|
Abstract:
|
Statistical estimation procedures can be viewed as (complex) queries to a database. When data are treated as a commodity, it may be necessary to quantify several properties of these queries. First, what is the risk that the query answers, combined with inference channels, will reveal confidential information (this risk is quantified using privacy definitions)? Second, what price should the data owner set for the answer to a query?
In this talk, we reveal structural connections between privacy definitions and arbitrage-free pricing functions, and show how these different points of view can be used to extend privacy definitions to domains such as smart camera privacy.
|
