Privacy Dependencies
summary
Argues that privacy about one person often depends on others, and gives a three‑part taxonomy: tie‑based (familial searches like Golden State Killer via 23andMe), similarity‑based (machine‑learning look‑alike audiences and demographic inference), and difference‑based (anomaly and elimination attacks and adverse inference). Shows tools used—DNA databases, social graphs, supervised ML—and cites surprising scale effects (small early disclosers enable population‑wide inference; combining signals can reidentify >70% of the U.S.). Mentions the Devious Defecator/GINA case as a legal surprise.
AI-written summary · about
chapters (19)
- 0:00Intro
- 2:41BEYOND INDIVIDUAL PRIVACY
- 5:04TYPES OF DEPENDENCIES
- 6:48TIE-BASED DEPENDENCIES
- 9:04INCIDENTAL OBSERVATION
- 10:074. INCENTIVIZED DISCLOSURE
- 12:43INDUCING A RULE
- 13:02APPLYING THE RULE
- 15:33IRREPRESSIBLY INDUCTIVE
- 17:46AVOIDING THE COSTS
- 19:29RELEVANTLY SIMILAR
- 34:10LIKE PARENT, LIKE CHILD
- 38:18THE COMPANY YOU KEEP
- 41:044. NON-SOCIALLY SALIENT GROUP
- 42:06PROCESS OF ELIMINATION
- 42:432. ANOMALY DETECTION
- 43:263. ADVERSE INFERENCE
- 46:07GENETIC DEPENDENCIES
- 48:53GENETIC INFORMATION NONDISCRIMINATION ACT
related talks
01h 10mENCODING IDENTITIES: THE CASE OF COMMERCIAL DNA DATABASE
01h 43mPerspective
01h 16mLaw and CS
43mGuarding the ML Galaxy: Beyond Accuracy to Privacy and Security
05mPrivacy, identity, and autonomy in the age of big data and AI
01h 17mTutorial: Privacy, Part I
29mAnti-Privacy Anti-Patterns
28mBiological Discovery and Consumer Genomics Databases Activate Latent Privacy Risk in...