Skip to content

Security: Cherie05/interflect

Security

SECURITY.md

Security Policy

Supported versions

Version Supported
0.1.x yes

Threat model

Interflect is an offline command-line renderer. It reads a scene file and writes a PNG. It opens no network connections, executes nothing it reads, and requires no elevated privileges.

The realistic attack surface is a malicious .rad file, which someone might open from an untrusted source. Malformed input produces a diagnostic and exit code 1 rather than a panic, and this is covered by unit tests and a CI gate.

Settings in the render {} block are bounded, because they are read straight from the scene file and reach allocation sizes directly. An audit found surfels: 999999999 aborting the process on a 76 GB allocation and a 200000x200000 film hanging indefinitely; both are now rejected with a message. The same limits are re-checked after CLI overrides, since --surfels bypasses the parser. Current ceilings are in scene.rs: 16384 px per side, 40M pixels, 2M surfels, 4096 bounces.

Things worth reporting:

  • A .rad file that causes a panic, an out-of-bounds access, or unbounded memory or CPU consumption
  • Any path traversal through -o or a scene reference
  • A crash reachable from a PNG the compare tool reads

There is no unsafe code in this repository.

Reporting a vulnerability

Use GitHub's private reporting: Security → Report a vulnerability on the repository. That keeps the report confidential until a fix ships.

If you would rather not use GitHub, email arunvpp24@gmail.com instead.

Please include the .rad file or input that triggers it, the version (interflect --help prints it) and your OS.

You can expect an acknowledgement within a week. Given the scope above, most findings will be treated as ordinary bugs and fixed in the open — but the private channel exists so that call is made after seeing the report, not before.

There aren't any published security advisories