CVE-2024-8939: vLLM Denial of Service via best_of parameter
vLLM
An improper handling vulnerability in the vLLM JSON web API allows attackers to cause a denial of service by supplying a large best_of parameter. The API becomes unresponsive due to excessive resource exhaustion.
disclosed CVENo confirmed exploitation, so this record is tracked as a disclosed vulnerability and not counted as an incident. It moves into the counts if CISA or the vendor confirms exploitation. Why
- CVSS
- 6.9
- EPSS
- <0.01 · top 86% · 2026-11-13
- fix status
- Fix status unknown
- first report
- 17 Sep 2024
- last update
- 17 Sep 2024
- CVE ids
Timeline · 2
Sources · 2
Advisories and databases
- 2024-09-17CVE-2024-8939: vLLM Denial of Service via the best_of parameterGitHub Advisory Database · github.com/advisories/GHSA-wc36-9694-f9rf
- 2024-09-17CVE-2024-8939: A vulnerability was found in the ilab model serve component, where improper handling of the best_of parameter in the vllm JSON…NVD · nvd.nist.gov/vuln/detail/CVE-2024-8939
Titles and links belong to their publishers. The summary above is written by gcdRadar in its own words; data is CC BY-SA 4.0.
Classification
- lane
- Patch now
- AI role
- Attack on AI
- categories
- Infrastructure vulnerability
- severity
- 2 · Disclosed vulnerability
- fix status
- Fix status unknown
- targets
- Framework / tooling
- owasp llm top 10
- LLM03:2025 Supply Chain
- mitre atlas
- no technique for these categories
- dates
- happened 17 Sep 2024 · first reported 17 Sep 2024
- origin
- GitHub Advisory DatabaseVulnerability record from NVD / the GitHub Advisory Database. This product uses the NVD API but is not endorsed or certified by the NVD.
- record
- confidence 0.95 · created 2026-10-18 07:00Z · updated 2026-10-18 10:04Z
Lane, AI role, categories and severity come from the classifier; the OWASP and ATLAS rows are derived from the categories. How these are assigned
Copy a brief
This record as a short Markdown brief, for a ticket, a chat or a coding agent.
the brief as text
# CVE-2024-8939: vLLM Denial of Service via best_of parameter https://gcdradar.vercel.app/incidents/cve-2024-8939-vllm-denial-of-service-via-best-of-parameter · AIR-2024-0166 · closed 2024-09-17 - Lane: Patch now - Severity: 2 (Disclosed vulnerability) - Fix status: Fix status unknown - First reported: 2024-09-17 - AI role: Attack on AI - Products: vLLM - CVE ids: CVE-2024-8939 - CVSS: 6.9 ## Summary An improper handling vulnerability in the vLLM JSON web API allows attackers to cause a denial of service by supplying a large best_of parameter. The API becomes unresponsive due to excessive resource exhaustion. ## What to do No action recorded; read the sources below and the vendor's advisory. ## Latest updates - 2024-09-17: CVE-2024-8939: A vulnerability was found in the ilab model serve component, where improper handling of the best_of parameter in the vllm JSON… (https://nvd.nist.gov/vuln/detail/CVE-2024-8939) - 2024-09-17: CVE-2024-8939: vLLM Denial of Service via the best_of parameter (https://github.com/advisories/GHSA-wc36-9694-f9rf) ## Sources - 2024-09-17 GitHub Advisory Database: CVE-2024-8939: vLLM Denial of Service via the best_of parameter https://github.com/advisories/GHSA-wc36-9694-f9rf - 2024-09-17 NVD: CVE-2024-8939: A vulnerability was found in the ilab model serve component, where improper handling of the best_of parameter in the vllm JSON… https://nvd.nist.gov/vuln/detail/CVE-2024-8939 Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.
Related
- closed8.7vLLM CVE-2024-8768: vLLM denial of service vulnerability
Patch now · Disclosed · since 17 Sep 2024 · CVE-2024-8768
AI - closed2.6vLLM CVE-2025-25183: vLLM vulnerable to hash collisions in prefix cache
Patch now · Disclosed · since 6 Feb 2025 · updated 7 Feb 2025 · CVE-2025-25183
AI - closed7.1vLLM CVE-2025-66448: vLLM remote code execution vulnerability in config loading
Patch now · Disclosed · since 2 Dec 2025 · updated 2 Dec 2025 · CVE-2025-66448
AI - closed8.8vLLM CVE-2025-62164: vLLM memory corruption via unvalidated prompt-embedding tensors
Patch now · Disclosed · since 20 Nov 2025 · updated 21 Nov 2025 · CVE-2025-62164
AI - closed6.5vLLM CVE-2025-62426: vLLM Vulnerable to DoS via Chat Completion Request Parameters
Patch now · Disclosed · since 20 Nov 2025 · updated 21 Nov 2025 · CVE-2025-62426
AI - closed8.3vLLM CVE-2025-62372: vLLM vulnerable to denial of service via multimodal embedding inputs
Patch now · Disclosed · since 20 Nov 2025 · updated 21 Nov 2025 · CVE-2025-62372
AI