NVIDIA Megatron-LM für alle Plattformen enthält eine Schwachstelle in einem Skript, wobei von einem Angreifer erstellte bösartige Daten zu einem Code-Injection-Problem führen können. Ein erfolgreicher Exploit dieser Schwachstelle kann

Successful exploitation of CVE-2025-23357 allows an attacker to inject and execute arbitrary code within the Megatron-LM environment. This could lead to a complete compromise of the system running the framework, including unauthorized access to sensitive data used for model training. The attacker could potentially steal training data, modify model parameters, or even use the compromised system to launch further attacks. The blast radius extends to any data processed by Megatron-LM, and the potential for lateral movement depends on the system's configuration and network access.

Organizations and researchers utilizing NVIDIA Megatron-LM for large language model training are at risk. This includes those deploying Megatron-LM in cloud environments, on-premise servers, or within development pipelines. Specifically, those using older, unpatched versions of the framework are most vulnerable.

• python / supply-chain:

import os
import subprocess

def check_megatron_version():
    try:
        result = subprocess.check_output(['python', '-c', 'import megatron_lm; print(megatron_lm.__version__)'], stderr=subprocess.STDOUT)
        version = result.decode('utf-8').strip()
        if version and version.startswith('0.13'):
            print(f"Vulnerability detected: Megatron-LM version {version} is vulnerable.")
        else:
            print(f"Megatron-LM version {version} is not vulnerable.")
    except FileNotFoundError:
        print("Megatron-LM not found.")
    except subprocess.CalledProcessError as e:
        print(f"Error checking version: {e}")

check_megatron_version()

1. 2025-11-11Disclosure

   disclosure

1. Reserviert2025-01-14
2. Veröffentlicht2025-11-11
3. Geändert2025-11-14
4. EPSS aktualisiert2026-03-23

The primary mitigation for CVE-2025-23357 is to upgrade to NVIDIA Megatron-LM version 0.14.0 or later, which includes the necessary fix. If upgrading is not immediately feasible, carefully review all input data provided to Megatron-LM, implementing strict validation and sanitization routines to prevent the injection of malicious code. Consider implementing a Web Application Firewall (WAF) to filter potentially malicious requests. After upgrading, confirm the vulnerability is resolved by attempting to reproduce the attack scenario with known malicious input; it should now be rejected or handled safely.