Usually, to retrieve an element from a database, a user will send a
query pointing out which element he wants to obtain, and the database
will send back the requested element. Which element a user is interested
in may be an information he would like to keep secret, even from the
database administrators. For example, the database may be :

– an electronic library, and which books we are interested in may
provide information about our politic or religious beliefs, or other
details about our personality it may be desirable to keep confidential,
– stock exchange share prices, and the clients may be investors
reluctant to divulge which share they are interested in,
– a pharmaceutical database, and some client laboratories may wish that
nobody may learn which are the active principles they may want to use,

To protect his privacy, a user accessing a database may therefore want
to retrieve an element without revealing which element he is interested
in. A trivial solution is for the user to download the entire database
and retrieve locally the element he wants to obtain. This is usually
unacceptable if the database is too large (for example, an electronic
library), quickly obsolete (for example, stock exchange share prices),
or confidential (for example, a pharmaceutical database).

Private Information Retrieval (PIR for short) schemes aim to provide the
same confidentiality to the user (with regard to the requested element)
than downloading the entire database, with sub-linear communication
cost. This research domain is split in two depending on whether the
database is supposed to be replicated or not. We will first quickly
introduce the approach of replicated-database protocols, and then we
will focus on single-database protocols. After a presentation of some
approaches we will describe innovative applications that highlight the
large span of applications in which these protocols may be used.