Security vulnerability reports will be accepted and acted upon for all released versions.

Allow the community to report potential security vulnerabilities to maintainers and repository owners privately. Learn more about private vulnerability reporting.

PyVista wraps VTK and meshio. Mesh parsers inherit upstream parser CVEs. Do not load mesh files from untrusted sources. PyVista is a viewer/processor for trusted scientific data; it is not a sandbox.

PyVista refuses .pkl / .pickle extensions in pyvista.read() and DataObject.save(). The top-level pyvista.read_pickle / pyvista.save_pickle API shims remain importable for backwards compatibility but always raise ValueError.

Why. Pickle is a Python serialization protocol, not a mesh file format. Unpickling untrusted data is arbitrary code execution (CWE-502). A pyvista.read(path) call with an attacker-influenced path (downloaded data, shared notebook, copy-pasted command) would be a one-shot RCE if pickle dispatch were enabled.

What still works. Python's pickle protocol via DataObject.__getstate__ / __setstate__ is unchanged — multiprocessing, dask, and joblib continue to work. pyvista.set_pickle_format() tunes that in-memory protocol.

Migration.

• For mesh files: use a real mesh format (.vtu, .vtp, .vtm, .vtk, .ply, .stl, ...).
• For fast single-blob round-trips: install pyvista-zstd (pip install pyvista[io]) and use the .pv format.