This repository contains only tutorial material: Jupyter notebooks, the images they embed, and the datasets they ship. It is executed by CI and published as static HTML, so the attack surface is a notebook that downloads or unpickles something at build time.
Report a vulnerability in the PINA library itself in the main repository: PINA-org/PINA. Use the "Report a vulnerability" button under the "Security" tab; it opens a private channel with the maintainers.
If the issue is specific to this repository — for example a tutorial that
downloads from an untrusted host, or a shipped dataset that is unexpectedly a
pickle — report it the same way, in the main repository, and mention
PINA-Tutorials in the description. There is no separate security release
cadence here.
- Prefer plain formats (
.mat,.npz,.txt) over pickles. A.ptfile loaded withtorch.loadexecutes arbitrary code, so a shipped.ptmust be treated as trusted and must not be silently replaced. - Download datasets from a
raw.githubusercontent.comURL pointing at a specific branch or tag, never from a redirecting or short-lived host. - Do not commit credentials, tokens, or private endpoints.