A differentially private algorithm for location data release
Web communities are making available an increasing volume of free, fresh, detailed and powerful information about living people. Among them, the Flickr photo-sharing service offers to researchers a database of several millions of geotagged pictures from users all around the world. Working on that opens the door to the study of meaningful mobility data, where title and description of a geotagged picture represent a mine from which extract labels to detect places and events, and useful information about user trends, behaviors and tastes. Our approach goes in the direction of developing an intelligence and unattended system able to extract and take advantage of up-to-date and spontaneous information embedded with pictures, making cities intelligent and able to reach user expectations. Such system, learning from past touristic user experiences, could make customized recommendations on “where to go”, and “what to see”, to people going to visit touristic places for the first time.