BlackTreeCVE Intelligence
← Back to the CVE catalogue
Full vulnerability report · 2026
CVE-2026-93592High confidence

vLLM before 0.28.0 Denial of Service via negative token ID

vllm-project · vllm

Official source article: GitHub GHSA-25Q3-V2HM-8VPF ↗. Check the applicable product and release in the original source.

8.7HighCVSS 4.0
Recommended action
Patch only the product branches with a verified fix

High technical severity with public exploit material referenced by a structured source; prioritise exposed affected systems while verifying vendor guidance. Verified remediation exists for at least one product or source, but 27 structured product or package states remain unresolved. Apply remediation only to the exact product branch confirmed by its source.

Fix availability varies by product
R
Operational reassessment

Published severity in operational context

Open reassessment dashboard →
Published severityHighOperational priority:Critical, raised one band.upgradedsince 18 Sep 2026

Evidence used

  • No CISA KEV confirmation is currently recorded.
  • A structured source references public exploit or proof-of-concept material.
  • The selected CVSS metric records a network-reachable, unauthenticated path with no user interaction.
  • EPSS is 0.54% for the current model date.

Compensating controls

  • Validate the affected product branch and deploy the verified fixed release.
  • Restrict the affected network interface to trusted sources where business-safe.
  • Increase monitoring for the attack path and post-exploitation behaviour described in the report.

Verification

  1. Confirm that the asset runs vllm-project vllm and falls inside the recorded affected range.
  2. Verify the installed build against the product-specific fixed version after deployment.
  3. Validate exposure, authentication requirements and compensating controls in the actual environment.
  4. Reopen this reassessment when CVSS, KEV, EPSS, exploit evidence or remediation changes.
Mitigation target: Within 3 daysRemediation target: Within 90 days

This automated reassessment organises public evidence. It does not know asset exposure, business impact or control effectiveness and does not replace CVSS or a human risk decision.

Cross-source reconciliation

Remediation availability differs by product scope

Verified remediation exists for at least one product or source, but 27 structured product or package states remain unresolved. Apply remediation only to the exact product branch confirmed by its source.

Open-source package ranges1 source-attributed range

These OSV and GitHub advisory ranges apply only to the named package and ecosystem. A listed fixed version is not a universal product patch or proof that an update is installed.

Ecosystem and packageAffected rangeFirst fixed versionEvidence
PyPIvllmECOSYSTEM: introduced 0; fixed 0.28.00.28.0OSV record ↗source linked ecosystem record · 29 Sep 2026
Direct vendor intelligence

Authoritative vendor CSAF and VEX advisories

Structured product status and remediation from the issuing vendor. Product-state explanations are always visible; large lists can be searched or downloaded.

1 current
CVE-2026-93592 · CSAF 2.0 · revision 3 · finalRed Hat Product Securityvllm: vLLM: Denial of Service via negative token ID input
27 known affected

The vendor explicitly identifies these products as affected by this CVE.

  • rhaii/vllm-cpu-rhel9 as a component of Red Hat AI Inference Server
  • rhaii/vllm-cuda-rhel9 as a component of Red Hat AI Inference Server
  • rhaii/vllm-gaudi-rhel9 as a component of Red Hat AI Inference Server
  • rhaii/vllm-neuron-rhel9 as a component of Red Hat AI Inference Server
  • rhaii/vllm-rocm-rhel9 as a component of Red Hat AI Inference Server
  • rhaii/vllm-spyre-rhel9 as a component of Red Hat AI Inference Server
  • rhaii/vllm-tpu-rhel9 as a component of Red Hat AI Inference Server
  • rhaiis/vllm-cpu-rhel9 as a component of Red Hat AI Inference Server
  • rhaiis/vllm-cuda-rhel9 as a component of Red Hat AI Inference Server
  • rhaiis/vllm-neuron-rhel9 as a component of Red Hat AI Inference Server
  • rhaiis/vllm-rocm-rhel9 as a component of Red Hat AI Inference Server
  • rhaiis/vllm-spyre-rhel9 as a component of Red Hat AI Inference Server
Summary
A flaw was found in vLLM. An unauthenticated attacker can exploit an input validation vulnerability by submitting negative token IDs to the /v1/embeddings and /pooling endpoints. This action triggers a CUDA device-side assertion, which poisons the GPU context and causes the engine to crash. Consequently, all subsequent requests will fail until the process is restarted, leading to a Denial of Service (DoS).
Remediation
Restrict access to the inference service or filter requests at an upstream reverse proxy or API gateway: 1. Limit network access: Deploy firewall rules or OpenShift NetworkPolicies so that the inference server ports and endpoints are accessible only by authenticated, trusted internal clients. 2. Ingress request validation: Configure an API gateway or reverse proxy placed in front of the model server to inspect request payloads for the `/v1/embeddings` and `/pooling` endpoints, rejecting requests where the `input` array contains negative numbers, or restricting untrusted clients to submitting text prompt strings rather than raw token lists. Caveats: Inspecting request bodies at an intermediary gateway layer may introduce slight processing latency. Modifying reverse proxy rules or applying network policies may require a service configuration reload or restart, which can temporarily disrupt in-flight connections.
Optional official sources

