Data Science · Beginner level

Beginner level · Module 5: Data Privacy and Ethics in Analysis · Lesson 2 of 3

Anonymising and Protecting Outputs

Remove or reduce identifying information in data and results before sharing.

In this lesson you will

  • Apply pseudonymisation, generalisation and suppression
  • Check outputs for re-identification risk

Key concepts

pseudonymisationgeneralisationsmall-cell suppressionre-identification risk

The lesson

Pseudonymisation replaces direct identifiers with codes; whoever holds the key can still re-identify, so the key must be protected. Anonymisation aims to make re-identification impossible in practice.

Reduce detail: ages into bands, GPS into areas, rare occupations into 'other'. Suppress or combine table cells with very few people — a common rule is to hide counts below five.

Before publishing, ask: could someone who knows the community recognise a household from this table or data set?

Step by step

  1. Replace names and phone numbers with codes.
  2. Band ages and remove GPS coordinates.
  3. Suppress small cells in the district tables.
  4. Review outputs for re-identification risk.
Step-by-step

Preparing a survey for safe sharing

The steps that turn a household survey into a data set and tables that can be shared without exposing anyone.

  1. Remove direct identifiers

    Delete names, phone numbers, ID numbers and household GPS points; keep a protected key only if follow-up is approved.

  2. Reduce indirect detail

    Group ages into bands, locations into payams or counties, rare occupations into 'other'.

  3. Check unique combinations

    Count households with a unique mix of area, household size and head's age band; generalise further where counts are tiny.

  4. Suppress small cells

    In published tables, hide or combine counts below five.

  5. Review sensitive variables

    Remove or aggregate health, displacement or protection details unless essential and approved.

  6. Document and share securely

    Write what was changed, who may use the data and for what, and send it through a protected channel.

Result

Analysts and partners get the information they need while no household can reasonably be recognised — including by neighbours who know the area.

Practice activity

Create the shareable data set and tables for the water project.

Expected result: Outputs with no direct identifiers, reduced indirect detail and suppressed small cells.

Check your understanding

  1. Is pseudonymised data anonymous?

    Show answer

    No; whoever holds the key can re-identify people.

  2. Why suppress small counts in tables?

    Show answer

    They can reveal information about identifiable individuals.

Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.