CVE-2022-0845

The code injection vulnerability in PyTorch Lightning arises from insufficient input validation within the repository. An attacker can craft malicious code and inject it into the system, leading to remote code execution (RCE). Successful exploitation could allow an attacker to gain full control over the affected system, including access to sensitive data, modification of system configurations, and installation of malware. The potential blast radius is significant, particularly in environments where PyTorch Lightning is used for training and deploying machine learning models, as attackers could compromise the entire training pipeline and potentially inject malicious models into production.

Organizations and individuals utilizing PyTorch Lightning for machine learning model training and deployment are at risk, particularly those using older versions (≤1.5.10.post0). This includes researchers, data scientists, and DevOps engineers working with PyTorch-based projects. Shared hosting environments where PyTorch Lightning is deployed could also be vulnerable if multiple users share the same environment and one user can inject malicious code.

• python / supply-chain:

import subprocess
result = subprocess.run(['pip', 'show', 'pytorch-lightning'], capture_output=True, text=True)
if 'Version' in result.stdout and result.stdout.splitlines()[2].startswith('1.5.'): 
    print('Vulnerable version detected!')

• python / server: Review PyTorch Lightning configuration files for any unusual or unexpected code snippets. • generic web: Inspect PyTorch Lightning model deployment pipelines for potential injection points.

1. 2022-03-05Disclosure

   disclosure

1. Reserviert2022-03-03
2. Veröffentlicht2022-03-05
3. Geändert2025-10-09
4. EPSS aktualisiert2026-04-02

The primary mitigation for CVE-2022-0845 is to immediately upgrade PyTorch Lightning to version 1.6.0 or later. If an immediate upgrade is not feasible due to compatibility issues or breaking changes, consider implementing stricter input validation on any user-provided data used within PyTorch Lightning workflows. While a direct WAF rule is unlikely to be effective, carefully reviewing and sanitizing any external data passed to PyTorch Lightning models can reduce the attack surface. Monitor PyTorch Lightning repositories for suspicious activity and review commit history for potentially malicious code.