Skip to content

Security: ADJ189/CompressZ

SECURITY.md

Security Policy

CompressZ processes files entirely in the browser and has no backend, so the most important security properties are: file content never leaves the device, user-controlled text never becomes executable markup, and the only remotely fetched executable code is third-party engine code from the documented CDNs in PRIVACY.md. Model weights and fonts are also downloaded as static assets, but they do not carry user file data. Some engines (for example FFmpeg.wasm via esm.sh and PDF.js/ pdf-lib via cdn.jsdelivr.net) are deliberately not bundled; the app downloads and runs them on first use. File bytes are never part of those requests, and the CDN hosts, versions and licenses are listed in CREDITS.md.

Supported versions

Security fixes land on main and in the latest deployed build at https://compressz.pages.dev. Older 1.x prereleases are not separately maintained; please reproduce against the latest main before reporting.

Reporting a vulnerability

Do not open a public issue for a security problem.

Use GitHub's private vulnerability reporting on the repository:

https://github.com/ADJ189/compressz/security/advisories/new

If private reporting is unavailable, open a minimal public issue asking for a private contact channel — without exploit details — and a maintainer will follow up.

Please include:

  • A description of the issue and its impact (data exposure, XSS, denial of service, dependency compromise, etc.).
  • Browser and OS versions, plus the CompressZ route/tool involved.
  • Reproduction steps or a proof of concept. A malicious file is ideal; if the file is large, describe its structure and attach a minimized version.
  • Whether any file data, localStorage value, or network request is involved.

What to expect

  • Acknowledgement on a best-effort basis, usually within 7 days.
  • Triage and a remediation plan after the report is reproduced.
  • Credit in the fix/changelog if you want it. There is no bug bounty.

Scope

In scope:

  • Cross-site scripting or HTML injection through file names, OCR text, SVG content, extracted office-document text, or error messages.
  • Any path that transmits file bytes or file-derived content to a server.
  • Leaks of localStorage settings beyond the documented theme, cz-settings-v2, and legacy cz-settings-v1 keys.
  • Supply-chain issues in dependencies that actually ship in the browser bundle or are loaded from the documented CDNs.
  • Denial-of-service bugs in the app's own parsers/engines that are reachable from a normal file drop.

Out of scope:

  • Third-party CDN outages or compromises outside the project's control (report those upstream, but tell us if the app should fail more safely).
  • Browser zero-days and OS-level attacks.
  • Social engineering, physical access, or issues requiring a compromised browser extension.
  • Missing hardening headers with no demonstrated impact.

Safe harbor

Good-faith research that respects user privacy, avoids destructive testing, and does not access other people's data is welcome. We will not pursue legal action for accidental issues found while following this policy.

There aren't any published security advisories