CVE-2026-116173

vLLM 0.22.0 before 0.24.0 Denial of Service via gRPC output_candidates.token_ids

Published Oct 25, 2026 CVSS 4.0 · 8.2 High EPSS < 1% KEV · Not listed
Description
vLLM from 0.22.0 before 0.24.0 contains a denial of service vulnerability that allows remote attackers to terminate EngineCore by bypassing logprob_token_ids limits in the Rust gRPC frontend. Attackers can send Generate or GenerateStream requests with output_logprobs enabled and more than 128 output_candidates.token_ids, triggering a fatal sampler exception that crashes the engine.

Analysis

AI-generated

Technical impact

A remote attacker can trigger a fatal sampler exception in vLLM versions 0.22.0 through 0.23.x by sending Generate or GenerateStream requests with output_logprobs enabled and more than 128 output_candidates.token_ids. The exploit bypasses input length limits in the Rust gRPC frontend, causing the EngineCore process to terminate unexpectedly and resulting in a service crash. This vulnerability exploits improper input validation (CWE‑20) and includes an error handling flaw (CWE‑248) that leads to a denial of availability of the host model service.

Affected Systems

The affected product is vLLM from the vllm-project organization. All releases of vLLM between 0.22.0 inclusive and before 0.24.0 are vulnerable. The impacted systems run the EngineCore component exposed through the gRPC interface.

Risk and Exploitability

The CVSS score of 8.2 indicates a high severity; the EPSS score is not available, so the current probability of exploitation is unknown. The vulnerability is not listed in the CISA KEV catalog. Attackers can exploit it remotely by sending crafted gRPC requests; no local privilege escalation is required. Because the vulnerability causes a crash of the entire EngineCore, it can destabilize or take down production services that rely on vLLM.

Remediation

No solution or workaround provided in the CVE record.

OpenCVE Recommended Actions

  • Upgrade to vLLM 0.24.0 or later, which removes the vulnerability.
  • If an upgrade is not immediately possible, configure the gRPC service to reject or enforce a maximum of 128 output_candidates.token_ids, or disable output_logprobs in all Generate and GenerateStream requests.
  • Monitor the log streams and system uptime for unusual crashes and apply any vendor‑issued patches promptly.

Generated by OpenCVE AI on October 26, 2026 at 01:52 UTC.

Tracking

Sign in to view the affected projects.

Advisories

No advisories available
History
Type Values Removed Values Added
Metrics ssvc

{'options': {'Automatable': 'no', 'Exploitation': 'none', 'Technical Impact': 'partial'}, 'version': '2.0.3'}

Type Values Removed Values Added
First Time appeared Vllm-project
Vllm-project vllm
Vendors & Products Vllm-project
Vllm-project vllm
Type Values Removed Values Added
Description vLLM from 0.22.0 before 0.24.0 contains a denial of service vulnerability that allows remote attackers to terminate EngineCore by bypassing logprob_token_ids limits in the Rust gRPC frontend. Attackers can send Generate or GenerateStream requests with output_logprobs enabled and more than 128 output_candidates.token_ids, triggering a fatal sampler exception that crashes the engine.
Title vLLM 0.22.0 before 0.24.0 Denial of Service via gRPC output_candidates.token_ids
First Time appeared Vllm
Vllm vllm
Weaknesses CWE-20
CPEs cpe:2.3:a:vllm:vllm:*:*:*:*:*:*:*:*
Vendors & Products Vllm
Vllm vllm
References
Metrics cvssV3_1

{'score': 5.9, 'vector': 'CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H'}

cvssV4_0

{'score': 8.2, 'vector': 'CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N'}

Weaknesses

Sources

Status
PUBLISHED
Assigner
VulnCheck
Published
Updated
Oct 26, 2026, 7:48 PM
Reserved
Oct 25, 2026, 12:25 PM
Status
Analyzed
Published
Oct 25, 2026, 2:16 PM
Modified
Oct 28, 2026, 3:43 PM
Severity
Moderate
Public date
Oct 25, 2026, 1:20 PM