vLLM is an inference and serving engine for large language models (LLMs). Prior to 0.22.1, the vLLM Dockerfile is vulnerable to a dependency confusion attack through the flashinfer-jit-cache package. The package is installed from a custom index (flashinfer.ai/whl/) using --extra-index-url, but the package name was not registered on PyPI, and UV_INDEX_STRATEGY="unsafe-best-match" is set globally. An attacker who registers flashinfer-jit-cache on PyPI with version 0.6.11.post2 can execute arbitrary code as root during the Docker build and backdoor every resulting container image, enabling exfiltration of all user prompts, API credentials, and model data from production vLLM deployments This vulnerability is fixed in 0.22.1.
Conclusion & alert: CVE-2026-54232 is rated High Exploit Risk (68.1/100): CVSS High severity, with medium exploitation likelihood (EPSS 0.27%). Core evidence: 1 public exploit reference(s) are indexed (Exploit-DB). Mandatory action: Public exploits are available—assess exposure, apply mitigations, and prioritize patching.
Risk is dynamic; we continuously reassess and refresh what is shown on this page as upstream context changes.
| EDB-ID | Source | Kind | Published | Link |
|---|---|---|---|---|
| — | nvd_ref | exploit_tag | Exploit-DB ↗ |
EPSS lead: Daily EPSS estimates relative likelihood of exploitation; percentile ranks this CVE among scored vulnerabilities (higher = more severe relative rank).
| # | Date | Old EPSS score | New EPSS score | Delta (New - Old) |
|---|---|---|---|---|
| 1 | 2026-06-13 | — | 0.27% | — |
Full EPSS history (1 record total)
CVSS metrics for this CVE.
| Base score | Version | Severity | Vector | Exploitability | Impact | Score source |
|---|---|---|---|---|---|---|
| 8.8 | 3.1 | HIGH |
|
2.8 | 5.9 | [email protected] |
| URL | Tags |
|---|---|
| https://github.com/vllm-project/vllm/security/advisories/GHSA-jrf6-vqxq-pjv2 | Exploit Third Party Advisory |