Database Reference
In-Depth Information
analysis of this kind of data can result in identifying person's movements without
the person being aware of this identification. In extreme cases, combining trajecto-
ries of an individual with external knowledge (e.g. the address of their home and
workplace) may identify a person uniquely from a large set of trajectories, ob-
tained from a huge population. Early work in this area has been a topic of a fo-
cused research project (Giannotti, Pedreschi et al. 2009). Some of the techniques
proposed aggregate trajectories into groups before releasing them for data mining,
in the spirit of
k
-anonymization applied on “static” data and described above.
Another enormous challenge is the growing universal use of social networks.
Clearly, there is a basic contradiction between privacy and the goal of social net-
works, which is to present information about a person, their opinions and their ac-
tivities. From a Computer Science perspective, social networks are often described
and analyzed as graphs. There exists a body of recent literature exploiting social
network mining as graph data mining, and proposing techniques that make unique
identification of a person from purely
structural
information hard. Numerous pa-
pers follow this approach, see e.g. (Liu and Terzi 2008) and (Hay, Miklau et al.
2008). Real social networks, however, supply a wealth of non-structural informa-
tion - names, photos, email addresses etc. - which can be used as explicit identifi-
ers. Therefore the practical value of graph-based social network privacy protection
research remains to be proven.
As no technical solutions for protecting privacy in social networks exist, it
seems this is not a purely technical problem. Perhaps the main tool to mitigate
potentially disastrous effects of social networks for privacy remains education.
Users, especially the teenage population, need to be explained the basic facts, e.g.
that posting anonymous photos of people for the world to see may cause automatic
tagging and identification of people in the photos. Oftentimes, many privacy
breaches could be prevented if the users of social networks were taking advantage
of setting privacy of their personal information using the existing privacy settings,
provided by social networks. For instance, users may allow only their direct
friends to see their tagged photos. Most users, however, never learn about these
privacy settings and never use them. In that realm, novel work by (Fang, Kim et
al. 2010) seems very interesting. The gist of it is to give users tools based on ma-
chine learning and recommender systems, and that make it relatively painless to
set the existing privacy settings in social networks such as Facebook. It is some-
thing most users are not doing, and it would protect against many privacy breaches
by limiting access to information the users provide.
Finally, cloud computing is a major challenge for data security, and hence data
privacy. In a cloud the data owner lose control over their data. The existing legal
safeguards are jurisdictional, and the cloud makes it hard, if possible at all, to de-
termine where the data resides and where is it processed, and therefore which legal
constraints - if any - on collecting, storing, and using the data apply. It has to be
observed, however, that if the research initiated by the paper (Gentry 2010) suc-
ceeds, it could provide a comprehensive solution for the privacy issues in a cloud
setting.
Search WWH ::
Custom Search