Skip to content

Security: ROCm/Quartz

Security

SECURITY.md

Security Policy

Reporting a Vulnerability

Do not open a public GitHub issue. Report privately via one of:

Please include: description and impact, steps to reproduce, and affected versions or commits.

We aim to acknowledge reports within 1 business day.

Scope

This policy covers code and configuration in this repository. For vulnerabilities in third-party dependencies, report upstream. For AMD product issues unrelated to this repo, use the AMD Product Security portal.

Automated security scanning

Alongside the reporting path above, this repository is scanned automatically. The scanners are not implemented here: Quartz calls the shared ROCm/rocm-security-gh security-baseline.yml reusable workflow, so the scanner versions and behavior are maintained centrally for ROCm, and this repository supplies only its own configuration (the *.yml / *.toml files at the repo root).

Scanner Looks for Configuration
gitleaks Secrets and credentials in tracked files and git history gitleaks.toml
bandit Unsafe patterns in our Python scripts bandit.yml
zizmor GitHub Actions workflow vulnerabilities zizmor.yml
trivy Dependency vulnerabilities and misconfigurations trivy.yml
CodeQL Semantic code analysis of our Python org-wide default (no local override)

Two workflows run them, and where a finding shows up depends on which one produced it:

  • security_scan_pr.yml runs on every pull request, scoped to what that pull request changed. It reports in the job summary and a build artifact, and deliberately does not upload to the Security tab, so pull requests from forks behave the same as those from branches in this repository.
  • security_scan_weekly.yml runs on a schedule across the whole repository and uploads SARIF to this repository's Security tab, which is the authoritative view of the current state. Quartz is a monorepo-adjacent, low-churn repository, so a weekly cadence (rather than on every push to develop/main) is enough to keep the Security tab current without adding a scan to every merge.

Important

A finding from these scanners is not a vulnerability report. If a scanner finding turns out to be an exploitable vulnerability in shipped ROCm software, report it through the AMD Product Security portal above rather than in a public issue or pull request.

Contributors can run every scanner locally against the same configuration CI uses; see the security scanners section in CONTRIBUTING.md.

There aren't any published security advisories