Abstract:
The rapid increase in popularity of location-based services have resulted in huge amount of
spatial textual data being generated by applications like Foursquare, Facebook Places, Flickr
etc. The location-based services o er convenience but threaten the location and query privacy of
the user. The data collected by such servers can be to used study user behaviour or for stalking
personal locations. A novel query which became popular in the past few years is Reverse k
Nearest Neighbour Query (RkNN). Given a set of database objects O and a query point Q, the
RkNN query returns those objects o 2 O, for which Q is one of its kth nearest neighbour, using an
appropriately de ned similarity function on pairs of database objects. We propose a generalized
framework for nding the reverse nearest neighbours of a query point which is independent
of the underlying hierarchical indexing structure used as well as the used similarity measure.
Our framework is independent of the type of database objects, but the only requirement is to
de ne lower and upper bound similarity between any two object/groups of objects E and E0
of the given index structure and calculate the number of objects for every group of objects.
We present two di erent approaches, namely, Lazy and Eager for performing monochromatic
Reverse Nearest Neighbour query on spatial textual data. We conduct extensive experiments
on real datasets and study the performance of both approaches. We address the problem of
performing Reverse Nearest Neighbour (RkNN) search while preserving the location privacy of
a user. Location Privacy can be preserved by anonymizing the location of a user using techniques
like k-anonymity[1] or l-diversity[2]. The idea is to send a cloaked region to the server instead
of the user's exact location so that location privacy is preserved. We formalize the problem
of performing Reverse Nearest Neighbour Search on spatial objects when the exact location of
database objects is not known to the server. A key challenge in performing such queries is to
strike a balance between maintaining the correctness of results versus maintaining the privacy
of a user.