Skip to content

Legal

Privacy

CGMPy is a research and analysis library. It does not collect, transmit, or store any user data. The library operates entirely locally on the data you provide.

Data you provide to CGMPy

When you call GlucoseData("my_file.csv"), CGMPy reads the file from your local disk and processes it in memory. No data ever leaves your machine as a result of using the library.

Data you share with the CGMPy project

CGMPy does not have a backend. Sharing data with the project is entirely under your control and happens only when you:

  • Open an issue (GitHub Issues).
  • Open a discussion (GitHub Discussions).
  • Open a pull request.

In all of these cases, you decide what to share. Please review the Security policy before sharing data, especially medical data.

Cookies and tracking

The CGMPy documentation site is hosted on GitHub Pages and does not set any cookies. No analytics, no tracking, no third-party scripts.

Telemetry

CGMPy does not phone home, does not collect usage statistics, and does not embed any analytics SDK.

What CGMPy does log

CGMPy uses the Python logging module. By default, log levels are WARNING and above. The library does not log glucose values, patient IDs, or any other potentially sensitive data.

If you enable INFO or DEBUG logging, you may see messages like:

INFO: cgmpy.data.loader: Loaded 1728 records from /path/to/file.csv
INFO: cgmpy.data.processor: Dropped 3 duplicate rows

These messages do not include glucose values or patient identifiers.

Compliance

CGMPy is released under the MIT License and ships without warranty. The library is not certified under HIPAA, GDPR, or any other medical- data regulation. It is the user's responsibility to ensure that their use of the library complies with applicable laws in their jurisdiction.

GDPR

CGMPy is a tool, not a service. The General Data Protection Regulation (GDPR) applies to data controllers and data processors that determine the purposes and means of processing personal data. CGMPy is none of these — it is software that runs locally on your data.

That said, if you use CGMPy to process personal data (including health data, which is a special category under GDPR Art. 9), you are likely a data controller and have obligations.

This section summarizes the library's posture, not legal advice.

Posture of the CGMPy library

  • CGMPy does not collect any data. It is a pure-Python library that runs locally.
  • CGMPy does not transmit any data. No telemetry, no analytics, no network calls.
  • CGMPy ships with anonymization helpers (scripts/anonymize_cgm.py) to help you comply with the data-minimization principle (Art. 5(1)©).
  • CGMPy explicitly forbids committing real patient data to the repository (see SECURITY.md).

What you (the user / controller) must do

If you use CGMPy on personal data:

  1. Have a lawful basis for processing (Art. 6).
  2. For health data, satisfy Art. 9 (special categories), typically through explicit consent or another Art. 9(2) basis.
  3. Inform the data subject (Art. 13/14) about how their data is used.
  4. Implement appropriate technical and organizational measures (Art. 32). CGMPy is one of those measures — it does not exempt you from the others.
  5. Honor data subject rights (access, rectification, erasure, restriction, portability — Arts. 15-20).
  6. Document your processing (Art. 30) and run a DPIA if high-risk (Art. 35).

PHI / special category data

CGM data is health data and therefore a special category under Art. 9. Treat it as such:

  • Anonymize before sharing for bug reports or examples.
  • Encrypt at rest and in transit.
  • Restrict access to those who need it.
  • Audit access logs.
  • Plan for deletion when the lawful basis expires.

CGMPy provides a basic anonymization script at scripts/anonymize_cgm.py. Use it on any real dataset before including it in a bug report.

Anonymization vs. pseudonymization

CGMPy's helper produces pseudonymization (patient IDs are replaced, but timestamps are only shifted, not removed). True anonymization requires irreversible transformations, which depend on the use case.

If you need stronger guarantees, consider:

  • Removing all timestamps (or binning to weeks/months).
  • Aggregating to daily summaries.
  • Using k-anonymity or differential-privacy techniques for cohort analyses.

International transfers

CGMPy is hosted on GitHub. If you are in the EU and use GitHub, your data may be transferred to the US. See GitHub's DPA for their commitments under the EU-US Data Privacy Framework.

Right to erasure

If a data subject requests erasure (Art. 17):

  • Erase the data from your local storage.
  • Erase any derived datasets (aggregates, models).
  • If you have published CGMPy results, retract or correct the publication.

CGMPy does not retain any of your data, so there is nothing to erase on the library side.

See also