs
TI-2026-32202 Aliases CVE-2026-44223GHSA-83vm-p52w-f9pw

vLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parameters

Priority MEDIUM 62CWE-131Published · Updated

Key facts

Attack vector
Network
Privileges required
Low
User interaction
None
Attack complexity
Low
Impact type
Denial of service
Confidentiality · integrity · availability
None · None · High
Fix status
Partial Red Hat: 2 fixed, 2 without a fix

Attack profile from CVSS 3.1 · CNA. Fix status from CNA, GitHub and Red Hat. Impact type from CWE-131 and the CVSS impact.

Summary

vLLM is an inference and serving engine for large language models (LLMs). From 0.18.0 to before 0.20.0, the extract_hidden_states speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The crash is triggered when any request in the batch uses sampling penalty parameters (repetition_penalty, frequency_penalty, or presence_penalty). A single request with a penalty parameter (e.g., "repetition_penalty": 1.1) is sufficient to crash the server. This vulnerability is fixed in 0.20.0.

Why it matters now

Exploitation Proof of concept CISA SSVC · not in CISA KEV
EPSS 0.00367 30th percentileScored 2026-11-07
Fix Partial vllm (vllm-project): 0.20.0vllm (PyPI): 0.20.0Red Hat: 14 advisoriesRed Hat: 2 fixed, 2 without a fix

SSVC decision points

CISA ADP · SSVC 2.0.3

Exploitation
Poc
Automatable
No
Technical impact
Partial

The decision (Track, Attend, Act) also depends on your mission and well-being impact; see CISA's SSVC guide.

Affected products

VendorProductAffected versionsFixed inStatusSource
vllm-projectvllm0.18.0 to before 0.20.00.20.0Patch availableCNA
—vllm (PyPI)0.18.0 to before 0.20.00.20.0Patch availableGHSA-83vm-p52w-f9pw
  • vllm Vendorvllm-project Affected 0.18.0 to before 0.20.0 Fixed in 0.20.0 Patch available CNA
  • vllm (PyPI) Vendor— Affected 0.18.0 to before 0.20.0 Fixed in 0.20.0 Patch available GHSA-83vm-p52w-f9pw

CPE configurations NVD · 1

CPEVersionsRoleVendor
cpe:2.3:a:vllm:vllm:*:*:*:*:*:*:*:*0.18.0 to before 0.20.0VulnerablevLLM
  • cpe:2.3:a:vllm:vllm:*:*:*:*:*:*:*:* Versions 0.18.0 to before 0.20.0 Vulnerable ·vLLM

Remediation

  • Fix · CNA vllm (vllm-project) Fixed in 0.20.0
  • Fix · GHSA-83vm-p52w-f9pw vllm (PyPI) Fixed in 0.20.0 Advisory →

