CVE-2026-22773— vLLM is vulnerable to DoS in Idefics3 vision models via image payload with ambiguous dimensions
Quick assessment
Affected
vllm-project vllm
Exploitation
No confirmed in-the-wild exploitation; assess based on exposure
Recommended action
Check the vendor advisory and references for a fixed version. If immediate upgrade is impossible, restrict exposure and increase monitoring.
vLLM是vLLM开源的一个适用于 LLM 的高吞吐量和内存高效推理和服务引擎。 vLLM 0.6.4版本至0.12.0之前版本存在安全漏洞，该漏洞源于发送特制的1x1像素图像会导致张量维度不匹配，可能导致服务器完全终止。
Possible ATT&CK Techniques 2 AI
T1059 · Command and Scripting Interpreter T1499 · Endpoint Denial of Service
Get alerts for future matching vulnerabilities
Log in to subscribe
I. Basic Information for CVE-2026-22773
Vulnerability Information
Have questions about the vulnerability? See if Shenlong's analysis helps!
View Shenlong Deep Dive ↗
Although we use advanced large model technology, its output may still contain inaccurate or outdated information.Shenlong tries to ensure data accuracy, but please verify and judge based on the actual situation.
Vulnerability Title
vLLM is vulnerable to DoS in Idefics3 vision models via image payload with ambiguous dimensions
Source: CVE Program / CVE List V5
Vulnerability Description
vLLM is an inference and serving engine for large language models (LLMs). In versions from 0.6.4 to before 0.12.0, users can crash the vLLM engine serving multimodal models that use the Idefics3 vision model implementation by sending a specially crafted 1x1 pixel image. This causes a tensor dimension mismatch that results in an unhandled runtime error, leading to complete server termination. This issue has been patched in version 0.12.0.
Source: CVE Program / CVE List V5
CVSS Information
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H
Source: CVE Program / CVE List V5
Vulnerability Type
不加限制或调节的资源分配
Source: CVE Program / CVE List V5
Vulnerability Title
vLLM 安全漏洞
Source: CNNVD (China National Vulnerability Database)
Vulnerability Description
vLLM是vLLM开源的一个适用于 LLM 的高吞吐量和内存高效推理和服务引擎。 vLLM 0.6.4版本至0.12.0之前版本存在安全漏洞，该漏洞源于发送特制的1x1像素图像会导致张量维度不匹配，可能导致服务器完全终止。
Source: CNNVD (China National Vulnerability Database)
CVSS Information
N/A
Source: CNNVD (China National Vulnerability Database)
Vulnerability Type
N/A
Source: CNNVD (China National Vulnerability Database)
Affected Products
| Vendor | Product | Affected Versions | CPE | Subscribe |
|---|---|---|---|---|
| vllm-project | vllm | >= 0.6.4, < 0.12.0 | - |
|
II. Public POCs for CVE-2026-22773
| # | POC Description | Source Link | Shenlong Link |
|---|
AI-Generated POC
Premium
No public POC found.
Login to generate AI POCIII. Intelligence Information for CVE-2026-22773
请登录查看更多情报信息。
Vendor Advisories for CVE-2026-22773 (1)
IV. Related Vulnerabilities
Same product: vllm
-
CVE-2026-105922
vllm-project vLLM Penalty utils.py get_token_bin_counts_and_ -
CVE-2026-105775
vllm-project vLLM Completions Request mamba_mixer2.py conv_s -
CVE-2026-105760
vLLM: GLMGA video sampling permits request-driven CPU and me -
CVE-2026-105759
vLLM: Unbounded Prometheus label cardinality from attacker-c -
CVE-2026-105758
vLLM: Qwen2-VL / Qwen3-VL video samplers bound on request-co
Same vendor: vllm-project
-
CVE-2026-105922
vllm-project vLLM Penalty utils.py get_token_bin_counts_and_ -
CVE-2026-105775
vllm-project vLLM Completions Request mamba_mixer2.py conv_s -
CVE-2026-105760
vLLM: GLMGA video sampling permits request-driven CPU and me -
CVE-2026-105759
vLLM: Unbounded Prometheus label cardinality from attacker-c -
CVE-2026-105758
vLLM: Qwen2-VL / Qwen3-VL video samplers bound on request-co
Same weakness: CWE-770
-
CVE-2026-107586
Allocation of Resources Without Limits or Throttling in hMai -
CVE-2026-107585
Allocation of Resources Without Limits or Throttling in hMai -
CVE-2026-16176
IBM DataPower Gateway Denial of Service -
CVE-2026-1403
Allocation of Resources Without Limits or Throttling in GitL -
CVE-2026-107212
Excelize: Unbounded row number in Rows.Columns makes GetRows
V. Comments for CVE-2026-22773
No comments yet