Anonymisation vs pseudonymisation vs data masking: three techniques, three labels, and plenty of confusion. Teams describe a dataset as “anonymised” when they’ve only swapped names for codes, or call test data “masked” without checking whether anyone could reverse it. The label matters, because it decides whether UK GDPR still applies.

This post explains anonymisation vs pseudonymisation vs data masking, how the law treats each, what the ICO’s guidance says, and how to choose between them. If you need the work done for you, our data anonymization services cover UK and global datasets.

Contents

  1. Why Does the Label Matter?
  2. What Is Anonymisation?
  3. What Is Pseudonymisation?
  4. What Is Data Masking?
  5. Anonymisation vs Pseudonymisation vs Masking: How Do They Compare?
  6. What Does the ICO Say?
  7. Which One Should You Use?
  8. FAQs

Why Does the Label Matter?

UK GDPR applies to personal data, meaning information about someone who can be identified. Anonymous information falls outside the law. Pseudonymised information stays inside it. Masked data can land on either side, depending on how it was done.

Getting it wrong works in both directions. Treat pseudonymised data as anonymous and you may share or keep it without a lawful basis. Treat anonymous data as personal and you take on duties, such as retention limits and access requests, that no longer apply. Either mistake is easier to avoid when you know what each technique does, which is why anonymisation vs pseudonymisation is the first question to settle.

What Is Anonymisation?

Anonymisation turns personal data into information that can’t identify anyone by any means reasonably likely to be used. The test considers the data itself, other information available, and the cost, time and technology someone would need to re-identify people.

Common methods include generalising values into ranges, removing rare records, aggregating figures and adding statistical noise. Larger programmes often run anonymisation alongside encryption and access control as part of data security as a service. Once the result passes the test, it isn’t personal data, so you can keep it beyond your retention period or share it without the restrictions UK GDPR places on personal data.

The Catch

True anonymisation is hard to achieve. If you hold a key, a lookup table or the original dataset that could re-link records, you probably haven’t anonymised anything.

Test Before You Rely On It

Try to re-identify the result yourself before calling it anonymous. Check for rare combinations of fields, link it against public data, and record what you found. That evidence is what regulators and partners will ask for, and it is the core of any data anonymisation services engagement.

What Is Pseudonymisation?

Article 4(5) of UK GDPR defines pseudonymisation as processing personal data so it can’t be linked to a person without additional information, kept separately and protected. In practice, that means replacing names or account numbers with codes and holding the key somewhere else.

Pseudonymised data is still personal data, because the link back to the person exists. The law rewards it anyway. Article 25 names it as a way to build privacy into your systems, and Article 32 lists it as a security measure.

Where the Key Lives Matters

Pseudonymisation only works if the key is genuinely separate. Store it in a different system, give different people access to it, and log every time someone uses it to re-link a record.

What Is Data Masking?

Data masking hides real values behind realistic substitutes. Picture a card number shown as 4532 XXXX XXXX 1234 (a common requirement under PCI DSS), or a customer’s name swapped for a made-up one in a test database. UK GDPR doesn’t define masking, so its legal status depends on how you do it.

Static vs Dynamic Masking

Static masking changes the stored copy, as when you build a test environment. Dynamic masking leaves the original untouched and hides values as people view them, as when a support agent sees only part of a card number.

How the Law Treats It

Dynamically masked data is still personal data, because the original remains behind the screen. Statically masked data may be anonymous, but only if nobody can reverse the substitution and the remaining fields can’t identify anyone.

Anonymisation vs Pseudonymisation vs Masking: How Do They Compare?

The table below compares the three techniques on the questions teams ask most.

Anonymisation Pseudonymisation Data masking
Legal status under UK GDPR Outside the law, if the test is met Personal data Depends on the method
Can the original be recovered? No Yes, with the separate key Dynamic: yes. Static: only if a mapping exists
Data quality kept Reduced detail Full record-level detail Realistic values, no real ones
Biggest risk Re-identification through linked datasets Key held too close to the data Assuming masked means anonymous
Best fit Statistics, research sharing, long-term trend data Analytics and research that need to re-link records Test environments, support screens, demos

What Does the ICO Say?

The ICO published its final anonymisation and pseudonymisation guidance in March 2025. It treats identifiability as a spectrum and doesn’t require zero risk. The question is whether re-identification is reasonably likely, judged against a motivated intruder: someone capable and determined, but without special access.

The guidance makes three points that catch businesses out. Anonymising personal data is itself processing, so you need a lawful basis to do it. The same dataset can be personal data in your hands and anonymous for a recipient who can’t link it back. Risk also changes over time, so anonymisation decisions need review. The ICO has said the guidance is under review following the Data (Use and Access) Act 2025.

The EU Position

In EDPS v SRB in September 2025, the Court of Justice held that pseudonymised data may not be personal data for a recipient with no realistic way to re-identify it. The ruling doesn’t bind UK courts, but it matters if you also handle EU residents’ data.

Which One Should You Use?

When weighing anonymisation vs pseudonymisation, start with one question: will anyone ever need to know who a record belongs to again?

Choose Anonymisation

Use it when you publish statistics, share data with researchers who never need to contact individuals, or want to keep trend data after its retention period ends.

Choose Pseudonymisation

Use it when analysts need record-level detail, when you track the same people over time, or when you may need to re-link a record, such as in a fraud investigation.

Choose Data Masking

Use it when engineers need realistic data to build and test systems, or when staff need to see part of a value to do their job but not all of it.

Combine Them

Many organisations use all three: masking for test environments, pseudonymisation for analytics, and anonymisation for anything shared externally or kept long term.

Questions to Ask First

Before you choose, ask who will see the data, whether anyone will need to re-link it, and how long you plan to keep it. Then check what other datasets the recipient holds, because that decides how strong your protection needs to be.

Anonymisation vs Pseudonymisation: FAQs

Anonymised data can no longer identify anyone by any means reasonably likely to be used, so it falls outside UK GDPR. Pseudonymised data has identifiers replaced with codes, but a separately held key can re-link it, so it is still personal data.
Yes. Article 4(5) defines pseudonymisation as processing that relies on additional information kept separately. Because that link back to the person exists, the data stays personal data, although UK GDPR treats pseudonymisation as a recognised privacy and security measure.
Not automatically. Dynamically masked data is still personal data because the original sits behind the screen. Statically masked data may be anonymous, but only if nobody can reverse the substitution and the remaining fields cannot identify anyone.
The ICO’s guidance, finalised in March 2025, treats identifiability as a spectrum and asks whether re-identification is reasonably likely, judged against a motivated intruder. Anonymising personal data is itself processing, so you need a lawful basis to do it.
Data masking is usually the best fit for test environments, demos and support screens, because engineers get realistic values without real ones. Use pseudonymisation for analytics that need to re-link records, and anonymisation for anything shared externally or kept long term.

Not Sure Which Technique Fits Your Data?

The right choice depends on how you use each dataset and who needs to see it. Our data anonymisation services test whether a dataset meets the ICO’s standard. We also cover data pseudonymisation for analytics and research.

For test and support environments, explore our data masking service.

Leave A Comment

Receive the latest news in your email
Table of content
Related articles