Re-identification: Why a Few Details Can Identify You

Additional resource for the lesson “Information Theory: The Basics for Anonymity” — Online Anonymity course

Re-identification is what happens when details that seem harmless on their own are put together and point back to one person. It does not take your name: a postcode, a birthday and a couple of habits can be enough. This resource explains, without formulas, why a few details can identify you, and helps you estimate how quickly your own crowd shrinks.

A. Why this matters

Privacy is deciding what others know about you. Anonymity, in this course, is something slightly different: acting without what you do being linked to who you are. So the question here is not “what have I shared?” but “how many people could this description fit?”

The original lesson called this information theory, the mathematics of how much a piece of data tells you. You do not need the maths. You need one intuition: every detail sorts people into those it fits and those it does not, and the group that is left gets smaller each time.

That is why “I never gave my name” and “the data was anonymised” are weaker reassurances than they sound. A name is only one way of picking someone out of a crowd. A few ordinary details can do the same job.

What the sources say

In a study published in 2000, the researcher Latanya Sweeney found that 87% of the US population was likely to be uniquely identified by three facts: a five-digit ZIP code (the US postcode), gender and date of birth. Using the town or city instead of the ZIP code still singled out about half of the population; using the county, 18%.

Later research found the same pattern in Europe. A 2013 study in Scientific Reports looked at fifteen months of mobile phone location data for one and a half million people in “a small European country” and found that “four spatio-temporal points are enough to uniquely identify 95% of the individuals”: four moments of where and when. In 2019, researchers at UCLouvain in Belgium and Imperial College London, writing in Nature Communications, estimated that “99.98% of Americans would be correctly re-identified in any dataset using 15 demographic attributes”. They concluded that even heavily sampled anonymised datasets are “unlikely to satisfy the modern standards for anonymization set forth by GDPR”, the EU’s General Data Protection Regulation.

European regulators describe the same risk in their own words. In 2014 the Article 29 Working Party, “an independent European advisory body on data protection and privacy”, named three risks to check when data is anonymised: singling out one person, linking records about the same person, and inferring details about them. It warned that “removing directly identifying elements in itself is not enough”. The UK’s Information Commissioner’s Office (ICO) uses the same vocabulary, describing linkability as “combining multiple records about the same person or group of people”. The European Commission explains that different pieces of information which together can identify someone count as personal data, and that “for data to be truly anonymised, the anonymisation must be irreversible.”

B. Key concepts

Five ideas that turn “a few details” into something you can reason about.

The anonymity set: the crowd you blend into

Your anonymity set is the group of people who match everything an observer knows about you. If a site knows only that a visitor came from Italy, the set is enormous. If it knows the visitor lives in one district of a city, is in their mid-thirties and runs along the same route every Sunday, the set may be a handful of people.

Why it matters to you: Your anonymity is only as strong as the size of that crowd. “How many people does this fit?” is a more useful question than “is my name on it?”

Each detail shrinks the crowd

A detail works like a filter. Gender roughly halves a crowd; a birth year cuts it much further; a full date of birth further still. The filters multiply rather than add up, so three modest details in a row can take a city down to a street.

Why it matters to you: No single detail needs to be sensitive for the combination to identify you. It helps to think about your details in groups, not one at a time.

Bits: an intuition, not a formula

Researchers measure identifying information in “bits”. In a primer for the Electronic Frontier Foundation (EFF), Peter Eckersley explained that one more bit “doubles the number of possibilities”; seen from your side, each yes-or-no detail halves the crowd. In his rough sums, picking out one person on the planet takes about 33 bits, a birthday gives about 8.5 and a star sign about 3.6. ZIP code, birthday and gender together came to just over 33 bits: enough, he wrote, that “we can probably say exactly who the person is.”

Why it matters to you: You do not need to calculate bits. The intuition is enough: about twenty halvings take a crowd of a million down to one person, and an everyday detail can account for several halvings at once.

Re-identification and “anonymised” data

Removing names from a dataset is not the same as making it anonymous. Re-identification means matching what remains, such as age, area and dates, with other information until a record points to one person again. The ICO calls someone who tries this a “motivated intruder”. Making details coarser helps, but less than you might expect: the 2013 mobility study concluded that “even coarse datasets provide little anonymity.”

