Skip to content

SECURITY.md

Latest commit

 

History

History
80 lines (63 loc) · 3.4 KB

SECURITY.md

File metadata and controls

80 lines (63 loc) · 3.4 KB
Raw

Security policy

Pelli Instant opens your photo folders, reads files from every kind of camera, writes next to your photos when you export, keeps a sign-in token for your Pelli account, downloads AI models when you ask for the Maximum precision, and installs its own updates. We take security reports seriously. Thank you for helping keep it safe.

Reporting a vulnerability

Please don't open a public issue. Email contact@pelli.io with Security in the subject.

Helpful details:

  • The Pelli Instant version (Settings → About) or the commit you built from, and your operating system and version.
  • What an attacker could do, and what they need to do it.
  • Steps to reproduce, or a proof of concept. For a file that crashes or misleads the decoder, the smallest file that shows it.
  • Whether you've shared the issue with anyone else.

Never include a Pelli access token in a report, nor photos you don't have the right to share.

What happens next

Pelli Instant is maintained by a small team, so we can't promise fixed deadlines. We'll do our best to:

  • acknowledge your report within a week;
  • tell you whether we can reproduce it, and keep you posted while we fix it;
  • publish the fix in a new release and credit you in its notes, if you'd like.

Please give us a reasonable time to ship a fix before disclosing the issue publicly.

Supported versions

Security fixes go into the latest release only, which installed copies are offered as an in-app update. Before the first release, they go into the main branch.

Scope

In scope: the code in this repository, including:

  • reading untrusted files: folder scanning, RAW containers, EXIF, embedded JPEG previews (crates/cull-decode);
  • what the app writes: XMP sidecars, copies of the selection, rejects moved to a sub-folder (crates/cull-analysis/src/xmp.rs, crates/cull-engine/src/export.rs);
  • the sign-in flow, how the token is stored (macOS Keychain, Windows Credential Manager) and the requests sent to the Pelli API;
  • the desktop shell: commands exposed to the interface, the asset protocol scope that serves previews (src-tauri);
  • the model files' checksum verification at load (crates/cull-ml);
  • the download of the Maximum pack: HTTP range resume, SHA-256 of each file against the manifest compiled into the app, the check of every model against reference outputs (crates/cull-engine/src/modelpack.rs, crates/cull-ml/src/maximum.rs), and PELLI_MODELS_BASE_URL, which only debug builds read;
  • the in-app updater: how latest.json is read, the check of each update's signature against the public key in src-tauri/tauri.conf.json, and when it installs (src-tauri/src/updates.rs);
  • the release workflow, the installers it produces, and how it signs them and the update bundles (the private update key is a CI secret, never in the repository).

Out of scope, please report them to the right place:

  • The Pelli web platform (app.pelli.io) and your Pelli account.
  • ONNX Runtime, the AI models, Tauri's updater plugin and other dependencies: report to their maintainers. Tell us too if Pelli Instant is affected.
  • Hugging Face and GitHub, which host models and releases: a file they serve is only used once its SHA-256 or its signature matches.
  • Builds made by someone else from this code, such as builds without the Pelli account.
  • Attacks that require an already compromised computer or user account.