National CERT insights
?CERT means Computer Emergency Response Team; CSIRT is the closely related term Computer Security Incident Response Team.

Choose official national sources for this report. Each advisory shows its original language. Your selection is remembered on this device and included in shared links.

Official European source

ENISA European Vulnerability Database

Official EUVD identifiers, advisory evidence and known-exploited context. Missing fields are not treated as evidence of low risk.

1 current
ENISA EUVD identifier

EUVD-2026-82911

No EUVD known-exploited evidence

vLLM versions before 0.28.0 fail to validate the lower bound of token IDs in the /v1/embeddings and /pooling endpoints, allowing unauthenticated attackers to crash the engine by submitting negative token IDs. A single request with a negative token ID triggers a CUDA device-side assertion that poisons the GPU context, causing all subsequent requests to fail until the process restarts.

EUVD state
Present in the current official mapping
Known exploitation
Not present in the current ENISA EUVD known-exploited dataset. This is not proof of no exploitation.
ENISA score
8.7 · CVSS 4.0
Advisory evidence
No linked advisory details stored yet
Recommended actionPatch only the product branches with a verified fix

High technical severity with public exploit material referenced by a structured source; prioritise exposed affected systems while verifying vendor guidance. Verified remediation exists for at least one product or source, but 27 structured product or package states remain unresolved. Apply remediation only to the exact product branch confirmed by its source.

Fix availability varies by product
01

What, why and how

vLLM versions before 0.28.0 fail to validate the lower bound of token IDs in the /v1/embeddings and /pooling endpoints, allowing unauthenticated attackers to crash the engine by submitting negative token IDs. A single request with a negative token ID triggers a CUDA device-side assertion that poisons the GPU context, causing all subsequent requests to fail until the process restarts.

What

vLLM versions before 0.28.0 fail to validate the lower bound of token IDs in the /v1/embeddings and /pooling endpoints, allowing unauthenticated attackers to crash the engine by submitting negative token IDs. A single request with a negative token ID triggers a CUDA device-side assertion that poisons the GPU context, causing all subsequent requests to fail until the process restarts.

Why

The product uses untrusted input when calculating or using an array index, but the product does not validate or incorrectly validates the index to ensure the index references a valid position within the array.

How

An attacker operating through a network path may attempt exploitation without authentication or user interaction. If successful, the issue may disrupt the affected service.

What

vLLM versions before 0.28.0 fail to validate the lower bound of token IDs in the /v1/embeddings and /pooling endpoints, allowing unauthenticated attackers to crash the engine by submitting negative token IDs. A single request with a negative token ID triggers a CUDA device-side assertion that poisons the GPU context, causing all subsequent requests to fail until the process restarts.

Why

The product uses untrusted input when calculating or using an array index, but the product does not validate or incorrectly validates the index to ensure the index references a valid position within the array.

How

An attacker operating through a network path may attempt exploitation without authentication or user interaction. If successful, the issue may disrupt the affected service.

02

Exploit reality and attack path

CVSS severity, EPSS forecast probability, public exploit material and CISA-confirmed exploitation are separate signals.

Observed exploitation
?Confirmed exploitation and public exploit material are separate signals. Attacks can occur without public proof-of-concept or exploit code.
No confirmed evidence

No CISA KEV match was present at the last successful refresh. This means no confirmation from that source, not proof of no exploitation.

