Summary
In an Internet of Things environment, the exploitation of data for prediction and recommendation purposes requires the aggregation of data of a different nature on centralized servers. This centralization causes the data to be exposed unnecessarily, forcing in some cases their owners to transfer them to obtain the required service. Likewise, when the response requires immediacy, centralization is also a problem. This proposal aims to develop a framework (based on the concept of osmotic computing and digital avatar) that guarantees the use of data in the layer in which they are generated, avoiding their centralized manipulation and therefore, its exposition and adapting the generation of the response to the context in which it is required.
Objectives
The Internet of Things (IoT) model is based on the arrangement of various devices and sensors
in our environment so that all of them offer an ecosystem that allows them to connect with each other, exchange
data and gain added value from their interaction. As a result of this phenomenon, what we call intelligent objects
are born, since they allow to be adapted according to the information available in real time. The data collected by
these devices can be displayed as open data, with the possibility of their reuse, as public data, accessible but with
some limitation (either of time or scope), or as private data, generated by citizens or owners who do not wish to lose control over them.
Until now, applications have chosen to exploit the prediction and recommendation capabilities of artificial intelligence, accessing all data
(without distinguishing its origin: open, public or private) through a centralized server, making private data exposed and the owner
(citizen) no longer have control over them.
This proposal aims to validate a proof of concept of a framework that guarantees the use of data at the level at which they are generated, avoiding the centralization (and, therefore, exposure) of them. To do this, we will use the notion of digital avatar, as a way to represent the data and behavior that each device exhibits. In this way, the system will have to decide autonomously and dynamically (osmotic computing) which data to use locally (within the devices themselves, edge computing), which centrally (cloud computing) and which in an intermediate environment (fog computing). This should take into account physical restrictions (network latency, device computing capacity or power limitations) and privacy considerations.
