What is the CVSS score of CVE-2025-46722?

CVE-2025-46722 has a CVSS 3.1 base score of 4.2 (Medium). CVSS vector: CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:L/I:N/A:L. EPSS exploitation probability: 0.2%.

Which versions of Vllm are affected by CVE-2025-46722?

Organizations using versions starting from 0.7.0 to should be aware of moderate risk from CVE-2025-46722 rated CVSS 4.2. Exploitation could compromise the security of affected deployments.. A patch is available.

How to fix or mitigate CVE-2025-46722?

During next maintenance window: Identify affected systems running versions starting from 0.7.0 to and apply vendor patches when convenient. Vendor patch is available.

Vllm CVE-2025-46722

MEDIUM

Incomplete Comparison with Missing Factors (CWE-1023)

2025-05-29 security-advisories@github.com

Information Disclosure Red Hat Vllm

4.2

CVSS 3.1

CVSS VectorNVD

CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:L/I:N/A:L

Attack Vector

Network

Attack Complexity

High

Privileges Required

Low

User Interaction

None

Scope

Unchanged

Confidentiality

Low

Integrity

None

Availability

Low

Lifecycle Timeline

Analysis Generated

Mar 28, 2026 - 18:44 vuln.today

Patch released

Mar 28, 2026 - 18:44 nvd

Patch available

CVE Published

May 29, 2025 - 17:15 nvd

MEDIUM 4.2

Blast Radius

ecosystem impact

† from your stack dependencies † transitive graph · vuln.today resolves 4-path depth

4 pypi packages depend on vllm (3 direct, 1 indirect)

Ecosystem-wide dependent count for version 0.7.0.

DescriptionNVD

vLLM is an inference and serving engine for large language models (LLMs). In versions starting from 0.7.0 to before 0.9.0, in the file vllm/multimodal/hasher.py, the MultiModalHasher class has a security and data integrity issue in its image hashing method. Currently, it serializes PIL.Image.Image objects using only obj.tobytes(), which returns only the raw pixel data, without including metadata such as the image’s shape (width, height, mode). As a result, two images of different sizes (e.g., 30x100 and 100x30) with the same pixel byte sequence could generate the same hash value. This may lead to hash collisions, incorrect cache hits, and even data leakage or security risks. This issue has been patched in version 0.9.0.

AnalysisAI

vLLM is an inference and serving engine for large language models (LLMs). Rated medium severity (CVSS 4.2), this vulnerability is remotely exploitable.

Technical ContextAI

This vulnerability is classified under CWE-1023. vLLM is an inference and serving engine for large language models (LLMs). In versions starting from 0.7.0 to before 0.9.0, in the file vllm/multimodal/hasher.py, the MultiModalHasher class has a security and data integrity issue in its image hashing method. Currently, it serializes PIL.Image.Image objects using only obj.tobytes(), which returns only the raw pixel data, without including metadata such as the image’s shape (width, height, mode). As a result, two images of different sizes (e.g., 30x100 and 100x30) with the same pixel byte sequence could generate the same hash value. This may lead to hash collisions, incorrect cache hits, and even data leakage or security risks. This issue has been patched in version 0.9.0. Affected products include: Vllm. Version information: before 0.9.0.

RemediationAI

A vendor patch is available. Apply the latest security update as soon as possible. Apply vendor patches when available. Implement network segmentation and monitoring as interim mitigations.