Skip to content

Security: AyaSlinc/TwinRoot

Security

SECURITY.md

Security Policy

Scope

TwinRoot operates on real disks, physical boot volumes, kernel filesystems, and guest boot chains. Security-relevant issues include, but are not limited to:

  • loss or corruption of user data on physical volumes;
  • unauthorized access to physical Windows or Linux installations;
  • privilege escalation through the host daemon, guest agents, or kernel module;
  • bypass of storage ownership or boot safety checks;
  • protocol forgery between host and guest components.

Reporting a Vulnerability

Please do not open a public GitHub issue for security vulnerabilities.

Report vulnerabilities privately to the TwinRoot maintainers by opening a GitHub Security Advisory on the repository, or by contacting the maintainers through the channels listed in the repository profile.

Include, where possible:

  • a description of the issue and its impact;
  • the components involved (host daemon, guest agent, TwinFS, hypervisor adapter, UI, protocol);
  • reproduction steps in a disposable environment;
  • suggested mitigation, if any.

You will receive an acknowledgement within a reasonable timeframe. Please allow time for assessment and coordination before any public disclosure.

Safe Disclosure

We credit responsible reporters when a fix is released, unless you prefer to remain anonymous.

Safety Philosophy

TwinRoot is designed to fail closed. Any operation involving a real disk, partition, filesystem, EFI environment, or boot volume must refuse to continue when safety cannot be verified. Refusing to start a VM is always preferable to risking data corruption. See AGENTS.md, section 10 "Storage Operations Must Fail Closed".

Unsupported Configurations

The following are explicitly unsupported and may cause data loss:

  • Windows Fast Startup or hibernation during TwinRoot transitions;
  • simultaneous block-level ownership of the same volume by two drivers;
  • using TwinRoot components with irreplaceable data before they are declared stable.

See the public documentation for the current project status.

There aren't any published security advisories