CVE-2026-73559: Uncontrolled Resource Consumption in vLLM API completions
Vulnerability ID: CVE-2026-73559
CVSS Score: 6.5
Published: 2026-08-13
CVE-2026-73559 is an uncontrolled resource consumption vulnerability in the vLLM engine, specifically within the /v1/completions API endpoint, allowing authenticated attackers to cause application-level denial of service via unbounded prompt arrays.
TL;DR
vLLM is vulnerable to memory and queue exhaustion via unbounded prompt arrays in the completions API, fixed in v0.26.0.
Technical Details
- CWE ID: CWE-400
- Attack Vector: Network
- CVSS v3.1: 6.5
- Impact: Denial of Service (DoS)
- Exploit Status: PoC / Theoretical
Affected Systems
- vLLM Engine
-
vllm: >= 0.19.0, < 0.26.0 (Fixed in:
0.26.0)
Code Analysis
Commit: 675f429
Add prompt list length validation in completions endpoint using environment variable VLLM_MAX_COMPLETION_PROMPTS
Mitigation Strategies
- Upgrade vLLM to version 0.26.0 or later
- Implement request size filtering at the API Gateway level
- Configure VLLM_MAX_COMPLETION_PROMPTS to a low threshold
Remediation Steps:
- Verify current vLLM version using pip show vllm
- Upgrade to the patched release via pip install --upgrade vllm>=0.26.0
- Restart the vLLM service process to apply changes
References
Read the full report for CVE-2026-73559 on our website for more details including interactive diagrams and full exploit analysis.
Top comments (0)