Skip to content

Security: brfid/edcloud

Security

SECURITY.md

Security policy

Security model

edcloud is a single-operator personal lab, not a multi-tenant platform.

Core assumptions:

  • Access is Tailscale-only.
  • The EC2 security group has no inbound rules.
  • The operator controls AWS and Tailscale identities.
  • Workloads are trusted by the operator.

What this project is designed to prevent

  • Public SSH exposure
  • Public exposure of Portainer or workload ports (Portainer binds to Tailscale interface only)
  • IMDSv1 usage (IMDSv2 is required, hop limit set to 1)
  • Avoidable idle spend (automatic idle shutdown)
  • Credentials in user-data (auth keys fetched from SSM at boot)

What this project does not try to prevent

  • Compromise of your AWS or Tailscale account
  • Malicious or vulnerable containers you choose to run
  • Physical compromise of devices in your tailnet
  • Multi-user isolation and tenant-level access control
  • Docker socket exposure in Portainer (accepted risk for single-operator convenience)

Required operator practices

  • Keep runtime secrets in AWS SSM Parameter Store.
  • Do not commit credentials, keys, or tokens to git.
  • Use MFA on AWS and your identity provider.
  • Rotate Tailscale auth keys and remove unused devices.
  • Keep AWS DLM backup policy enabled and review edc backup-policy status periodically.
  • Run restore drills and validate backup recovery.

Vulnerability reporting

Do not open public issues for security vulnerabilities.

Report privately to @brfid on GitHub and include:

  • A clear description
  • Reproduction steps
  • Expected impact
  • Suggested remediation (optional)

Response targets:

  • Initial acknowledgment: within 7 days
  • Fix priority: based on impact and exploitability
  • Public disclosure: after a fix is available

Supported code line

Security fixes are applied on main.

Dependency scope

Security issues in upstream dependencies (AWS, Ubuntu, Docker, Tailscale, Portainer) should also be reported to the relevant maintainers.

There aren’t any published security advisories