Security Model

Security Model

Condensed from the repo's SECURITY.md — read that file for the full version, including code examples.

What VoidGuard will never do

• Arbitrary shell / cmd.exe / PowerShell execution
• Accept a free-form command string from the UI and spawn it
• Remote control, remote shell, or covert channel of any kind
• Hidden/undocumented persistence
• Credential collection, keystroke logging, clipboard scraping
• Disable or bypass Windows Defender or other security software
• Hide its own process, files, or network activity

How privileged actions work

Every mutating action (kill a process, open/close a port, later: change DNS, edit firewall rules, start/stop services) is:

1. A named, typed Tauri command — never a raw string sent to the OS.
2. Confirmed explicitly in the UI before the backend is even called, for anything destructive.
3. Validated again server-side (never trust the frontend).
4. Executed via a native Windows API or COM interface — no netsh, no shell.
5. Recorded in the append-only audit_logs table, success or failure, with before/after state.

Detection philosophy

Unknown or unsigned software is reported with evidence and a confidence level — never auto-labeled "malware," auto-quarantined, or auto-deleted. All remediation is a suggestion the user acts on explicitly.

Elevation

The app runs as a standard user by default. Operations that need elevation trigger a scoped Windows elevation prompt for that single operation — VoidGuard does not hold a permanent elevated token.

Reporting a concern

Use the Security concern issue template in the repo, or see CONTRIBUTING.md for private-disclosure guidance on live vulnerabilities.