Treat identifiability as a threat model
Explain why synthetic data privacy depends on quasi-identifiers, sensitive fields, and recipient context, not only direct identifiers.
Identifiability is a threat model. It depends on fields, combinations, recipient knowledge, and release context. Four things to inspect Direct identifiers: explicit links to a person or account, such as names, emails, phone numbers, device IDs, or production IDs. Quasi-identifiers: fields that look harmless alone but identify people in combination, such as age, location, dates, role, tenure, or rare event history. Sensitive fields: the protected facts, such as diagnosis, salary, fraud status, complaint reason, or performance rating. Recipient knowledge: what the recipient can combine with the synthetic file, including customer lists, public records, or insider knowledge. Why it works Synthetic…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in