Security Policy
Supported versions
Inkvec is pre-1.0. Only the latest published 0.x release is supported with security
fixes; please upgrade before reporting an issue against an older version.
Reporting a vulnerability
Please do not open a public issue for a security vulnerability. Instead, use GitHub's private vulnerability reporting:
- Go to the Security tab of this repository.
- Click "Report a vulnerability" to open a private advisory.
The report is visible only to you and the maintainers until an advisory is published. This is the one channel for security reports: there is no security mailbox. If the button is missing or you cannot use it, open a public issue that says only that you have a security report and asks a maintainer to contact you — no details, no reproduction — and we will arrange a private channel from there.
Please include:
- the Inkvec version (
inkvec --version) and platform, - the command line or API call used,
- the input file (or a minimized reproduction) if you can share it, and
- what you expected versus what happened.
Scope
In scope:
- Malformed or adversarial image input — crashes, panics, excessive memory/CPU use, or memory-safety issues while decoding or tracing a PNG, JPEG, WebP, GIF, BMP or TIFF.
- Command-line and library argument handling — flag parsing, path handling, and any
external command invoked via flags such as
--sr-command. - The browser build (
crates/inkvec-wasm,web/) — anything that would let a hosted page misbehave beyond what the user's own input authorizes, given tracing runs entirely client-side and nothing is uploaded.
Out of scope:
- The optional trained restorer network itself (
restorer.onnx) — it is not part of this repository and is distributed separately. - Issues that require an attacker to already control the machine running
inkvecor to modify the binary or its dependencies. - The benchmark harness (
bench/) and research tooling, which are developer-only and not part of any shipped artifact.
Response expectations
This is a small open-source project without a dedicated security team. We aim to acknowledge new reports within a week and to keep you updated as we investigate, but we do not commit to a fixed resolution SLA before 1.0. Coordinated disclosure is appreciated; we will credit reporters (with permission) in the release notes.