Why it matters to you: When a service says your data is shared “anonymously”, it is reasonable to ask which details stay in it. The answer tells you more than the word does.

Blending in: the other way to grow the crowd

You can make your crowd bigger by sharing fewer details, and also by sharing the same details as many other people. Some tools are built on this second idea: the Tor Project says that “Tor Browser aims to make all users look the same.” Your everyday browser is itself a bundle of details; the fingerprinting resource in this section shows how to test it and what blending in looks like in practice.

Why it matters to you: Looking ordinary can protect you as well as staying quiet. An unusual combination stands out, even when each choice in it was made for privacy.

C. A practical example: Luca’s public profile

Luca is 36, lives in Milan and travels often. His social media profile shows no surname, only a nickname and a photo of a mountain, and he has always thought of it as fairly anonymous. Over a year, it filled up with ordinary moments: a birthday cake with “36” on it, posted on the day; his Sunday run along the same canal; check-ins at his gym; and a photo from the airport each time he left for a trip.

One evening, after reading how ordinary details combine, he looked at his profile as a stranger might. “Lives in Milan” fits a very large crowd. A man in his mid-thirties shrinks it a great deal. A birthday on a known day shrinks it again. A regular route and one particular gym leave, at a rough guess, a few people. Add the dates he was away, and the crowd is probably one: him. No single post said who he was. Together, they did.

What changed when Luca counted his crowd

He was not alarmed. He simply saw that his profile was linked to him through details, not through his name, and decided which details to keep.

• He kept the travel photos, but started posting them after he got home rather than from the airport.
• He removed the gym check-ins, which added little for his friends and a lot for anyone counting.
• He made older posts visible to friends only, so strangers had fewer details to combine.
• He left the birthday post: friends had written under it, and he decided it was worth its trace.

Luca made his crowd bigger in three places and left the rest as it was. Anonymity, for him, became something to adjust in proportion to what he cares about, not something you either have or lack.

D. Try it yourself: how fast does your crowd shrink? (15 minutes)

Four short steps with round numbers. Do not write down your real details: use categories (“my city”, “my birth year”) and rough fractions. The numbers are for intuition, not precision.

Step 1 — Pick a starting crowd (3 minutes)

• Choose a place someone could easily learn about you, such as your city or region, from a profile, a photo or a username.
• Give it a round size: 100,000, 1 million or 10 million. There is no need to look it up; the right order of magnitude is enough.
• Write only the number, for example “start: 1,000,000”.

Step 2 — List four public details (4 minutes)

• Think of four details a stranger could find out about you without asking: from a profile, a post, a photo or an old comment.
• Write the type of detail, never its value: “gender”, “birth year”, “birthday”, “neighbourhood”, “employer”, “sport”, “dog breed”.
• Prefer details that people you do not know can see.

Step 3 — Shrink the crowd (5 minutes)

• Apply a rough fraction for each detail, one after the other: a detail with two common answers, divide by 2; a birth year, divide by about 50; a birthday (day and month), divide by about 365; a neighbourhood, school, employer or club, divide by somewhere between 10 and 100.
• Worked example: 1,000,000 ÷ 2 = 500,000; ÷ 50 = 10,000; ÷ 365 ≈ 27. One local detail later, the crowd may be a single person.
• Real populations are uneven, so read your result as a rough idea of size, not a measurement.

Step 4 — Choose one detail to loosen (3 minutes)

• Look at which detail shrank your crowd the most. It is often the most specific one, not the most personal one.
• Choose one change: make it visible to fewer people, make it coarser (a month instead of a date, a city instead of a street), or keep it as it is, knowingly.
• Note your choice in one line.

A crowd is something you can grow again, one detail at a time.

Already covered in the Online Privacy course

Your Digital Footprint: Map the Data Traces of One Ordinary Day
It maps the declared, observed and inferred data that one day produces; this resource asks how many people those traces, once combined, could still describe.

E. Videos, articles and further resources

Research on re-identification first, then institutional guidance and digital rights tools. All in English.