Public PoC / exploit material
?Confirmed exploitation and public exploit material are separate signals. Attacks can occur without public proof-of-concept or exploit code.
Reference recorded

A structured CVE source labels at least one public reference as exploit material. BlackTree has not independently validated that it is safe, reliable or weaponised.

Likely attack path
a network path → Improper Validation of Array Index → disrupt the affected service
Attack surface
Network
Privileges required
None: unauthenticated exploitation is possible
User interaction
None
Attack complexity
Low: no specialised conditions are recorded
Security boundary
Not a CVSS 4.0 base metric
Weakness
?CWE means Common Weakness Enumeration: a standard category for the underlying weakness.
CWE-129 ↗

CWE-129: Improper Validation of Array Index. The product uses untrusted input when calculating or using an array index, but the product does not validate or incorrectly validates the index to ensure the index references a valid position within the array.

CVSS vector
?CVSS means Common Vulnerability Scoring System. The vector records the metric values used to calculate technical severity.
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N

Common Vulnerability Scoring System 4.0: the compact vector below is decoded into plain language.

AVNetworkAttack vector: The vulnerable component can be reached over a network.ACLowAttack complexity: No specialised conditions are required beyond attacker-controlled input.ATNoneAttack requirements: No additional deployment or execution condition is required.PRNonePrivileges required: The attacker does not need an account or existing privileges.UINoneUser interaction: No action by another user is required.VCNoneVulnerable-system confidentiality: No direct loss is represented by this metric.VINoneVulnerable-system integrity: No direct loss is represented by this metric.VAHighVulnerable-system availability: A successful attack can cause a major loss.SCNoneSubsequent-system confidentiality: No direct loss is represented by this metric.SINoneSubsequent-system integrity: No direct loss is represented by this metric.SANoneSubsequent-system availability: No direct loss is represented by this metric.
Post-exploitation / living off the land
Stolen credentials, tokens and legitimate administration functions may provide continued access without deploying a large custom toolset.
NetworkUnauthenticatedDenial of serviceCWE-129Public exploit reference
A

Official authority intelligence

Only matched European and national findings are included. Language selectors and unavailable sources are omitted.

ENISA EUVD · EUVD-2026-82911Official EUVD mapping

vLLM versions before 0.28.0 fail to validate the lower bound of token IDs in the /v1/embeddings and /pooling endpoints, allowing unauthenticated attackers to crash the engine by submitting negative token IDs. A single request with a negative token ID triggers a CUDA device-side assertion that poisons the GPU context, causing all subsequent requests to fail until the process restarts.

Official EUVD record ↗
BSI · German · WID-SEC-2026-3187vllm: Schwachstelle ermöglicht Denial of Service

Ein entfernter, anonymer Angreifer kann eine Schwachstelle in vllm ausnutzen, um einen Denial of Service Angriff durchzuführen.

Official advisory ↗
JVN iPedia · Japanese · JVNDB-2026-035803vLLMにおける配列インデックスの検証に関する脆弱性

vLLMの0.28.0未満のバージョンには、/v1/embeddingsおよび/poolingエンドポイントでトークンIDの下限を検証しない脆弱性があります。そのため、認証されていない攻撃者が負のトークンIDを送信することでエンジンをクラッシュさせる可能性があります。負のトークンIDを含む単一のリクエストにより、CUDAデバイス側のアサーションがトリガーされ、GPUコンテキストが破損し、プロセスが再起動されるまで以降のすべてのリクエストが失敗する状態になります。

Official advisory ↗
03

Patch and workaround

Operational remediation based on structured source evidence.

Status
?Patch availability is based on structured fixed-version fields and authoritative update references. If no fix is verified, check the vendor advisory before making a change.
Fix availability varies by product
Affected
Fixed
vllm: 0.28.0
Action
Use the product-specific evidence above. Patch only products with a verified fixed release, and keep every affected or under-investigation state without a matching fix in the remediation queue.
Workaround
Restrict access to the inference service or filter requests at an upstream reverse proxy or API gateway: 1. Limit network access: Deploy firewall rules or OpenShift NetworkPolicies so that the inference server ports and endpoints are accessible only by authenticated, trusted internal clients. 2. Ingress request validation: Configure an API gateway or reverse proxy placed in front of the model server to inspect request payloads for the `/v1/embeddings` and `/pooling` endpoints, rejecting requests where the `input` array contains negative numbers, or restricting untrusted clients to submitting text prompt strings rather than raw token lists. Caveats: Inspecting request bodies at an intermediary gateway layer may introduce slight processing latency. Modifying reverse proxy rules or applying network policies may require a service configuration reload or restart, which can temporarily disrupt in-flight connections.
Official vendor articles1 stored document

Linked articles are downloaded and versioned as source evidence. A CVE mention or approved update relationship does not, by itself, verify a fix for every product branch.

