Free workshop series

Differential privacy in healthcare

A free five-session workshop series on sharing and analysing patient data with differential privacy, built for health organisations in Africa.

5 sessions60 minutes eachOnline, live with Q&AThursdays, 29 October to 26 November 2026, 14:00 GMT

What you'll be able to do

By the end of the series you can decide when differential privacy fits a health data problem, apply it to statistics and machine learning, and explain the choice to a regulator or ethics board.

The sessions

  1. Why “anonymised” health data isn’t

    Thursday, 29 October 2026, 14:00 GMT

    You'll be able to: Explain why removing names fails, where k-anonymity breaks, and what differential privacy promises instead, including epsilon in plain terms.

    Hands-on: Re-identify "patients" in a small synthetic clinic dataset using only age, sex and district.

  2. Publishing health statistics safely

    Thursday, 5 November 2026, 14:00 GMT

    You'll be able to: Add calibrated noise to counts and rates, handle small numbers in rural districts, and track a privacy budget across monthly releases.

    Hands-on: Release district case counts with the open-source OpenDP library and compare accuracy at different epsilon values.

  3. Training clinical models with differential privacy

    Thursday, 12 November 2026, 14:00 GMT

    You'll be able to: Describe differentially private training, estimate its accuracy cost on small or under-represented datasets, and decide when to pair it with federated learning.

    Hands-on: Train the same simple risk model with and without differential privacy and compare the results.

  4. Sharing data with researchers and vendors

    Thursday, 19 November 2026, 14:00 GMT

    You'll be able to: Judge when differentially private synthetic data is good enough to share, and test it for usefulness and leakage.

    Hands-on: Generate a small synthetic dataset and check what it preserves and what it loses.

  5. Governance: choosing, documenting and defending epsilon

    Thursday, 26 November 2026, 14:00 GMT

    You'll be able to: Write a privacy decision record and map differential privacy to Ghana, Nigeria, Kenya and Egypt's data protection laws and the GDPR.

    Hands-on: Complete a one-page decision record using NIST's "differential privacy pyramid" (SP 800-226).

Who it's for

  • Data protection officers and compliance leads in hospitals, insurers and health-tech companies
  • CTOs, CISOs and data engineers who hold patient data
  • Ministry of Health and programme teams that publish health statistics
  • Researchers and NGOs who need to share data across borders

Level: Comfortable with spreadsheets and basic statistics. No programming needed to follow; optional notebooks for technical attendees.

Format

Cost
Free
Each session
35 minutes teaching, 15 minutes hands-on demo, 10 minutes Q&A
Replays
Recording, transcript and slides published on this site after each session, without a sign-up wall
Materials
Slides, a synthetic health dataset, and optional notebooks
Certificate
Attend a session live for at least 45 minutes and a certificate of attendance is emailed to you automatically within 24 hours, showing your minutes attended and a verification link. You may claim CPE hours under your certifying body's continuing education policy. Register and join with the same email address. Watching the replay does not qualify.

Read before the first session