CVE-2026-54234

Public on

Last Modified: UTC

Important Important severity What does this mean?
7.5 CVSS v3 Score Breakdown

Description

A flaw was found in vLLM, a high-throughput and memory-efficient inference and serving engine for Large Language Models (LLMs). A remote attacker can exploit this vulnerability by sending a specially crafted multi-request speculative decoding workload through public gRPC Generate and Abort endpoints. This malformed workload can cause the rejection sampler to produce an out-of-vocabulary token, which then crashes the engine worker. This leads to a service-wide Denial of Service (DoS) for all clients until the worker is restarted.

Statement

This Important denial of service flaw in vLLM, utilized by Red Hat AI Inference Server, Red Hat Enterprise Linux AI, and Red Hat OpenShift AI, allows a remote attacker to crash the engine worker. By sending a specially crafted multi-request speculative decoding workload to public gRPC endpoints, an attacker can trigger an out-of-vocabulary token, leading to a service-wide disruption for all connected clients.

Mitigation

To mitigate this issue, restrict network access to the vLLM inference engine's gRPC Generate and Abort endpoints. Configure firewall rules to limit incoming connections to trusted clients or internal networks only. This will prevent remote, unauthenticated attackers from sending malformed workloads and triggering a denial of service. If the service is exposed via a proxy or load balancer, ensure that access controls are in place at that layer.

Additional information

External References

Affected Packages and Issued Red Hat Security Errata

Unless explicitly stated as not affected, all previous versions of packages in any minor update stream of a product listed here should be assumed vulnerable, although may not have been subject to full analysis.

Products / Services Components State Errata
Red Hat AI Inference Server rhaii/vllm-cpu-rhel9 Affected
Red Hat AI Inference Server rhaii/vllm-gaudi-rhel9 Will not fix
Red Hat AI Inference Server rhaii/vllm-neuron-rhel9 Will not fix
Red Hat AI Inference Server rhaii/vllm-tpu-rhel9 Will not fix
Red Hat AI Inference Server rhaiis/vllm-cpu-rhel9 Will not fix
Red Hat AI Inference Server rhaiis/vllm-neuron-rhel9 Will not fix
Red Hat AI Inference Server rhaiis/vllm-tpu-rhel9 Will not fix
Red Hat AI Inference Server 3.2 rhaiis/vllm-cuda-rhel9 Fixed RHSA-2026:61627
Red Hat AI Inference Server 3.2 rhaiis/vllm-rocm-rhel9 Fixed RHSA-2026:61629
Red Hat AI Inference Server 3.3 rhaiis/vllm-spyre-rhel9 Fixed RHSA-2026:60363

Common Vulnerability Scoring System (CVSS) Score Details

Important note

CVSS scores for open source components depend on vendor-specific factors (e.g. version or build chain). Therefore, Red Hat's score and impact rating can be different from NVD and other vendors. Red Hat remains the authoritative CVE Naming Authorities (CNA) source for its products and services (see Red Hat classifications ).

CVSS v3 Score Breakdown

Red Hat NVD cve.org
CVSS v3 Base Score 7.5 N/A 7.5
Attack Vector Network N/A Network
Attack Complexity Low N/A Low
Privileges Required None N/A None
User Interaction None N/A None
Scope Unchanged N/A Unchanged
Confidentiality Impact None N/A None
Integrity Impact None N/A None
Availability Impact High N/A High

CVSS v3 Vector

Red Hat CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H

cve.org CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H

Frequently Asked Questions

Why is Red Hat's CVSS v3 score or Impact different from other vendors?

 For open source software shipped by multiple vendors, the CVSS base scores may vary for each vendor's version depending on the version they ship, how they ship it, the platform, and even how the software is compiled. This makes scoring of vulnerabilities difficult for third-party vulnerability databases such as NVD that only provide a single CVSS base score for each vulnerability. Red Hat scores reflect how a vulnerability affects our products specifically.

For more information, see https://access.redhat.com/solutions/762393.

My product is listed as "Under investigation" or "Affected", when will Red Hat release a fix for this vulnerability?
  • "Under investigation" doesn't necessarily mean that the product is affected by this vulnerability. It only means that our Analysis Team is still working on determining whether the product is affected and how it is affected.
  • "Affected" means that our Analysis Team has determined that this product is affected by this vulnerability and might release a fix to address this in the near future.

What can I do if my product is listed as "Will not fix"?

 A "will not fix" status means that a fix for an affected product version is not planned or not possible due to complexity, which may create additional risk.

Available options depend mostly on the Impact of the vulnerability and the current Life Cycle phase of your product. Overall, you have the following options:
  • Upgrade to a supported product version that includes a fix for this vulnerability (recommended).
  • Apply a mitigation (if one exists).
  • Open a This content is not included.support case to request a prioritization of releasing a fix for this vulnerability.

What can I do if my product is listed as "Fix deferred"?

 A deferred status means that a fix for an affected product version is not guaranteed due to higher-priority development work.

Available options depend mostly on the Impact of the vulnerability and the current Life Cycle phase of your product. Overall, you have the following options:
  • Apply a mitigation (if one exists).
  • Open a This content is not included.support case to request a prioritization of releasing a fix for this vulnerability.
  • Red Hat Engineering focuses on addressing high-priority issues based on their complexity or limited lifecycle support. Therefore, lower-priority issues will not receive immediate fixes.

What is a mitigation?

 A mitigation is an action that can be taken to reduce the impact of a security vulnerability, without deploying any fixes.

I have a Red Hat product but it is not in the above list, is it affected?

 The listed products were found to include one or more of the components that this vulnerability affects. These products underwent a thorough evaluation to determine their affectedness by this vulnerability. Note that layered products (such as container-based offerings) that consume affected components from any of the products listed in this table may be affected and are not represented.

Why is my security scanner reporting my product as vulnerable to this vulnerability even though my product version is fixed or not affected?

 In order to maintain code stability and compatibility, Red Hat usually does not rebase packages to entirely new versions. Instead, we backport fixes and new features to an older version of the package we distribute. This can result in some security scanners that only consider the package version to report the package as vulnerable. To avoid this, we suggest that you use an approved vulnerability scanner from our This content is not included.Red Hat Vulnerability Scanner Certification program.

My product is listed as "Out of Support Scope". What does this mean?

 When a product is listed as "Out of Support Scope", it means a vulnerability with the impact level assigned to this CVE is no longer covered by its current support lifecycle phase. The product has been identified to contain the impacted component, but analysis to determine whether it is affected or not by this vulnerability was not performed. The product should be assumed to be affected. Customers are advised to apply any mitigation options documented on this page, consider removing or disabling the impacted component, or upgrade to a supported version of the product that has an update available.