Research

Data Privacy Lab — Simple Demographics Often Identify People Uniquely
Latanya Sweeney’s study of how ZIP code, gender and date of birth single out most of the US population.
https://dataprivacylab.org/projects/identifiability

Scientific Reports — Unique in the Crowd: The privacy bounds of human mobility
A peer-reviewed study (2013), based on mobile phone data from a European country, of how few location points make a person’s movements unique.
https://www.nature.com/articles/srep01376

Nature Communications — Estimating the success of re-identifications in incomplete datasets using generative models
A peer-reviewed study (2019) by researchers in Belgium and the UK of why sampled, “anonymised” datasets can still be re-identified; open-access copy on PubMed Central.
https://pmc.ncbi.nlm.nih.gov/articles/PMC6650473

Institutional guidance

Article 29 Working Party — Opinion 05/2014 on Anonymisation Techniques
European guidance (2014) on singling out, linkability and inference, and why removing names or using pseudonyms is not enough to anonymise data.
https://ec.europa.eu/justice/article-29/documentation/opinion-recommendation/files/2014/wp216_en.pdf

Information Commissioner’s Office (UK) — Anonymisation guidance: Glossary
Plain definitions of anonymisation, identifiability, singling out, linkability and the “motivated intruder”.
https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/data-sharing/anonymisation/glossary/

European Commission — Data protection explained
What counts as personal data, why pseudonymised data still does, and why anonymisation must be irreversible.
https://commission.europa.eu/law/law-topic/data-protection/data-protection-explained_en

Digital rights guides and tools

Electronic Frontier Foundation — A Primer on Information Theory and Privacy
The “bits” intuition, worked through with birthdays, star signs and ZIP codes.
https://www.eff.org/deeplinks/2010/01/primer-information-theory-and-privacy

The Tor Project — Tor Browser
How Tor Browser tries to make its users look alike, so that fewer details single anyone out.
https://www.torproject.org/

Links checked in November 2026. If an address changes, searching for the title on the organisation’s site usually finds it.

F. The Cyber Welfare Framework: Skills, Awareness, Secure Behaviour

This resource sits on the Awareness pillar at level FL2 – Beginner: seeing anonymity as a crowd size you can estimate, rather than a switch that is on or off.

Skills

• Estimating, with round numbers, how much a detail shrinks the crowd you belong to.
• Spotting which of your public details is the most specific.
• Reading claims of “anonymised” data by asking which details remain.

Awareness

• Understanding the anonymity set: the people who match everything known about you.
• Recognising that ordinary details, combined, can identify you without your name.
• Knowing that looking like many others can protect you as well as sharing less.

Secure Behaviour

• Making one specific detail coarser or less visible, rather than trying to hide everything.
• Sharing time and place after the fact when real-time sharing adds nothing for you.
• Repeating the crowd estimate now and then, because profiles grow quietly.

G. Questions to sit with

• Which detail about you, harmless on its own, would shrink your crowd the most if someone combined it with others?
• Where do you already blend into a large crowd, and where do you stand out?
• When an organisation says it uses your data “anonymously”, what would you want it to tell you?

H. What to do now

The recommendations (R) and security measures (MS) from the Cyber Welfare database that apply most directly here.

1. The details you share yourself

• MS7 — Limit access to your account, photos and posts by changing privacy settings, so that fewer strangers can combine the details you share.
• Prefer coarser details where they serve you just as well: a city rather than a street, a month rather than a date, a trip shared after you are home.

Minimum commitment: today, loosen the one detail you chose in step 4.

2. The details others hold about you

• When a form asks for more than it seems to need, such as a full date of birth for a newsletter, check whether the field is optional before you fill it in.
• When an app or survey says your data is shared “anonymously”, look in its privacy notice for which details are kept: dates, places and age bands tell you how large each record’s crowd might be.

Minimum commitment: the next time you read the word “anonymised” in a privacy notice, look for one sentence on which details remain.

In short

• Details that seem harmless can be combined and linked back to one person, without needing a name.
• Your anonymity set is the crowd you blend into; each detail shrinks it, and details multiply.
• Removing names does not make data anonymous; research has re-identified people from a few ordinary details.
• You can grow your crowd by sharing fewer or coarser details, and by looking like many others.

Related resources in this course

• What Your IP Address Reveals: Location, Provider and Metadata
• How Websites Recognise You: Cookies, Ad IDs and Logins
• Browser Fingerprinting Test: Check Yours with Cover Your Tracks

Discover more companion resources from the online courses of the Protect Your Digital Privacy programme.