2026-09-26 13:23CVE-2026-100654VulnCheck
PUBLISHED5.2CWE-129

vLLM before 0.29.0 Denial of Service via out-of-range stop_token_ids

vLLM before 0.29.0 accepts user-controlled stop_token_ids on the OpenAI-compatible POST /v1/completions and POST /v1/chat/completions endpoints but validates only that the values are integers, not that each token id is within the model vocabulary/logits range. When min_tokens > 0, the stop token ids are used as logits indices to suppress stop tokens, so an out-of-range id reaches a CUDA indexing operation (index_put_) and triggers a device-side assertion. An authenticated API user can send a single malformed completion request that returns 500 Internal Server Error and puts EngineCore into a fatal state, causing subsequent requests to fail until the service is restarted (denial of service).

Problem type

Affected products

vllm-project

vllm

< 0.29.0 - AFFECTED

0.29.0 - UNAFFECTED

References

GitHub Security Advisories

GHSA-9fp6-6v59-w2jx

vLLM before 0.29.0 accepts user-controlled stop_token_ids on the OpenAI-compatible POST /v1...

https://github.com/advisories/GHSA-9fp6-6v59-w2jx

vLLM before 0.29.0 accepts user-controlled stop_token_ids on the OpenAI-compatible POST /v1/completions and POST /v1/chat/completions endpoints but validates only that the values are integers, not that each token id is within the model vocabulary/logits range. When min_tokens > 0, the stop token ids are used as logits indices to suppress stop tokens, so an out-of-range id reaches a CUDA indexing operation (index_put_) and triggers a device-side assertion. An authenticated API user can send a single malformed completion request that returns 500 Internal Server Error and puts EngineCore into a fatal state, causing subsequent requests to fail until the service is restarted (denial of service).

JSON source

https://cveawg.mitre.org/api/cve/CVE-2026-100654
Click to expand
{
  "dataType": "CVE_RECORD",
  "dataVersion": "5.2",
  "cveMetadata": {
    "cveId": "CVE-2026-100654",
    "assignerOrgId": "83251b91-4cc7-4094-a5c7-464a1b83ea10",
    "assignerShortName": "VulnCheck",
    "dateUpdated": "2026-09-26T13:23:23.312Z",
    "dateReserved": "2026-09-26T02:33:07.899Z",
    "datePublished": "2026-09-26T13:23:23.312Z",
    "state": "PUBLISHED"
  },
  "containers": {
    "cna": {
      "providerMetadata": {
        "orgId": "83251b91-4cc7-4094-a5c7-464a1b83ea10",
        "shortName": "VulnCheck",
        "dateUpdated": "2026-09-26T13:23:23.312Z"
      },
      "datePublic": "2026-09-12T00:00:00.000Z",
      "title": "vLLM before 0.29.0 Denial of Service via out-of-range stop_token_ids",
      "descriptions": [
        {
          "lang": "en",
          "value": "vLLM before 0.29.0 accepts user-controlled stop_token_ids on the OpenAI-compatible POST /v1/completions and POST /v1/chat/completions endpoints but validates only that the values are integers, not that each token id is within the model vocabulary/logits range. When min_tokens > 0, the stop token ids are used as logits indices to suppress stop tokens, so an out-of-range id reaches a CUDA indexing operation (index_put_) and triggers a device-side assertion. An authenticated API user can send a single malformed completion request that returns 500 Internal Server Error and puts EngineCore into a fatal state, causing subsequent requests to fail until the service is restarted (denial of service)."
        }
      ],
      "affected": [
        {
          "vendor": "vllm-project",
          "product": "vllm",
          "defaultStatus": "unaffected",
          "versions": [
            {
              "version": "0",
              "status": "affected",
              "versionType": "semver",
              "lessThan": "0.29.0"
            },
            {
              "version": "0.29.0",
              "status": "unaffected",
              "versionType": "semver"
            }
          ]
        }
      ],
      "problemTypes": [
        {
          "descriptions": [
            {
              "lang": "en",
              "description": "Improper Validation of Array Index",
              "cweId": "CWE-129",
              "type": "CWE"
            }
          ]
        }
      ],
      "references": [
        {
          "url": "https://github.com/vllm-project/vllm/security/advisories/GHSA-v5gm-qgmv-gc6c",
          "name": "GitHub Security Advisory (GHSA-v5gm-qgmv-gc6c)",
          "tags": [
            "vendor-advisory"
          ]
        },
        {
          "url": "https://www.vulncheck.com/advisories/vllm-before-0.29.0-denial-of-service-via-out-of-range-stop-token-ids",
          "name": "VulnCheck Advisory: vLLM before 0.29.0 Denial of Service via out-of-range stop_token_ids",
          "tags": [
            "third-party-advisory"
          ]
        }
      ],
      "metrics": [
        {
          "format": "CVSS"
        },
        {
          "format": "CVSS",
          "cvssV3_1": {
            "version": "3.1",
            "vectorString": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H",
            "attackVector": "NETWORK",
            "attackComplexity": "LOW",
            "privilegesRequired": "LOW",
            "userInteraction": "NONE",
            "scope": "UNCHANGED",
            "confidentialityImpact": "NONE",
            "integrityImpact": "NONE",
            "availabilityImpact": "HIGH",
            "baseScore": 6.5,
            "baseSeverity": "MEDIUM"
          }
        }
      ],
      "credits": [
        {
          "lang": "en",
          "value": "passer12",
          "type": "reporter"
        },
        {
          "lang": "en",
          "value": "jperezdealgaba",
          "type": "coordinator"
        },
        {
          "lang": "en",
          "value": "QwertyJack",
          "type": "finder"
        }
      ]
    }
  }
}