Security and Privacy

Mind Archive exists to keep personal material private. Security is a product requirement here, not a checklist.

The privacy model in one paragraph

Your archive lives on your computer. Mind Archive has no account system, no hosted service, and no telemetry. It makes no outbound network request as part of its normal operation. Nothing is uploaded anywhere unless you explicitly configure cloud storage, which is off by default and cannot be enabled silently.

What we guarantee

No silent uploads. MIND_ARCHIVE_CLOUD_ENABLED defaults to false. No cloud adapter exists yet, and when one does it will require explicit configuration. The API reports cloud status so the interface can tell you plainly where your data is.

No telemetry. No analytics, no crash reporting, no usage statistics.

No secret ever reaches the browser. The frontend receives only non-sensitive configuration. API keys, when future features need them, stay server-side.

Your content stays readable and yours. Markdown and JSON on your filesystem. No proprietary format, no encryption you cannot undo, no lock-in.

Conversation content is not logged. Application logs record what happened — “imported 412 conversations” — never what was said.

Imported files are untrusted input

This is the most important attack surface in the project. A provider export is a file from outside, and from Milestone 2 onward we parse them.

Every importer must:

• Validate structure before trusting it. Never assume a field exists, has the expected type, or is the expected size.
• Guard against path traversal. Filenames inside an archive are attacker controlled. Resolve every path and confirm it is inside the archive root before writing. This is centralised in mind_archive/paths.py — use it.
• Guard against zip-slip and zip bombs. Check entry names and uncompressed sizes before extracting.
• Cap resource use. Bound file sizes, entry counts and nesting depth.
• Fail clearly. A malformed archive produces a plain-language error, not a stack trace and not a partial, corrupted import.

Treat every imported archive as if it were written by someone hostile, because you cannot prove it was not.

Secrets

Never commit API keys, passwords, tokens, private keys, credentials or personal archives.

.env is git-ignored; .env.example holds placeholders only and is the file that gets committed. If a secret is ever committed, rotate it — removing it from history is not enough.

.gitignore also excludes data/, *.db, *.sqlite and *.zip so that your archive and any provider export you are working with cannot be committed by accident.

Network exposure

By default the API binds to localhost. Do not expose it to a network you do not control: there is no authentication, because it is a single-user local application.

CORS is restricted to the local frontend origin and is configurable through MIND_ARCHIVE_CORS_ORIGINS. Widen it only if you understand the consequence.

If you ever want to reach Mind Archive from another machine, put it behind a reverse proxy with authentication that you control. That is out of scope for the project itself until Milestone 7.

Dependencies

Dependencies are kept deliberately few, which is itself a security measure. Each one must be justifiable in a sentence. CI runs on every push and pull request.

For contributors

Before opening a pull request:

• Run git diff --cached and read it. Confirm no secret, no .env, no personal archive and no database file is included.
• If you touched anything that handles a path or an imported file, say so in the pull request description.
• If you added a dependency, explain why.

Reporting a vulnerability

Please do not open a public issue for a security vulnerability.

Use GitHub’s private vulnerability reporting on the repository (Security → Report a vulnerability), or contact the maintainers directly.

Include what you found, how to reproduce it, and what an attacker could achieve. We will confirm receipt, agree a disclosure timeline with you, and credit you unless you would rather we did not.