manos/the-last-thing
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

summary: FInished the perspectives

Browse Source
This commit is contained in:
Manos Katsomallos 2021-10-25 17:48:02 +02:00
parent 4d828db98e
commit a1b3eda5d8
1 changed files with 31 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

43
text/conclusion/perspectives.tex
View File

@ -1,18 +1,37 @@
\section{Perspectives}
\label{sec:persp}
\mk{WIP}
\paragraph{Global scheme}
statistical data
\paragraph{Streaming mode}
\paragraph{Global {\thething} privacy}
For now, we have applied {\thething} privacy in the local scheme and for microdata due to the advantages of the local scheme over the global as we discussed in detail in Section~\ref{subsec:data-publishing}.
The adaptation of {\thething} privacy to support the global processing and publishing scheme would allow for the studying of more diverse scenarios including statistical data publishing.
\paragraph{Event categories}
\paragraph{Spatiotemporal continuity}
\paragraph{More correlations}
\paragraph{{\Thething} privacy over infinite event sequences}
So far, we considered for our problem setting finite time series that are processed in batch mode.
This was a decision that we made for the shake of simplicity in order to facilitate a more straightforward definition of {\thething} privacy.
In the future, we plan to explore more dynamic scenarios where data are processed and published in streaming mode, which will lead to adoption of time critical crowdsensing applications.
In the future, we aim to work on automatically learning the initial landmark set by analyzing the input data sets,semantics, and user preferences.
We also plan to introduce learning for the tuning of ourAdaptive scheme parameters.
\paragraph{{\Thething} privacy and spatiotemporal continuity}
In mereology, the formal study on the relation between parts and the entities they form, it is generally held that the identity of an observable object depends to its \emph{spatiotemporal continuity}~\cite{wiggins1967identity, scaltsas1981identity, hazarika2001qualitative}.
That is, the property of well-behaved objects that alter their state in harmony with space and time.
Considering events that span the entirety of the user-generated series of events thereof ensures the spatiotemporal continuity of the users.
This way, it is possible to acquire more information regarding individuals' identities, and thus design privacy schemes that offer improved privacy and utility guarantees.
\paragraph{Diversification of event categories}
With the proposal of {\thething} privacy we introduced {\thething} events in privacy-preserving continuous data publishing.
This categorization in regular and significant events enabled the development of a configurable differential privacy notion for time series.
Variation of the existing event categories, e.g.,~weighted {\thethings}, or the introduction of new ones, would allow for an even more fine-grained configuration of privacy protection and the development of variations of {\thething} privacy.
\paragraph{More data correlation types}
In the current state of our work, we consider {\thethings} as one-dimensional elements in our problem setting.
Consequently, we have explored {\thething} privacy under temporal correlation and examined the behavior of temporal privacy loss for different {\thething} percentages and distributions.
Accounting for other possible dimensions, e.g.,~location, can introduce more aspects to the current use case of {\thething} privacy.
Indicatively, as we have extensively studied in Section~\ref{sec:correlation}, there are many types of data correlation in time series to further research in the context of {\thething} privacy.
\paragraph{Incorporation of machine learning}
Until now, we consider the {\thething} discovery and selection process orthogonal to our work.
In the future, we aim to work on automatically learning the initial {\thething} set by analyzing the input data sets, semantics, and user preferences.
We also plan to introduce learning for the tuning of our \texttt{Adaptive} scheme parameters, which will further improve its sampling component and overall utility performance.