Database Reference
In-Depth Information
user inside a hospital must be less than 0.1. Next, coarse regions are generated
satisfying the privacy preferences, independently from the user's position, in
order to prevent possible inferences on their reciprocal positions. A sample set
of obfuscated regions is shown in Figure
2.9
b. Finally, at runtime if the user's
position falls inside one of the coarse regions, that region is delivered instead of
the exact position. This solution is grounded on a conceptually founded privacy
metric. Moreover, an additional metric is defined, the utility metric, providing a
measure of the spatial accuracy of the cloaked regions. Unlike more traditional
obfuscation techniques, the utility measure can be computed prior to any service
request. In this way users can tune and balance the amount of privacy with the
quality of service.
2.6 Conclusions
In this chapter, we presented techniques for collecting mobility data and han-
dling them appropriately (applying data cleansing, data compression, and map
matching) so as to produce noise-free and meaningful trajectories (trajectory
reconstruction). Finally, privacy issues in mobility data collection and handling
were discussed.
We outline next a few research directions that originate in the discussion
provided in this chapter.
With respect to
trajectory reconstruction
, future work may include the explo-
ration of intelligent ways to automatically extract proper values of trajectory
reconstruction parameters according to a number of characteristics of data sets,
as well as the extension of this technique so as to be able to identify different
movement types (pedestrian, bicycle, motorbike, car, truck, etc.) and hence to
apply customized trajectory reconstruction.
With respect to
privacy issues
, major research directions include
privacy
usability
, that is, how to provide personalizable, conceptually founded, and
simple-to-use privacy mechanisms so to enhance user experience; and
context-
aware location privacy
, that is, tailoring privacy protection based on the context
in which individuals are located. While semantic location privacy is a first
attempt to introduce the contextual dimension in privacy, this notion can be
extended along several directions; for example, to account for the temporal and
social dimension of privacy.
2.7 Bibliographic Notes
In this section, we distinguish and annotate some works from the literature.
With regard to the data-handling approaches,
Ya n e t a l .
(
2010
) proposed
a Gaussian kernel-based local regression model to smooth out GPS feeds.
Brakatsoulas et al.
(
2005
) proposed the methodology for map matching that is
