CVE-2025-50461

A deserialization vulnerability exists in Volcengine's verl 3.0.0, specifically in the scripts/model_merger.py script when using the "fsdp" backend. The script calls torch.load() with weights_only=False on user-supplied .pt files, allowing attackers to execute arbitrary code if a maliciously crafted model file is loaded. An attacker can exploit this by convincing a victim to download and place a malicious model file in a local directory with a specific filename pattern. This vulnerability may lead to arbitrary code execution with the privileges of the user running the script.

CVSS v3 6.5 MEDIUM

6.5^/10

CVSS v3 : MEDIUM

Vector :

Exploitability : 3.9 / Impact : 2.5

Attack Vector NETWORK

Attack Complexity LOW

Privileges Required NONE

User Interaction NONE

Confidentiality Impact LOW

Integrity Impact LOW

Availability Impact NONE

Scope UNCHANGED

References

Link	Resource
https://github.com/Anchor0221/CVE-2025-50461
https://github.com/pytorch/pytorch/blob/main/SECURITY.md#loading-untrusted-data
https://github.com/volcengine/verl
https://github.com/volcengine/verl/blob/main/scripts/model_merger.py#L152
https://pytorch.org/docs/stable/generated/torch.load.html

Configurations

No configuration.

History

No history.

Information

Published : 2025-08-19 14:15

Updated : 2025-08-20 14:40

NVD link : CVE-2025-50461

Mitre link : CVE-2025-50461

CVE.ORG link : CVE-2025-50461

JSON object : View

Products Affected

No product.

CWE

CWE-77

Improper Neutralization of Special Elements used in a Command ('Command Injection')

{"id": "CVE-2025-50461", "cveTags": [], "metrics": {"cvssMetricV31": [{"type": "Secondary", "source": "134c704f-9b21-4f2e-91b3-4a467353bcc0", "cvssData": {"scope": "UNCHANGED", "version": "3.1", "baseScore": 6.5, "attackVector": "NETWORK", "baseSeverity": "MEDIUM", "vectorString": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:N", "integrityImpact": "LOW", "userInteraction": "NONE", "attackComplexity": "LOW", "availabilityImpact": "NONE", "privilegesRequired": "NONE", "confidentialityImpact": "LOW"}, "impactScore": 2.5, "exploitabilityScore": 3.9}]}, "published": "2025-08-19T14:15:39.533", "references": [{"url": "https://github.com/Anchor0221/CVE-2025-50461", "source": "cve@mitre.org"}, {"url": "https://github.com/pytorch/pytorch/blob/main/SECURITY.md#loading-untrusted-data", "source": "cve@mitre.org"}, {"url": "https://github.com/volcengine/verl", "source": "cve@mitre.org"}, {"url": "https://github.com/volcengine/verl/blob/main/scripts/model_merger.py#L152", "source": "cve@mitre.org"}, {"url": "https://pytorch.org/docs/stable/generated/torch.load.html", "source": "cve@mitre.org"}], "vulnStatus": "Awaiting Analysis", "weaknesses": [{"type": "Secondary", "source": "134c704f-9b21-4f2e-91b3-4a467353bcc0", "description": [{"lang": "en", "value": "CWE-77"}]}], "descriptions": [{"lang": "en", "value": "A deserialization vulnerability exists in Volcengine's verl 3.0.0, specifically in the scripts/model_merger.py script when using the \"fsdp\" backend. The script calls torch.load() with weights_only=False on user-supplied .pt files, allowing attackers to execute arbitrary code if a maliciously crafted model file is loaded. An attacker can exploit this by convincing a victim to download and place a malicious model file in a local directory with a specific filename pattern. This vulnerability may lead to arbitrary code execution with the privileges of the user running the script."}, {"lang": "es", "value": "Existe una vulnerabilidad de deserializaci\u00f3n en la versi\u00f3n 3.0.0 de Volcengine, espec\u00edficamente en el script scripts/model_merger.py al usar el backend \"fsdp\". El script llama a torch.load() con weights_only=False en archivos .pt proporcionados por el usuario, lo que permite a los atacantes ejecutar c\u00f3digo arbitrario si se carga un archivo de modelo malicioso. Un atacante puede explotar esto convenciendo a la v\u00edctima para que descargue y coloque un archivo de modelo malicioso en un directorio local con un patr\u00f3n de nombre de archivo espec\u00edfico. Esta vulnerabilidad puede provocar la ejecuci\u00f3n de c\u00f3digo arbitrario con los privilegios del usuario que ejecuta el script."}], "lastModified": "2025-08-20T14:40:17.713", "sourceIdentifier": "cve@mitre.org"}