Vulnerability in vllm (CVE-2026-54234)
Summary
Vulnerability in vllm (CVE-2026-54234). Risk of unauthorized operations or information disclosure. Exploitable via vocab_size. Mitigation: upgrade to 0.24.0 or later.
Original description
The description as published by the source.
vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Prior to 0.24.0, a frontend-legal multi-request speculative decoding workload can cause the rejection sampler to produce a recovered token equal to the model vocabulary size boundary value, which is then converted to negative one when the engine selects the next live token for a request and is written back into the drafter's input ids; that out-of-vocabulary value is later consumed by the model's embedding and attention path and crashes the engine worker with a GPU device-side assertion. The same triggering request sequence is reachable through the public gRPC Generate and Abort endpoints, so a remote client that can send generation requests can crash the shared engine worker, aborting concurrent requests and causing a service-wide denial of service for other clients of the deployment until the worker is restarted. This issue is fixed in version 0.24.0.
Source: NIST National Vulnerability Database NVD GitHub Advisory
Auto-generated summary
Assembled by SecPulse from the source data (CVSS, CWE, affected versions) using templates — no AI involved. Check the original description and references below for exact details.
A vulnerability tracked as CVE-2026-54234 has been found in vllm.
Attackers can target a specific entry point like vocab_size over the network to misuse the product.
Risk of unauthorized operations or information disclosure. CVSS score: 7.5/10.
What to do: upgrade vllm to 0.24.0 or later.
If unsure, ask your IT team or search "vllm CVE-2026-54234" on the vendor's site.
CVE-2026-54234 (vllm) — CWE-20 / CVSS v3 7.5
Attack vector: remote (network-reachable) / unauthenticated / no user interaction
Attack surface: vocab_size / Generate / Abort / SamplingParams
Patched: 0.24.0 — apply immediately
Plan: 1) Audit SBOM/dependencies, 2) Stage→prod upgrade, 3) Add WAF/proxy monitoring on affected endpoints, 4) Hunt IOCs in logs.
Refs: see the GHSA / vendor advisory / patched release linked on this page.
A vulnerability (CWE-20) exists in vllm. Attackers reach the vulnerable code path via vocab_size without authentication.
Check the vendor advisory for the affected versions of vllm. Attack surface: vocab_size / Generate / Abort / SamplingParams.
Severity: High (CVSS 7.5/10). Risk of unauthorized operations or information disclosure.
Update to 0.24.0.
Until the patch is applied: disable the affected feature, apply WAF rules, or restrict access via network ACLs.
Search webserver/proxy logs for unusual request patterns matching this CVE's known IOCs. Run grep -r 'vllm' . against your dependency files (package-lock.json, requirements.txt, go.sum) to find affected services.
Response actions
Concrete steps and command examples for SOC/SRE teams to execute in order
-
1Identify exposure identify
grep -r 'vllm' . | grep -v node_modulesリポジトリと本番環境の依存ファイル (package-lock.json / requirements.txt / go.sum / Gemfile.lock 等) で
vllmを grep し、稼働しているサービス・バージョンを把握する。 -
2Apply patch patch
Upgrade vllm to 0.24.0ステージング環境で 0.24.0 に上げて回帰テスト → 本番反映。回帰テストはアプリの主要ハッピーパスを含めること。
-
3Post-deployment verification verify
Confirm patched version is live in productionパッチ適用後、ステージングで PoC または同等の悪用パターンを再現して脆弱性が閉じたことを確認。本番ではアクセスログ・WAF でアラート再発が無いか継続監視。
Affected packages
References
- advisory github.com
- advisory nvd.nist.gov
- advisory github.com
- advisory github.com
- package github.com
- package pypi.org
- patch github.com
- patch github.com