Computer Science › Section J · Database systems › Topic J.8
Social impact of database technology: privacy
The same technology that lets a shop restock its shelves can assemble a detailed portrait of your life. This topic looks at how that happens and at the laws, pressures and design choices that push back.
Records within easy reach
Information that once sat in scattered paper files is now stored, searchable and easy to combine. A library system can list everything a patron has borrowed; a shop knows every purchase made with a store card; a search engine keeps the questions its users type; a phone records where it has been. Each record may seem harmless, but together they interest marketers, employers, insurers, campaigns and police. Database technology makes collecting and comparing data cheap, so the legal and ethical consequences are everyday matters.
Obvious and hidden collection
Some data collection is open: a survey, a competition entry form, a tax return. Whether giving the data is voluntary can depend on circumstances: filling in a loan application is a choice only if you can manage without the loan. Other collection is quiet. A card company records where and when its customers spend; websites log their visitors' devices and pages; cameras can read the number plates of cars in a car park. A loyalty card is a revealing middle case. The shopper gets discounts, and the shop gets a complete, named record of what that shopper buys, often worth far more to the shop than the discounts it gives away.
The power of linking
Data gains its value, and its danger, when databases are combined. Linking purchases, locations, browsing and public records produces a profile that no single source reveals: a person's likely income, health worries, religion or daily routine. Data brokers are companies that build and sell such profiles. Linking can serve the public, as when agencies cross-check records to catch benefit fraud, but it also turns information given for one purpose to quite another. The relational join of "Querying with SQL" is exactly the tool that does this matching, which is why removing names from a data set does not make it anonymous.
Worked example: re-identifying "anonymous" data
A clinic publishes research data with names removed: Clinic(Zip, Born, Sex, Diagnosis). A public voter list holds Voter(Name, Zip, Born, Sex). Steps: (1) Notice the shared columns: ZIP code, birth date and sex. (2) Join on all three: SELECT Voter.Name, Clinic.Diagnosis FROM Clinic JOIN Voter ON Clinic.Zip = Voter.Zip AND Clinic.Born = Voter.Born AND Clinic.Sex = Voter.Sex; (3) On a small test, three clinic rows and three voters, this returns two rows, naming Dana Cole with asthma and Sam Ortiz with diabetes. (4) The third patient matches no voter and stays hidden. In well-known research, computer scientist Latanya Sweeney estimated that most Americans, about 87%, could be singled out by those three facts alone.
Metadata and surveillance
Data about communications can be as revealing as the communications themselves. In 2013, documents leaked by Edward Snowden showed that the US National Security Agency had been collecting records of Americans' phone calls in bulk: which numbers called which, when and for how long, but not what was said. Supporters argued that no conversations were recorded. Critics replied that call records alone can expose a person's doctor, lawyer, religious community or relationships. Metadata, met in "Database fundamentals", can reveal a great deal.
When the data is wrong
Privacy problems are serious when records are accurate and worse when they are not. An error in one database, such as a debt wrongly recorded on a credit report, can spread to every organisation that buys or shares that data, and automated systems may then refuse a loan, a flat or a job without anyone checking. In the United States, the Fair Credit Reporting Act of 1970 gives people the right to see and dispute information held by consumer reporting agencies. The general lesson applies everywhere: people need ways to find out what is held about them and to correct it.
Law in the United States
As of 2026, US privacy law is a patchwork of laws for particular sectors rather than one law for all personal data. The Privacy Act of 1974 covers federal agencies: they must publish notice of their record systems in the Federal Register and let people see and correct their own records. HIPAA, from 1996, protects health information held by health plans, most health-care providers and the businesses working for them, but not most consumer apps. FERPA (1974) covers student education records, and COPPA (1998) limits online data collection from children under 13. California's Consumer Privacy Act, passed in 2018 and later expanded, gives residents rights to know about, delete and stop the sale of their data, and other states have followed.
The EU's GDPR
Across the EU, the General Data Protection Regulation (GDPR) has applied since May 2018 to organisations handling personal data of people in the EU, including many based elsewhere. Processing needs a lawful basis, of which consent is one. Its principles include purpose limitation, using data only for the purposes it was collected for, and data minimisation, collecting no more than is needed. People gain rights to access their data, correct it, have it erased in some circumstances, take it to another provider and object to some uses. The most serious breaches can bring fines of up to €20 million or 4% of a company's worldwide annual turnover, whichever is higher. Laws change; check current rules.
Public opinion and privacy by design
Laws make abuse illegal; they do not stop it, and enforcement is slow. Public opinion can act faster, because companies fear damage to their reputation. In 2010 Google launched Buzz, a social tool that by default made public the people many Gmail users contacted most; after fierce criticism it was changed within days, and it closed in 2011. The strongest protection, though, is built in from the start, an approach called privacy by design: collect the minimum, make sharing opt-in rather than opt-out, keep data only as long as needed, and protect what is kept, as "Cybersecurity: encryption and authentication" describes.
Worked example: reviewing a fitness app
A running app publishes every user's routes on a public map by default. Steps: (1) Identify the data: GPS tracks, times and a profile name. (2) Identify the harm: routes starting and ending at the same spot reveal a home address, and regular times reveal when the home is empty. (3) Check the law: for users in the EU, GDPR requires a lawful basis and data minimisation, and public-by-default sharing is hard to justify; in the US, HIPAA generally does not cover such an app. (4) Redesign: make sharing opt-in, hide the first and last few hundred metres of each route, explain plainly what is shared and let users delete their history.
Common mistakes to avoid
Assuming data without names is anonymous; linking on a few ordinary attributes can re-identify people. Thinking metadata is harmless because it lacks content. Believing a privacy law, once passed, ends the abuse it forbids. Stating that HIPAA protects all health data in the US; it covers specific organisations, not most apps. Treating GDPR as applying only to companies based in Europe; it can reach organisations elsewhere that handle data of people in the EU. Forgetting that inaccurate data does extra harm once it is shared.
Key takeaways
Database technology makes collecting, storing and linking personal data cheap, and linking reveals far more than any one source. Removing names does not guarantee anonymity, and metadata can expose habits and relationships. Errors spread when data is shared, so people need rights to see and correct records. In the US, privacy law is sector by sector: the Privacy Act, FCRA, HIPAA, FERPA, COPPA and state laws such as California's. In the EU, GDPR sets principles such as data minimisation and gives individuals strong rights. Public opinion and privacy by design are powerful complements to law.
Flashcards
Flashcard
People's ability to control information about themselves
Test your understanding
Question 1 of 5
A research file lists ZIP code, birth date, sex and diagnosis but no names. Why might it still identify people?
Question 2 of 5
Which statement about HIPAA is accurate?
Question 3 of 5
A company collects users' birth dates, contacts and locations 'in case they are useful later'. Which GDPR principle does this most directly break?
Question 4 of 5
Why can phone-call metadata raise privacy concerns even without recordings?
Question 5 of 5
Which change best reflects privacy by design for an app that shares runs publicly by default?
