Privacy is being increasingly eroded as technological advances make it possible to capture, store, and process large amounts of personal data. In this talk, I will overview the current privacy landscape and outline some of the important problems in this area. I will also describe some of my work in this area, on privacy-protecting statistical analysis.

Suppose a client wishes to compute some aggregate statistics on a privately-owned data base. The data owner wants to protect the privacy of the personal information in the data base, while the client does not want to reveal his selection criteria. Privacy-protecting statistical analysis allows the client and data owner to interact in such a way that the client learns the desired aggregate statistics, but does not learn anything further about the data; the data owner leans nothing about the client's query.

Motivated by this application, we consider the more general problem of "selective private function evaluation," in which a client can privately compute an arbitrary function over a database. We present various approaches for constructing efficient selective private function evaluation protocols, both for the general problem and for privacy-protecting statistical analysis.