Skip to content

Security: hate/keyless

Security

SECURITY.md

Security Policy

Supported Versions

We actively support security updates for the latest stable release. Please update to the latest version to ensure you have the latest security patches.

Version Supported
0.3.x
0.2.x
0.1.x
< 0.1.0

For Users

  • Always download releases from the official GitHub releases page
  • Verify checksums when available
  • Keep your system and dependencies updated
  • Review permissions requested by the application (microphone access, etc.)

For Developers

  • Follow the Contributing Guide
  • Run security checks before submitting PRs: cargo audit and cargo deny check
  • Never commit secrets, API keys, or tokens
  • Use environment variables for sensitive configuration

Security Considerations

Privacy-First Architecture

keyless is designed with privacy as a core principle:

  • 100% local processing - All audio and transcription stays on your device
  • No network required - After initial model download, works completely offline
  • No telemetry - Zero tracking or analytics
  • Open source - Full code auditability

Data Handling

  • Audio is processed in memory and never persisted unless explicitly configured
  • Models are cached locally in ~/.cache/keyless/models/
  • Configuration is stored locally in OS-specific config directories
  • Log files may contain debugging information but no sensitive audio data

Permissions

keyless requires:

  • Microphone access - For audio capture (system-level permission)
  • Accessibility/Automation permissions - For paste output mode (macOS/Windows)

These permissions are necessary for core functionality and are clearly documented.

Security Audits

We use automated security scanning:

  • cargo audit - Checks against RustSec advisory database
  • cargo deny - License compliance and security checks
  • CI/CD runs these checks on every commit

To run locally:

cargo install cargo-audit cargo-deny --locked
cargo audit
cargo deny check

Known Limitations

  • Model downloads are not cryptographically verified (future improvement)

There aren’t any published security advisories