04

Evidence and provenance

Published 18 Sept 2026 · Last source change 18 Sept 2026, 18:09 UTC · CWE-129 · Improper Validation of Array Index

CVE recordCVE.org · 5.2
CVSS sourceCNA
EPSS source
?The date BlackTree first stored a score for this CVE from the daily FIRST EPSS feed.
FIRST · tracked since 2026-09-19
European sourceENISA EUVD · EUVD-2026-82911
Product sourceVendor CSAF · Red Hat Product Security
Remediation sourceVendor CSAF · Red Hat Product Security
CWE sourceCNA
NVD statusNVD enriched

Core structured fields are present and their contributing authorities are shown above.

Material change intelligence

What changed after publication

View recent updates ↗
  1. Vendor guidanceAuthoritative vendor guidance changed from remediation: access.redhat.com/CVE-2026-93592 to remediation: access.redhat.com/CVE-2026-93592.
    Before
    remediation: access.redhat.com/CVE-2026-93592
    After
    remediation: access.redhat.com/CVE-2026-93592
    Red Hat Product Security ↗
  2. Remediation statusRemediation status changed from Awaiting fix to Mitigation available.
    Before
    Awaiting fix
    After
    Mitigation available
    Red Hat Product Security ↗
  3. ENISA EUVD mappingEUVD-2026-82911 was added to the official ENISA EUVD mapping for this CVE.
    Before
    not recorded
    After
    {"euvdId":"EUVD-2026-82911"}
    ENISA EUVD ↗
  4. Vendor guidanceAuthoritative vendor guidance changed: added patch: github.com/GHSA-25q3-v2hm-8vpf.
    Before
    remediation: github.com/GHSA-25q3-v2hm-8vpf
    After
    patch: github.com/GHSA-25q3-v2hm-8vpf · remediation: github.com/GHSA-25q3-v2hm-8vpf
    github.com ↗
  5. Public exploit evidencePublic exploit evidence changed from Public exploit reference to Public exploit reference.
    Before
    Public exploit reference
    After
    Public exploit reference
    github.com ↗
  6. Public exploit evidencePublic exploit evidence changed from No public exploit to Public exploit reference.
    Before
    No public exploit
    After
    Public exploit reference
    CVE.org ↗
  7. Catalogue recordCVE added to the BlackTree catalogue.
    CNA ↗
Material fields only · duplicate refreshes suppressed · history retained for the configured operational retention period
CVE published
Fixed release or patch reference recorded
ENISA EUVD · EUVD-2026-82911↗Vendor CSAF · Red Hat Product Security↗CISA Vulnrichment SSVC evidence↗Public exploit reference · GitHub GHSA-25Q3-V2HM-8VPF↗Official CVE record (CNA and ADP data)↗NVD record↗
Technical terms and abbreviations used in this report
CVE
Common Vulnerabilities and Exposures: the public identifier for one disclosed vulnerability.
CVSS
Common Vulnerability Scoring System: a technical severity framework; it is not patching priority by itself.
EPSS
Exploit Prediction Scoring System: FIRST's estimate of the probability that exploitation activity will be observed in the next 30 days; it is a forecast, not confirmation.
CWE
Common Weakness Enumeration: the standard category describing the underlying software or hardware weakness.
CNA
CVE Numbering Authority: an organisation authorised to assign and publish CVE records.
CISA ADP
Cybersecurity and Infrastructure Security Agency Authorized Data Publisher: structured enrichment added to a CVE record.
NVD
National Vulnerability Database: NIST's enrichment service for CVE records.
CERT / CSIRT
A computer security incident response team that publishes warnings or coordinates incident response.
PoC
Proof of concept: public material that demonstrates or helps reproduce exploitation.
CSAF
Common Security Advisory Framework: a machine-readable format for security advisories.
LoTL
Living off the land: abuse of legitimate tools or system functions during an attack.
Free version - for non-commercial use only.CVE-2026-93592 · cve.blacktree.nl