Bastion is a security tool, so bugs in Bastion itself are taken seriously.
Please do not open a public issue for a vulnerability.
Use GitHub's private reporting instead: open the repository's Security tab and choose Report a vulnerability. Include:
- the Bastion version and your OS (Windows / macOS, version, CPU architecture),
- what an attacker can do and what they need (local user, same network, malicious web page, malicious AI response, ...),
- steps to reproduce or a proof of concept,
- any suggested fix.
You should get an acknowledgement within 7 days. We will keep you posted while a fix is prepared and credit you in the release notes unless you prefer not to be named.
Only the latest release receives security fixes. Bastion is pre-1.0; please update before reporting.
- Anything that lets a web page, a network peer, a router response or an AI model response run code, read files, or call privileged IPC in the app.
- Leaks of stored AI API keys (keys are encrypted at rest with the operating system keychain via
Electron
safeStorage) or of network data to any host other than the ones listed in the README's Privacy section. - Bastion changing a setting on your computer or router without your explicit approval.
- Command injection through values parsed from system tools (
netsh,ipconfig,arp,airport, ...). - Checks that report "OK" when the risk is demonstrably present (false reassurance).
- Vulnerabilities in your router, ISP equipment or other devices. Bastion helps you find some of them; please report those to the vendor.
- Findings that require an already compromised machine or administrator access.
- Unsigned builds triggering SmartScreen or Gatekeeper warnings (signed builds are on the roadmap).
- The renderer runs sandboxed with context isolation and no Node access; the only bridge is a small,
typed
window.bastionAPI (seedocs/CONTRACT.md). - System tools are run with
execFileand fixed argument lists, never through a shell string. - Bastion never logs in to your router, never tries passwords and never changes settings on its own. AI agents can only answer and suggest steps; they cannot change anything themselves.