8. Thinking Like the Adversary

Module 4: Risk Assessment · Runtime 45:41 · YouTube title: Thinking Like the Adversary: HIPAA Re-identification Risk

You cannot measure re-identification risk until you name the adversary. The risk equation, overall risk = P(attempt) × P(success | attempt), is introduced and then actually multiplied out. Prosecutor, journalist, and marketer risk are each computed, population class sizes are estimated responsibly, and real attack scenarios are mapped onto the right model.

In this video

• The risk equation and the two levers: context lowers the probability of an attempt, data properties set the probability of success
• What moves attempt probability: public release, contract terms, recipient motive, security posture, payload sensitivity
• Prosecutor risk (target known to be in the data): maximum risk 1/smallest k, Springfield raw = 1.0, mean 0.90
• Journalist risk (membership unknown): anchored to population class K; computed on a 100-of-10,000 sample, 0.20 versus prosecutor 1.0
• Marketer risk (bulk, average): 0.12 on the same file, and the hierarchy prosecutor ≥ journalist ≥ marketer
• Estimator rule: sampling fraction below 10 percent uses Pitman, at or above uses Zayatz
• Reasonably available information: voter files, Census and ACS tables, data brokers
• Maximum versus average risk, how tools compute it by Monte Carlo, and attacks beyond the models
• Scenario to model to metric, and the full equation evaluated: public 1.0 × 0.09 = 0.09, enclave 0.05 × 0.20 = 0.01

Worked example

Springfield (prosecutor and marketer risk) plus a second 100-record toy sample for journalist risk.