Quasi-Identifier

What is a quasi-identifier?

A quasi-identifier is a piece of information that does not identify a person on its own but can identify them when combined with other data. ZIP code, date of birth, gender, job title and hospital admission date are common examples. Quasi-identifiers are the main reason that removing names alone does not make a dataset anonymous.

For example, a hospital shares a research extract with names and record numbers removed. The extract still lists each patient’s ZIP code, birth date and sex. A researcher matches those three fields against a public voter list, puts names back on several records and can now read those patients’ diagnoses.

Why combinations identify people

Each quasi-identifier on its own is shared by many people. Thousands of people have the same birthday, and thousands live in the same ZIP code. The combination is what narrows it down. Latanya Sweeney’s research on 1990 US census data estimated that about 87% of the US population could be uniquely identified by five-digit ZIP code, gender and full date of birth.

The risk grows with every extra field and with every outside dataset an attacker can match against, such as voter lists, social media profiles or data broker files.

Direct identifiers vs quasi-identifiers

Direct identifiers Quasi-identifiers
Examples Name, Social Security number, email, phone number, medical record number ZIP code, date of birth, gender, job title, employer, dates of service
Identifies a person alone Yes No, only in combination
Usual treatment Remove or tokenize Generalize, suppress or tokenize, based on risk

How teams reduce quasi-identifier risk

  • Generalization: keep the birth year instead of the full date, or the first three digits of a ZIP code instead of all five.
  • Suppression: drop rare records that stand out, such as the only 97-year-old patient in a small town.
  • k-anonymity checks: confirm that each combination of quasi-identifiers is shared by at least k records.
  • HIPAA Safe Harbor: removes all date elements except the year and geographic units smaller than a state, with a narrow exception for three-digit ZIP codes covering more than 20,000 people. When more detail must stay, an expert can certify that re-identification risk is very small under HIPAA Expert Determination.

Quasi-identifiers in AI systems

In AI workflows, quasi-identifiers hide in free text. A prompt that says “the 34-year-old CFO of our Tulsa branch” names no one, yet most colleagues would know who it means. Support tickets, call transcripts and documents pulled into RAG are full of these details. Language models are good at linking clues, and an AI agent that calls several tools can combine details from different systems in one answer. Column rules written for structured data do not see any of this.

How Protecto helps

Protecto’s agentic detection and classification engine reads context, so it finds dates of birth, addresses and other identifying details inside free text as well as in labeled columns. Teams can add details specific to their business, such as job titles or internal IDs, as custom entities in days. Detected values become consistent tokens such as <DOB>tok_4c1e</DOB>, so models keep the structure of the record. See PHI de-identification for data lakes.

Frequently asked questions

Is a ZIP code personal information?

Often, yes. A ZIP code alone rarely identifies someone, but combined with a birth date or gender it can. Laws such as GDPR and CCPA cover any information that can reasonably be linked to a person, so quasi-identifiers usually count.

What is the difference between an identifier and a quasi-identifier?

A direct identifier, such as a name or Social Security number, points to one person by itself. A quasi-identifier only identifies someone when combined with other fields.

How do you protect quasi-identifiers?

Generalize them (birth year instead of birth date), suppress rare records, or tokenize them. Then test the result, for example with a k-anonymity check, before sharing the data.

Related terms

k-anonymity · Data anonymization · Data de-identification · HIPAA Safe Harbor · HIPAA Expert Determination · Personally identifiable information (PII)

Book a demo

See Protecto on your own data

Book a demo to see how Protecto finds and masks sensitive data before it reaches any AI model.

New: Protecto now runs inside LiteLLM. Control what sensitive data your models and tools see.
Read the launch blog