Red Hat 14 advisories

  • RHSA-2026:70995 Released 2026-09-23 · Red Hat rating Important
    Red Hat AI Inference 3.4.5 (rocm)
    • Red Hat AI Inference Server 3.4 registry.redhat.io/rhaii/vllm-rocm-rhel9
  • RHSA-2026:70979 Released 2026-09-23 · Red Hat rating Important
    Red Hat AI Inference 3.4.5 (cuda)
    • Red Hat AI Inference Server 3.4 registry.redhat.io/rhaii/vllm-cuda-rhel9
  • RHSA-2026:70969 Released 2026-09-23 · Red Hat rating Important
    Red Hat AI Inference 3.4.5 (spyre)
    • Red Hat AI Inference Server 3.4 registry.redhat.io/rhaii/vllm-spyre-rhel9
Show 11 more Red Hat advisories
  • RHSA-2026:70965 Released 2026-09-23 · Red Hat rating Important
    Red Hat AI Inference 3.4.5 (cpu)
    • Red Hat AI Inference Server 3.4 registry.redhat.io/rhaii/vllm-cpu-rhel9
  • RHSA-2026:69469 Released 2026-09-21 · Red Hat rating Important
    Red Hat AI Inference 3.4.5 (rocm)
    • Red Hat AI Inference Server 3.4 registry.redhat.io/rhaii/vllm-rocm-rhel9
  • RHSA-2026:69467 Released 2026-09-21 · Red Hat rating Important
    Red Hat AI Inference 3.4.5 (cuda)
    • Red Hat AI Inference Server 3.4 registry.redhat.io/rhaii/vllm-cuda-rhel9
  • RHSA-2026:69466 Released 2026-09-21 · Red Hat rating Important
    Red Hat AI Inference 3.4.5 (cpu)
    • Red Hat AI Inference Server 3.4 registry.redhat.io/rhaii/vllm-cpu-rhel9
  • RHSA-2026:69464 Released 2026-09-21 · Red Hat rating Important
    Red Hat AI Inference 3.4.5 (spyre)
    • Red Hat AI Inference Server 3.4 registry.redhat.io/rhaii/vllm-spyre-rhel9
  • RHSA-2026:59151 Released 2026-08-24 · Red Hat rating Important
    Red Hat Enterprise Linux AI 3.4.4 enhancement update
    • Red Hat Enterprise Linux AI 3.4 registry.redhat.io/rhelai3/bootc-aws-cuda-rhel9, registry.redhat.io/rhelai3/bootc-azure-cuda-rhel9, registry.redhat.io/rhelai3/bootc-azure-rocm-rhel9 +3
  • RHSA-2026:59144 Released 2026-08-24 · Red Hat rating Important
    Red Hat Enterprise Linux AI 3.4.4 enhancement update
    • Red Hat Enterprise Linux AI 3.4 registry.redhat.io/rhelai3/disk-image-cuda-rhel9
  • RHSA-2026:57390 Released 2026-08-20 · Red Hat rating Important
    Red Hat AI Inference 3.4.4 (rocm)
    • Red Hat AI Inference Server 3.4 registry.redhat.io/rhaii/vllm-rocm-rhel9
  • RHSA-2026:57389 Released 2026-08-20 · Red Hat rating Important
    Red Hat AI Inference 3.4.4 (cuda)
    • Red Hat AI Inference Server 3.4 registry.redhat.io/rhaii/vllm-cuda-rhel9
  • RHSA-2026:57387 Released 2026-08-20 · Red Hat rating Important
    Red Hat AI Inference 3.4.4 (spyre)
    • Red Hat AI Inference Server 3.4 registry.redhat.io/rhaii/vllm-spyre-rhel9
  • RHSA-2026:57380 Released 2026-08-20 · Red Hat rating Important
    Red Hat AI Inference 3.4.4 (cpu)
    • Red Hat AI Inference Server 3.4 registry.redhat.io/rhaii/vllm-cpu-rhel9

Vendor assessments

Red Hat

Red Hat rating: Important. Red Hat's page for this CVE →

ProductStatusPlan or fix
Red Hat AI Inference ServerAffectedNo fix yet
Red Hat OpenShift AI (RHOAI)AffectedNo fix planned: Will not fix
Red Hat AI Inference Server 3.4Fixed RHSA-2026:57380, RHSA-2026:57387, RHSA-2026:57389, RHSA-2026:57390, RHSA-2026:69464, RHSA-2026:69466, RHSA-2026:69467, RHSA-2026:69469, RHSA-2026:70965, RHSA-2026:70969, RHSA-2026:70979, RHSA-2026:70995
Red Hat Enterprise Linux AI 3.4Fixed RHSA-2026:59144, RHSA-2026:59151
Not affected: 1 product
  • Red Hat OpenShift AI (RHOAI)

Timeline

  1. 2026-10-02 Threatist record created from CVE-2026-44223
  2. 2026-10-03 Alias added: GHSA-83vm-p52w-f9pw
  3. 2026-10-12 Affected versions changed · Red Hat

References

Patch 1

Advisory 2

Other 2

  • TI-2024-10229 Remote Code Execution in vllm-project/vllm CVSS 3.0CRITICAL 9.8
  • TI-2024-08404 Remote Code Execution in vllm-project/vllm CVSS 3.1CRITICAL 9.8
  • TI-2026-36017 vLLM: OpenAI auth bypass CVSS 3.1CRITICAL 9.1
  • TI-2026-63056 vLLM before 0.28.0 Denial of Service via negative token ID CVSS 4.0HIGH 8.7
  • TI-2026-04308 Hardcoded trust_remote_code=True in vllm-project/vllm Bypasses User Security Control CVSS 3.0HIGH 8.8
  • TI-2026-39137 vLLM: Remote DoS in vLLM via Invalid Recovered Token Reinjection CVSS 3.1HIGH 7.5
See all 124 →

Sources

This product uses the NVD API but is not endorsed or certified by the NVD. Sources and licences