Skip to main content

vLLM CVE-2026-37237

| EUVDEUVD-2026-67775 HIGH
Uncontrolled Resource Consumption (CWE-400)
2026-08-28 mitre GHSA-63mw-rppr-f2pf
7.5
CVSS 3.1 · Vendor: mitre
Share

Severity by source

Vendor (mitre) PRIMARY
7.5 HIGH
AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
vuln.today AI
7.5 HIGH

Network fetch triggered by unauthenticated API request; no confidentiality or integrity impact, only availability via memory exhaustion.

3.1 AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
4.0 AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N

Primary rating from Vendor (mitre).

CVSS VectorVendor: mitre

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
Attack Vector
Network
Attack Complexity
Low
Privileges Required
None
User Interaction
None
Scope
Unchanged
Confidentiality
None
Integrity
None
Availability
High

Lifecycle Timeline

5
Source Code Evidence Fetched
Aug 28, 2026 - 22:34 vuln.today
Analysis Generated
Aug 28, 2026 - 22:34 vuln.today
CVSS changed
Aug 28, 2026 - 22:22 NVD
7.5 (HIGH)
CVE Published
Aug 28, 2026 - 00:00 cve.org
HIGH 7.5
CVE Published
Aug 28, 2026 - 00:00 cve.org
UNKNOWN (no severity yet)

DescriptionCVE.org

vLLM up to and including 0.17.0 allows remote attackers to cause a Denial of Service via memory exhaustion. The AsyncMediaIO.fetch_audio and AsyncMediaIO.fetch_image functions in multimodal/inputs.py fetch user-supplied media URLs using aiohttp and call r.read() without enforcing a maximum response size, allowing an attacker to exhaust server memory by providing a URL to an arbitrarily large file.

AnalysisAI

Memory exhaustion via unbounded URL fetch in vLLM up to 0.17.0 allows remote unauthenticated attackers to crash the inference server by supplying a multimodal API request pointing to an arbitrarily large file. The AsyncMediaIO.fetch_audio and AsyncMediaIO.fetch_image functions in multimodal/inputs.py use aiohttp to retrieve user-controlled URLs but call r.read() with no response-size cap, meaning the server will buffer the entire remote payload into process memory. …

Unlock full vulnerability intelligence

  • Risk assessment & exploitation conditions
  • Attack chain visualization
  • Remediation with exact patch versions
  • Threat intelligence from 22 sources
  • Personal watchlist & email alerts

Free forever · No credit card required

Attack ChainAIDerived

Hypothetical attack flow derived from CVE metadata

Access
technique details hidden
Delivery
technique details hidden
Exploit
technique details hidden
Execution
technique details hidden
Impact
technique details hidden

Vulnerability AssessmentAI

Exploitation Exploitation requires that the vLLM instance is serving a multimodal model (one that processes image or audio inputs) so that the AsyncMediaIO code path is reachable. … Additional conditions and limiting factors are described in the full assessment.
Risk Assessment The CVSS 3.1 score of 7.5 (AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H) accurately characterizes the exploitability: the attack requires no authentication, no special timing, and no user interaction beyond submitting a well-formed API request. … Full risk analysis with EPSS, KEV, and SSVC signal comparison available after sign-in.
Exploit Scenario Full exploit scenario with step-by-step reproduction available after sign-in.
Remediation The primary fix is to upgrade vLLM beyond 0.17.0 once a patched release incorporating GitHub PR #36506 is available; the exact released fix version is not independently confirmed from the provided data, so monitor the vLLM GitHub releases page. … Detailed patch versions, workarounds, and compensating controls in full report.

Recommended ActionAI

Within 24 hours, inventory all vLLM deployments running version 0.17.0 or earlier and immediately restrict network access to multimodal API endpoints using firewall rules or API gateway policies to trusted internal clients only; apply network segmentation if multimodal endpoints are not business-critical. …

Sign in for detailed remediation steps and compensating controls.

Threat intelligence, references, and detailed analysis are available after sign-in.

More in N A

View all
CVE-2026-30623 CRITICAL POC
9.8 Jul 15

Command injection in LiteLLM 1.18.10 lets attackers who can configure MCP (Model Context Protocol) servers supply arbitr

CVE-2026-67917 CRITICAL POC
9.8 Aug 17

SQL injection in AzuraCast's backup restore flow (versions up to and including 0.23.7) lets an attacker achieve privileg

CVE-2026-75337 CRITICAL POC
9.8 Aug 27

Arbitrary file read in Yu AI Code Mother v4.3 lets anonymous, remote attackers retrieve files outside the intended previ

CVE-2026-31072 CRITICAL POC
9.8 May 19

Remote code execution in APScheduler (all versions through 3.10.x and 4.0.0a5) is achievable when applications deseriali

CVE-2026-45185 CRITICAL POC
9.8 May 12

Remote code execution in the Exim mail transfer agent (versions before 4.99.3) built against certain GnuTLS configuratio

CVE-2025-50455 CRITICAL POC
9.1 Jul 27

SQL injection in EasyAppointments <=1.5.1 allows unauthenticated attackers to extract and modify database contents via t

CVE-2026-36356 CRITICAL POC
9.1 May 05

Unauthenticated remote OS command injection in MeiG Smart FORGE_SLT711 cellular gateway firmware MDM9607.LE.1.0-00110-ST

CVE-2026-31071 CRITICAL POC
9.1 May 19

Unauthenticated API access in LalanaChami Pharmacy Management System (commit 5c3d028) allows remote attackers to dump al

CVE-2025-66391 HIGH POC
8.8 Jun 17

Privilege escalation via broken access control in Citrix Cloud (through 2025-11-10) lets a user holding only read-only p

CVE-2026-75419 HIGH POC
8.8 Aug 27

go-wind-cms (GoWind) before 1.0.0 exposes all administrative APIs to any authenticated user because NewAuthorizer() retu

CVE-2026-26740 HIGH POC
8.2 Mar 18

Giflib 5.2.2 contains a buffer overflow in the EGifGCBToExtension function that fails to validate allocated memory when

CVE-2025-60464 HIGH POC
7.8 Jun 25

Denial of service in GPAC's MP4Box multimedia tool (versions before 26.02.0) arises from a use-after-free in the gf_sei_

Share

CVE-2026-37237 vulnerability details – vuln.today

This site uses cookies essential for authentication and security. No tracking or analytics cookies are used. Privacy Policy