CVE-2026-114055 · GHSA-gx7p-2j49-hfq4
vLLM : DoS via unvalidated mm_processor_kwargs in multimodal image processors (Isaac, InternVL, H2OVL, etc.)
- Risk score
- 60/100
- CVSS
- 7.5 v3.1
- Exploit chance (30 days)
- 0.3%
- Published
- 21 Oct 2026
Open in the live forecast Today's forecast
What to know
- What it is
- vLLM from 0.7.3 before 0.31.0 contains an uncontrolled resource consumption vulnerability that allows remote clients to control multimodal processor parameters by supplying per-request mm_processor_kwargs. Attackers can set large image size, patch count, crop count, or frame count values in Isaac, InternVL, H2OVL, and…
- Affected
- vllm (pip) >= 0.7.3, < 0.31.0
- Fix
- Upgrade vllm to 0.31.0 or later.
- Exploitation
- No known exploitation
Story map
At a glance
vllm: Amber, patch this week. vLLM from 0.7.3 before 0.31.0 contains an uncontrolled resource consumption vulnerability that allows remote clients to control multimodal processor parameters by supplying per-request mm_processor_kwargs. Attackers can set large image size, patch count, crop count, or frame count values in Isaac, InternVL, H2OVL, and…
The flaws
- CVE-2026-114055: VLLM : DoS via unvalidated mm_processor_kwargs in multimodal image processors (Isaac, InternVL, H2OVL, etc.) (CVSS 7.5 · EPSS 0%)
- CVE-2026-114050: VLLM: Forged multimodal UUID + P0/P1 processor-cache eviction drift: unauthenticated remote crash of the whole vLLM engine (CVSS 7.5 · EPSS 0%)
- CVE-2026-73557: VLLM: Incomplete CVE-2025-62164 remediation can be bypassed by concurrent prompt parts (CVSS 6.3)
The fix
Upgrade vllm: 0.31.0 or later on your release line
Who is exposed: vllm (pip) >= 0.7.3, < 0.31.0
Where it stands
- No known exploitation
Connects to
- Same weakness: Denial of service (CWE-400). 256 other warnings in 30 days
Why it is Amber
- CVSS 7.5 high
- Remotely exploitable without logging in
- On your watchlist: PyPI
Affected packages
| Package | Ecosystem | Vulnerable | Fixed in |
|---|---|---|---|
| vllm | pip | >= 0.7.3, < 0.31.0 | 0.31.0 |
References
- https://github.com/vllm-project/vllm/security/advisories/GHSA-gx7p-2j49-hfq4
- https://www.vulncheck.com/advisories/vllm-0.7.3-before-0.31.0-dos-via-unvalidated-mm-processor-kwargs
- https://github.com/advisories/GHSA-gx7p-2j49-hfq4
- https://nvd.nist.gov/vuln/detail/CVE-2026-114055
- https://github.com/vllm-project/vllm/pull/58830
- https://github.com/vllm-project/vllm/commit/31f2e70cd3e403aaa471fb7fe3d33a7207aaf3cc
- https://github.com/vllm-project/vllm/releases/tag/v0.31.0
- https://github.com/advisories/GHSA-gx7p-2j49-hfq4
Original description
vLLM from 0.7.3 before 0.31.0 contains an uncontrolled resource consumption vulnerability that allows remote clients to control multimodal processor parameters by supplying per-request mm_processor_kwargs. Attackers can set large image size, patch count, crop count, or frame count values in Isaac, InternVL, H2OVL, and other processors to exhaust memory and terminate the shared server.
Last activity 23 Oct 2026. Patch Forecast updates every hour.