EUVD-2025-208974
GHSA-qm3j-hh5g-jmq2
Severity by source
AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
Primary rating from NVD · only source for this CVE.
CVSS VectorNVD
CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
Lifecycle Timeline
3DescriptionCVE.org
NVIDIA Megatron LM contains a vulnerability in quantization configuration loading, which could allow remote code execution. A successful exploit of this vulnerability might lead to code execution, escalation of privileges, information disclosure, and data tampering.
AnalysisAI
NVIDIA Megatron LM contains an insecure deserialization vulnerability (CWE-502) in its quantization configuration loading mechanism that enables remote code execution. Attackers with local access and low privileges can exploit this flaw to execute arbitrary code, escalate privileges, disclose sensitive information, and tamper with data. The vulnerability has a CVSS score of 7.8 and affects all versions of NVIDIA Megatron LM based on available CPE data.
Technical ContextAI
NVIDIA Megatron LM is a large-scale transformer model training framework used for natural language processing tasks. The vulnerability stems from unsafe deserialization (CWE-502) during the loading of quantization configuration files. Insecure deserialization occurs when untrusted data is used to reconstruct objects without proper validation, allowing attackers to inject malicious serialized objects that execute arbitrary code when deserialized. The affected product is identified via CPE as cpe:2.3:a:nvidia:megatron_lm:*:*:*:*:*:*:*:*, indicating all current versions are vulnerable. Quantization is a model compression technique, and the configuration loading process for this feature appears to lack input validation, creating an attack surface where malicious configuration files can be crafted to exploit the deserialization mechanism.
RemediationAI
Consult the official NVIDIA security advisory at https://nvidia.custhelp.com/app/answers/detail/a_id/5769 for patches and updated versions of Megatron LM that address this deserialization vulnerability. Until patching is completed, implement defense-in-depth controls including restricting access to quantization configuration files through file system permissions, validating and sanitizing all configuration inputs before processing, limiting local system access to only trusted users, and monitoring for suspicious configuration file modifications or unexpected process behavior. Organizations should also review access controls on systems running Megatron LM and implement principle of least privilege to reduce the attack surface, as the vulnerability requires local authenticated access to exploit.
More from same product – last 7 days
Arbitrary file read leading to remote code execution affects Langflow versions prior to 1.9.2 in any flow that uses Base
Unauthenticated remote code execution in NVIDIA Spatial Intelligence Lab's GEN3C inference API server allows network att
Code injection in NVIDIA NeMo Framework across all supported platforms allows a local attacker with low privileges to ex
Local code execution in NVIDIA NeMo Framework on Linux allows an authenticated low-privileged attacker to abuse unsafe d
Share
External POC / Exploit Code
Leaving vuln.today