Please report security issues privately using the vulnerability submission form.

Reports will be triaged by the vLLM project vulnerability management team.

LLM Compressor is a tool for compressing large language models through quantization and sparsification. Security considerations include:

• Model integrity: Ensuring compressed models maintain expected behavior
• Supply chain security: Safe handling of model weights and checkpoints
• Deployment security: Secure integration with vLLM and other inference engines

Please see the vLLM Security Guide for broader security assumptions and recommendations when deploying compressed models.

Please see PyTorch's Security Policy for recommendations on how to securely interact with models.

We determine the risk of each issue using the following severity categories:

Vulnerabilities that allow remote attackers to execute arbitrary code, take full control of the system, or significantly compromise confidentiality, integrity, or availability without any interaction or privileges needed. Generally those issues rated as CVSS ≥ 9.0.

Serious security flaws that allow elevated impact—like RCE in specific, limited contexts or significant data loss—but require advanced conditions or some trust. These issues typically have CVSS scores between 7.0 and 8.9.

Vulnerabilities that cause denial of service or partial disruption, but do not allow arbitrary code execution or data breach and have limited impact. These issues have a CVSS rating between 4.0 and 6.9.

Minor issues such as informational disclosures, logging errors, non-exploitable flaws, or weaknesses that require local or high-privilege access and offer negligible impact. These issues often have CVSS scores less than 4.0.

When a security report is accepted, the fix process depends on the severity:

• CRITICAL and HIGH severity: Fixes are developed in a private security fork and coordinated with the prenotification group before public disclosure.
• MODERATE and LOW severity: Fixes are developed and submitted as public pull requests. These issues do not require embargo since they do not enable arbitrary code execution or significant data breach, and public visibility accelerates community review and adoption of the fix.

The vulnerability management team reserves the right to adjust the disclosure approach on a case-by-case basis, taking into account factors such as active exploitation, unusual attack surface, or coordination requirements with downstream vendors.

For certain security issues of CRITICAL, HIGH, or MODERATE severity level, we may prenotify certain organizations or vendors that ship LLM Compressor or vLLM. The purpose of this prenotification is to allow for a coordinated release of fixes for severe issues.

• This prenotification will be in the form of a private email notification. It may also include adding security contacts to the GitHub security advisory, typically a few days before release.

• If you wish to be added to the prenotification group, please send an email copying all the members of the vulnerability management team. Each vendor contact will be analyzed on a case-by-case basis.

• Organizations and vendors who either ship or use LLM Compressor are eligible to join the prenotification group if they meet at least one of the following qualifications:

  • Substantial internal deployment leveraging the upstream LLM Compressor project.
  • Established internal security teams and comprehensive compliance measures.
  • Active and consistent contributions to the upstream LLM Compressor project.

• We may withdraw organizations from receiving future prenotifications if they release fixes or any other information about issues before they are public. Group membership may also change based on policy refinements for who may be included.