CVE-2026-105759: CWE-400: Uncontrolled Resource Consumption in vllm-project vllm
Description
CVE-2026-105759 is an uncontrolled resource consumption vulnerability in the vllm-project's vllm inference engine for large language models. Versions prior to 0.30.0 have a flaw in the Rust frontend's track_http_metrics middleware, which records raw HTTP method tokens as Prometheus labels. An unauthenticated attacker can exploit this by sending arbitrary HTTP method tokens to unprotected routes, causing unbounded creation of Prometheus metric label sets. This leads to increased memory usage and enlarged /metrics responses, potentially exhausting service or monitoring resources. The issue is fixed in version 0.30.0.
CVSS v3.1
Score 5.9medium
Affected software
vllm-project
vllm
pkg:cargo/github/vllm-project/vllmRun on your own infrastructure? Check whether these packages are installed with threat-finder — our free open-source scanner.
Weaknesses
AI-Powered Analysis
Machine-generated threat intelligence
Technical Analysis
The vulnerability arises because the track_http_metrics middleware in vllm versions before 0.30.0 records raw HTTP method tokens as Prometheus label values without validation or limitation. An attacker can send many unique HTTP method tokens to routes like /tokenize, triggering Prometheus's Family::get_or_create function to create permanent counter and histogram label sets for each unique token. This uncontrolled growth in label sets causes increased memory consumption and bloated /metrics responses, which can exhaust the service or monitoring infrastructure. The vulnerability is classified as CWE-400 (Uncontrolled Resource Consumption) and has a CVSS 3.1 base score of 5.9 (medium severity). The issue is resolved in vllm version 0.30.0.
Potential Impact
Exploitation of this vulnerability results in denial of service conditions due to resource exhaustion. Specifically, the process memory usage grows as Prometheus metric label sets accumulate without limit, and the /metrics endpoint response size increases, potentially exhausting service or monitoring resources. There is no impact on confidentiality or integrity, only availability is affected.
Mitigation Recommendations
Upgrade to vllm version 0.30.0 or later, where this issue is fixed. Prior versions are vulnerable. No other mitigations or workarounds are indicated in the available data.
Technical Details
- Data Version
- 5.2
- Assigner Short Name
- GitHub_M
- Date Reserved
- 2026-10-05T19:11:07.948Z
- Cvss Version
- 3.1
- State
- PUBLISHED
Threat ID: 6ac430632cdf04f65645c895
Added to database: 10/05/2026, 23:18:59 UTC
Last enriched: 10/05/2026, 23:33:19 UTC
Last updated: 10/05/2026, 23:33:54 UTC
Views: 5
Community Reviews
0 reviewsCrowdsource mitigation strategies, share intel context, and vote on the most helpful responses. Sign in to add your voice and help keep defenders ahead.
Want to contribute mitigation steps or threat intel context? Sign in or create an account to join the community discussion.
Actions
Updates to AI analysis require Pro Console access. Upgrade inside Console → Billing.
Need more coverage?
Upgrade to Pro Console for AI refresh and higher limits.
For incident response and remediation, OffSeq services can help resolve threats faster.
Latest Threats
Check if your credentials are on the dark web
Instant breach scanning across billions of leaked records. Free tier available.