cve.bz

CVE-2026-116176 vllm vllm: SSRF

8.8 highEPSS 0.3%Fix available

High (8.8) in vllm vllm. EPSS 0.3% estimated probability of exploitation activity in the next 30 days. Fixed in vllm 0.24.0.

On this page
vLLM from 0.9.2 before 0.24.0 contains a server-side request forgery vulnerability that allows API callers to force outbound connections by embedding a decode address in the X-Request-Id header. Attackers can make a P2pNcclConnector prefill worker ZMQ-connect to an attacker host, leaking the NCCL unique_id and KV-cache tensors or blocking the worker.

Severity

CVSS 4.0

8.8 high
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:N/VA:L/SC:N/SI:N/SA:N

CVSS 3.1

8.2 high
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:L
EPSS
0.3% (unchanged from 0.3% a week ago) estimated probability of exploitation activity in the next 30 days; 25th percentile. Scored 2026-11-17. A forecast for this CVE across the internet, not for your own systems.
CISA's assessment (SSVC)
Exploitation: none · Automatable: yes · Technical impact: partial, assessed 2026-11-04. From CISA's Vulnrichment in the CVE List: whether it is exploited, whether exploiting it can be automated, and how much control it gives.
CWE
CWE-918 Server-Side Request Forgery (SSRF)
OWASP Top 10
A01:2025 Broken Access Control
Published
Modified
NVD analysis
Analyzed
Record from
the NVD, synced

Fix and patched versions

Ask an AI about this CVE

Show the promptExplain CVE-2026-116176 for an engineer deciding whether to act. Use https://cve.bz/api/public/cves/CVE-2026-116176?via=ask-ai-copy and https://cve.bz/cve/CVE-2026-116176?via=ask-ai-copy as the source of truth: what is affected and which versions fix it, whether it is in CISA KEV, public exploits, and its EPSS (an internet-wide 30-day estimate of exploitation activity, not a forecast for my systems). If they give no fixed versions, use the CVE record https://cveawg.mitre.org/api/cve/CVE-2026-116176 or the vendor advisory, not third-party mirrors. Tell me how to check whether I am affected and how to verify a fix. Cite the sources and say where they are uncertain.

Opens a third-party assistant; its answer is its own, not cve.bz's. The prompt holds only this CVE's id and public links.

Affected

Versions

From the CNA's record in the CVE List.

ProductAffected
vllm-project vllmfrom 0.9.2 before 0.24.0

Packages (OSV)

AdvisoryEcosystemPackageAffectedFixed in
PYSEC-2026-4351PyPIvllm>=0.9.2 <0.24.00.24.0

Red Hat

Red Hat rates this Important.

Red Hat's statement

This vulnerability is rated as Important because remote API users can trigger unauthorized outbound network connections from inference workers, allowing exfiltration of internal cluster identifiers and sensitive key-value cache data, or causing worker denial of service. In Red Hat OpenShift AI and Red Hat Enterprise Linux AI, exploitation is contingent upon explicitly enabling the optional P2pNcclConnector within the KV-transfer configuration. Default deployments do not enable this disaggregated caching feature and are therefore not exposed to the vulnerable connection handler.

Red Hat's mitigation

To mitigate this vulnerability, apply one or more of the following operational controls:

1. Disable disaggregated KV-transfer:

Ensure that vLLM model serving workloads do not configure `--kv-transfer-config` to use `P2pNcclConnector`. Standard model serving without this optional connector avoids invoking the vulnerable address parser. Note that changing container deployment arguments will require restarting the model serving pod, resulting in temporary service interruption.

2. Sanitize ingress request headers:

Configure the ingress controller, API gateway, or reverse proxy fronting the inference service to strip or overwrite incoming `X-Request-Id` HTTP headers with internally generated unique IDs before proxying traffic to the backend vLLM containers.

3. Restrict pod network egress:

Apply OpenShift NetworkPolicy resources or firewall egress rules to restrict outbound network connections from the model serving namespace or pods, permitting communication only to trusted internal endpoints and blocking arbitrary external connections.

Caveats: Disabling `P2pNcclConnector` disables disaggregated prefill and decode functionality. Overwriting request headers may affect distributed request tracing if upstream clients rely on custom tracking headers.

--- CME Mitigation Guidance (2026-11-02) ---

Disclaimer: The following mitigation solutions are provided as a courtesy, pending the application of an official patch. This content was generated utilizing AI tools in conjunction with data from https://cmetaxonomy.org. Customers are advised to independently evaluate and test these temporary mitigations within their own infrastructure.

