Database Reference
In-Depth Information
While several social sensing applications are already deployed, ex-
citing research opportunities remain in order to help understand their
emergent behavior, optimize their performance, redesign the networks
on which they run, and provide guarantees to the user, such as those on
bounding unwanted information leakage.
4. Privacy Issues in Social Sensing
Social sensing offers interesting new challenges pertaining to privacy
assurances on data. General research on privacy typically focuses on
electronic communication as opposed to ramifications of increasing sen-
sory instrumentation in a socio-physical world. In contrast, traditional
embedded systems research typically considers computing systems that
interact with physical and engineering artifacts and belong to the same
trust domain. A need arises to bridge the gap in privacy research by
formulating and solving privacy-motivated research challenges in the
emerging social sensing systems, where users interact in the context of
social networks with embedded sensing devices in the physical world.
Sharing sensor data creates new opportunities for loss of privacy
(and new privacy attacks) that exploit physical-side channels or a priori
known information about the physical environment. Research is needed
on both privacy specification and enforcement to put such specification
and enforcement on solid analytic foundations, much like specification
and enforcement of safety requirements of high-confidence software.
Specification calls for new physical privacy specification interfaces that
are easy to understand and use for the non-expert. Enforcement calls for
two complementary types of privacy mechanisms; (i) protection mech-
anisms from involuntary physical exposure , and (ii) control of volun-
tary information sharing . The former enforce physical privacy .They
are needed to prevent “side-channel” attacks that exploit physical and
spatio-temporal properties, characteristic of embedded sensing systems,
to make inferences regarding private information. Control of voluntary
information sharing must facilitate privacy-preserving exchange of time-
series data. A predominant use of data in social sensing applications is
for aggregation purposes such as computing statistical information from
many sources. Mathematically-based data perturbation and anonymiza-
tion schemes are needed to hide user data but allow fusion operations
on perturbed or partial data to return correct results to a high degree
of approximation.
While privacy-preserving statistics and privacy-preserving data min-
ing are mature fields with a significant amount of prior research, shar-
ing of sensor data offers the additional challenge of dealing with cor-
Search WWH ::




Custom Search