Skip to content

Security: luohoa97/ramble-voice

Security

SECURITY.md

Security policy

Reporting a vulnerability

Please report vulnerabilities privately through GitHub private vulnerability reporting: open the Security tab of luohoa97/ramble-voice and choose Report a vulnerability. Please do not open a public issue for security problems.

Include the affected version, platform (Android or Linux, desktop and compositor), steps to reproduce, and the impact you expect. You will get an acknowledgement within a few days. This is a volunteer project, so fix times vary with severity.

Supported versions

Only the latest release receives security fixes while the project is pre-1.0.

Scope

In scope:

  • Prompt injection. Ramble Voice is text-only. Whatever you say is transcribed and cleaned up; it must never be executed, obeyed, or turned into an action (no shell commands, no tool calls, no opening apps, no changing settings). Any input, spoken or pasted, that makes the app do something other than output cleaned text is a vulnerability.
  • API key storage. The Gemini key is stored in the platform secret store (libsecret on Linux, Android Keystore-backed storage on Android). Bugs that leak it to logs, files, the clipboard, crash reports, or the network (other than requests to the Gemini API) are in scope. Note that the optional GEMINI_API_KEY environment variable and flatpak override --env store the key in plain text by design; that is documented, not a vulnerability.
  • Model download integrity. Bypassing the SHA-256 verification of model files, or downloading from an unexpected origin.
  • Privacy. Any network traffic other than Gemini requests (cloud mode, opt-in) and model downloads, or any form of telemetry.
  • Text insertion and portals. Inserting text into the wrong window, or abusing the RemoteDesktop portal session or the optional ydotool socket beyond typing dictated text.
  • Build and release pipeline. Unpinned or tampered engine binaries, workflow injection, leaked signing secrets.

Out of scope:

  • Content quality or accuracy of the model output.
  • Problems in Google's Gemini service, Gemma weights, Flutter, or other upstream projects (report those upstream; tell us if we can mitigate).
  • Attacks that need an already compromised device or desktop session.
  • Free-tier data use by Google, which is disclosed in the app and README.

There aren't any published security advisories