Skip to content

Security: kixel-ai/rind

SECURITY.md

Security policy

Reporting a vulnerability

Report it privately through GitHub: on the repository's Security tab, choose Report a vulnerability. Don't open an issue, pull request, or quest for it, since those are public.

Say which package and version it affects, the target and CPU, and how to reproduce it. If the flaw is in code rind took from another project, which ALGORITHMS.md names, it may affect that project too: report it there as well.

What counts

These are the promises rind's code makes, and breaking one is a vulnerability on every target, wasm included:

  • Hostile input. A malformed key, signature, ciphertext, or encoding returns an error. A panic, or a cost the input can make unbounded, is a vulnerability.
  • Constant time. No branch, memory index, or variable-time instruction in rind's secret-handling code depends on a secret.
  • Correct results. rind computes what the algorithm's specification says, and accepts no forged signature or tag.
  • Wiping. A key, keyed state, or heap scratch derived from a private key is wiped before it is freed, where Wiping keys says rind wipes it.

Out of scope:

  • The workspace's unpublished crates: rind-bench, rind-cpuid, rind-fips, rind-timing, and the fuzz targets. A flaw in rind's own code that rind-fips builds is a flaw in rind.
  • What a wasm engine's compiler does to rind's constant-time code, which PLATFORMS.md says is unchecked. A secret-dependent branch or index in rind's own source is in scope.
  • A flaw in a crate an adapter plugs into, such as rustls: report it to that project.

Reviewing rind

AUDIT.md is where a security review of rind starts: its threat model, where its code came from, what each check establishes, and where the checks are thinnest.

Supported versions

Only the latest release of each package rind publishes gets fixes: the rind crate and its rind-<consumer> adapter crates on crates.io, and its packages on pub.dev and PyPI. A fix ships as a new release of each package it affects.

A fixed vulnerability gets a GitHub security advisory and a RustSec advisory.

There aren't any published security advisories