QuantUI is an educational teaching tool designed for classroom and local research use. By default it runs calculations inside your own Python session — there is no hosted server and no user accounts, and results, logs, settings and checkpoints stay on the machine (or cluster account) you run it on. The optional SLURM batch mode writes job folders under your own cluster home and submits them with your own sbatch.

If you find a security issue (e.g. a path-traversal bug, an unsafe use of subprocess, or a dependency with a known CVE), please do not open a public GitHub issue.

Instead, report it privately via one of these channels:

• GitHub private vulnerability reporting — use the Security tab on this repository (preferred).
• Email — contact the lab maintainer through the GitHub profile @The-Schultz-Lab.

Please include:

1. A short description of the issue and its potential impact
2. Steps to reproduce (or a minimal proof-of-concept)
3. The version of QuantUI affected
4. Your suggested fix, if you have one

We aim to acknowledge reports within 5 business days and to release a fix within 30 days for confirmed vulnerabilities.

This project is a local teaching tool. The threat model is limited — there is no multi-user server and no user accounts. It does keep persistent data in the user's own account, protected only by its file permissions: saved results (./results/ by default, QUANTUI_RESULTS_DIR in the Apptainer image and Viewer), and under ~/.quantui/ logs (logs/perf_log.jsonl, logs/event_log.jsonl), user settings (settings.json), calculation checkpoints (checkpoints/), SLURM job folders and staging (jobs/, staging/) and the generated launcher notebook. The log, settings and checkpoint paths can be redirected with QUANTUI_LOG_DIR, QUANTUI_SETTINGS_PATH and QUANTUI_CHECKPOINT_DIR.

The primary security concerns are:

• Access to the running app. quantui run app, quantui view, the native launchers and the Apptainer image's app runscript bind Voilà to 127.0.0.1 and require a per-run token (QUANTUI_APP_TOKEN when set, otherwise a fresh random one); the printed/opened URL carries it (?token=...). Standalone Voilà reads no token from the environment, so on a shared host the token is visible to other local users in the process list (ps). launch-native-jupyter.sh uses JUPYTER_TOKEN (off the command line). The Apptainer image's default JupyterLab mode keeps Jupyter's own token handling. Anyone holding the token can drive the session as you — read History, submit SLURM jobs with your account.
• Untrusted files the app reads: the Viewer's .zip upload of a results folder (quantui/app_viewer.py), structure-file upload (quantui/structure_upload.py: XYZ/PDB/CIF/MOL/SDF/MOL2/Gaussian), and ingestion of SLURM job folders and result data written outside the session (quantui/backends/slurm_ingest.py, quantui/results_storage.py).
• Path traversal in file I/O (the uploads/ingest above, molecule XYZ parsing, script and bundle export)
• Unsafe subprocess calls — PySCF runs in-process via the Python API, but subprocess is used for SLURM (sbatch, squeue, sacct, scancel; see quantui/backends/slurm.py and the quantui-batch launcher), for starting Voilà (quantui run app / quantui view), and for small CLI and diagnostic helpers. Job-folder paths and resource requests for SLURM are validated in quantui/backends/cluster_security.py.
• Dependency vulnerabilities in the scientific Python stack (numpy, pyscf, ase, etc.) — monitored via Dependabot

Issues that are explicitly out of scope:

• Vulnerabilities that require physical access to the machine running the notebook
• Vulnerabilities in Jupyter/JupyterHub itself (report those upstream)
• Social-engineering or phishing attacks

QuantUI relies on PySCF, ASE, NumPy, Matplotlib, ipywidgets, and py3Dmol. Security advisories for these packages are tracked automatically via Dependabot. If you become aware of a critical CVE in one of these dependencies before Dependabot picks it up, please report it using the channels above.