The highest-value action is to disable P2pNcclConnector in KV-transfer configurations to eliminate unvalidated address parsing from the X-Request-Id header and prevent backend workers from initiating outbound ZMQ connections: upgrade to vLLM 0.24.0 (or the fixed Rhoai/Odh Vllm Cuda Rhel/Rhoai/Odh Vllm Rocm Rhel/Rhaii/Vllm Neuron Rhel/Rhaii/Vllm Tpu Rhel/Rhoai/Odh Vllm Gaudi Rhel/Rhaii Fast/Vllm Neuron Rhel/Rhoai/Odh Kserve Agent Rhel/Rhoai/Odh Kserve Controller Rhel/Rhoai/Odh Kserve Router Rhel/Rhoai/Odh Kserve Storage Initializer Rhel/Rhaii/Vllm Gaudi Rhel/Rhelai3/Bootc Gaudi Rhel/Rhelai3/Bootc Aws Cuda Rhel/Rhelai3/Bootc Azure Cuda Rhel/Rhelai3/Bootc Azure Rocm Rhel/Rhelai3/Bootc Cuda Rhel/Rhelai3/Bootc Gcp Cuda Rhel/Rhelai3/Bootc Rocm Rhel/Rhaii/Vllm Cpu Rhel/Rhaii/Vllm Cuda Rhel/Rhaii/Vllm Rocm Rhel/Rhaii/Vllm Spyre Rhel/Rhoai/Odh Llm D Kv Cache Rhel/Rhoai/Odh Th Torch Cuda Py312 Rhel/Rhaii Fast/Vllm Gaudi Rhel/Rhaii Fast/Vllm Omni Cuda Rhel/Rhaii Fast/Vllm Spyre Rhel/Rhaii Fast/Vllm Tpu Rhel/Rhaii Fast/Vllm Cpu Rhel/Rhaii Fast/Vllm Rocm Rhel/Rhaii Fast/Vllm Cuda Rhel builds), which remove the vulnerable P2pNcclConnector component, closing crafted X-Request-Id HTTP header containing an arbitrary decode network address. Until patched, this is only a concern on vLLM deployments exposing the HTTP API endpoint configured with disaggregated KV-transfer using P2pNcclConnector — Default vLLM deployments running standard model serving without P2pNcclConnector enabled are unaffected — so Do not configure --kv-transfer-config to use P2pNcclConnector, strip or overwrite incoming X-Request-Id HTTP headers at ingress gateways, and apply egress network policies to block outbound connections from inference pods. As a compensating measure, Enforce allowlist validation on incoming HTTP headers to reject crafted X-Request-Id values containing arbitrary network addresses and Deploy Falco eBPF runtime probes to detect anomalous outbound network connections initiated by vLLM worker processes to unexpected external endpoints. These controls limit blast radius and contain exploitation, but they do not fix the underlying server-side request forgery defect — apply the upstream patch to fully remediate.

Highest-impact CME: CME-1304 SSRF Prevention (Outbound Request Allowlist) (high)

ProductPackageStateAdvisory
Red Hat AI Inference Server 3.5rhaii/vllm-cpu-rhel9:1793444554Fixed 2026-11-03RHSA-2026:94702
Red Hat AI Inference Server 3.5rhaii/vllm-rocm-rhel9:1793444456Fixed 2026-11-03RHSA-2026:94709
Red Hat AI Inference Server 3.5rhaii/vllm-spyre-rhel9:1793444409Fixed 2026-11-03RHSA-2026:94717
Red Hat AI Inference Server 3.5rhaii/vllm-gaudi-rhel9:1793444510Fixed 2026-11-03RHSA-2026:94721
Red Hat AI Inference Server 3.5rhaii/vllm-cuda-rhel9:1793444600Fixed 2026-11-03RHSA-2026:94738
Red Hat AI Inference Serverrhaii-fast/vllm-cpu-rhel9Affected
Red Hat AI Inference Serverrhaii-fast/vllm-cuda-rhel9Affected
Red Hat AI Inference Serverrhaii-fast/vllm-gaudi-rhel9Affected
Red Hat AI Inference Serverrhaii-fast/vllm-neuron-rhel9Will not fix
Red Hat AI Inference Serverrhaii-fast/vllm-omni-cuda-rhel9Will not fix
Red Hat AI Inference Serverrhaii-fast/vllm-rocm-rhel9Affected
Red Hat AI Inference Serverrhaii-fast/vllm-spyre-rhel9Affected
Red Hat AI Inference Serverrhaii-fast/vllm-tpu-rhel9Will not fix
Red Hat AI Inference Serverrhaii/vllm-neuron-rhel9Will not fix
Red Hat AI Inference Serverrhaii/vllm-tpu-rhel9Will not fix
Red Hat Enterprise Linux AI (RHEL AI) 3rhelai3/bootc-aws-cuda-rhel9Affected
Red Hat Enterprise Linux AI (RHEL AI) 3rhelai3/bootc-azure-cuda-rhel9Affected
Red Hat Enterprise Linux AI (RHEL AI) 3rhelai3/bootc-azure-rocm-rhel9Affected
Red Hat Enterprise Linux AI (RHEL AI) 3rhelai3/bootc-cuda-rhel9Affected
Red Hat Enterprise Linux AI (RHEL AI) 3rhelai3/bootc-gaudi-rhel9Affected
11 more in the JSON and on Red Hat's page.

Contains Red Hat Security Data © Red Hat, Inc., licensed under CC BY 4.0. Source: https://access.redhat.com/security/data. Modified: reformatted. Red Hat's page for CVE-2026-116176.

Attack patterns (CAPEC)

  • CAPEC-664 Server Side Request Forgery Likelihood: High · Severity: High

The attack patterns MITRE's CAPEC relates to this record's weaknesses, as MITRE names them. Credits.

References

Check your stackJSON