The queries, social network and history of movements of an individual are among the most sensible Personal Identifiable Information (PII) and can act as quasi-identifiers. In particular, if an unauthorized entity can access this information, he can use it to perform an inference attack and learn additional knowledge regarding an individual, which may lead to a privacy breach. For instance, from the queries of an individual it is possible to build a precise profile of his interests and even sometimes directly discover his identity (e.g. in case of a "vanity query"). Moreover, from his network of friends, it is possible to deduce information about his hobbies, religious beliefs, political opinions and even sexual preferences.

Simply removing the identifiers of individuals or replacing them by a pseudonym is usually not sufficient to protect their privacy. Therefore privacy researchers have developed more sophisticated techniques for protecting privacy, such as privacy-enhancing technologies (PETs) or sanitization methods which add uncertainty to the data and remove some sensible information. During this talk, I will review some possible inference attacks on query, geolocated and social network data, thus highlighting some of the privacy risks incurred by making this data publicly available. I will also discuss some privacy-preserving mechanisms developped for protecting this data (including some failed attempts) before concluding with an overview of recent developments in the area of privacy.