The Internet of Things (IoT) world is composed by a huge number of di erent so called \smart devices" and every year new and di erent models are released on the mass market. Most of those devices are intended to be used by professional people or by companies. Thanks to the constant growth of the \smart objects", end users and people with low or no-knowledge of the IT-world get in touch with these pieces of technology. Those people are expected to use the smart devices \out of the box" and in a very simple and easy way so, the human-device interaction needs to be as easiest as possible. Despite of this need, end users are commonly faced with thousand of di erent technological standards which are hard to evaluate without a solid IT background. Thus, the comparison to understand which IoT device performs better in a particular situation become complicated. In this paper we propose a comparison of two di erent IoT solutions using an IoT model. The model assesses the di erent technical speci cations of the devices and then extracts a \score" for each technological aspect. The end user can use the score to better understand the points of strength and the weaknesses of the devices.
The mass market o ers to expert and non-expert users thousands of di erent
Internet of Things devices. Those devices are generally composed by sensors,
electronics boards and antennas that allows them to communicate with the \outside"
world. Due to the enormous number of combinations of those three components,
the number of the available devices gets higher and higher. Hardware,
programming, security, protocols etc. are terms that often are not clear to the end user
that has to face the choice among multiple IoT devices. Indeed, the complexity
of the choice increases when multiple objects can be used in multiple application
contexts. Often, speci c devices that are built to work in certain domains, thanks
to their characteristics, can be also used in more and di erent contexts. For
example, a temperature monitoring device for the home environment can be also
used in a pharmacy store to check and control the temperature of the medicine
room, using a \sink" node to transmit the data over the Internet. To correctly
con gure an ecosystem made by multiple di erent devices, end users need to
understand most of the device's characteristics [
In this paper we rst propose a model to assess smart devices in order to give a way to assess the suitability of a speci c smart device to a certain application domain. Each device characteristic can be assessed and compared in order to choose the one that best satis es the end user needs. We also report a rst validation of the model by comparing two di erent smart devices. 2
The proposed model has been developed by analyzing the characteristics of a
number of di erent devices. It is shaped as a star diagram (Figure 1). A rst
version has been proposed in [
The rst one represents the device communication element and concerns how the data are transmitted by the device. The \Target" element considers the context of use of the device, and it is associated to the device hardware implementation. The \Data Manipulation" element covers all the characteristics related to the data generated and managed by the device. Finally, the \Development" element analyses the programming characteristics of the device. \Target" and \Development" are detailed with a group of more sub-elements that help the assessment of the device. 2.1
The Communication element represents how the communication between the device and the external world occurs. It analyses the technical aspect of many device's characteristics, such as: the Security, the communication Protocol, and the Destination of the communication.
The Cryptography is an aspect related to the security of the communication.
Cypher data allow them to be sent and received over the net, keeping the content
safe from unauthorized users [
The Protocol is related to the application protocol in the communication. All the application protocols generate some overhead in the communication, as they add more \control" information to be sure that the data reach the receiver uncorrupted. To receive a good assessment in this element, a low-overhead protocol must be used. For example, COAP (COnstrained Application Protocol) is 81
Firmware / Low level design
Professional
Developer Cryptography Destination Protocol
Communication
API Domain
Experts Development
IoT device Target
User Engagement Functionatlities
addition Customization
End Users
Data
Manipulation Personal
Professional
Mixed Hardware Communication type
Microcontroller
Cost Size
Standard Compliant
Accuracy Remote
Application
preferred to HTTP (Hyper Text Transfer Protocol) since it generates a very low
overhead [
The Destination is related to \where" the data are sent. This is a critical aspect as it makes the di erence in an IoT ecosystem when multiple similar device communicate each other at the same time. Sending the data directly to a remote server may generate excessive tra c and interference on the network, thus it is important that a coordination (such as the usage of a sink node) is present. 2.2
The Target element represents the main usage of the device and is related to
the hardware characteristics of the device. We have classi ed the Target use in
three main area, according its use and/or deploy: Personal, Professional and
Mixed. All those aspects consider, in the assessment process, the hardware
characteristics, including the physical communication type. As a general guideline
for the assessment process, an hardware that uses low energy to work, supports
a communication type IoT-compliant (low transfer rate and low power) and is
usually exploited in multiple contexts, performs better than the others [
The Personal element considers some technological and commercial aspects of the device such as the cost and the size, two valuable aspects in the assessment process. The Mixed element considers the cost and the possibility to use a remote application combined to the device. The Professional element assesses the accuracy of the sensor used on the device and its \IoT standard compliant".
Each of the above elements of the target perspective, share one more subelement: Hardware communication type, Microcontroller, Cost, Accuracy, Size, Remote Application and Standard compliant. While Hardware communication type sub-element refers to the low-level technology used by the device to communicate, the Microcontroller assesses the type of the hardware used. The device cost is particularly important for personal and mixed usages, while for professionals, the accuracy in its measures should be considered. Personal devices should be small in size in order to t everywhere, while for a mixed usage, the possibility to have a remote application is an important factor. Finally, for professionals it is imperative to use standard-compliant devices. 2.3
The Data Manipulation element is related to the data, from the Data Collection
and Data Transformation, to the Visualization involving also the Data Privacy.
This part is critical as it is the connection between the device low level sensor
and the \outside world" represented by mobile applications, servers, services etc.
[
The Data Collection is an important aspect related to how the data are
temporary stored into the device. There are two di erent approaches to this
aspect: the bu ering (which is the temporary store of multiple data instances)
83
or the \immediate send", which involve the transmission of the data as soon as
it has been produced. The rst one is less energy consuming [
The Data Transformation involves the conversion or the enrichment of the data. These kind of operations are supposed to be made on an external server rather than on the device itself.
The Data Visualization analyses the aspect of the representation of
information on the device. To save energy, the device should not show anything on
it, but it should send the data to a remote service. An example of visualization
available on a web server is CBP, that shows home energy consumption detected
by IoT devices [
The aspect of the Data Privacy is related to the type of the data collected
by the device [
The Development element involves the possibility of the device to be programmed by people. It has three target users: the Professional Developers, Domain Expert and End Users.
Professional Developers are computer-literate people that are familiar with
programming languages, low-level programming and model driven testing [
Domain expert users are people familiar with the IT world. They know how to program, thus they can use the API to develop a small piece of software that is able to perform the required need.
End users are not familiar with the computer or (in general) with the
technology. They are not able to program, thus they need to be supported by possibly
visual paradigms (such as ATOOMA, IFTTT, Tasker or EFESTO)[
This element has multiple sub-elements: Firmware/Low level design, API,
Customization, Functionalities addition and User engagement. The former
belongs to the world of Professional Developers: as their ability to program with
low level languages, the source of the project and the possibility to edit the
\core" of the system is essential. Domain experts, can use the API to program
their solution to their own needs. End Users need to Customize and add or
remove some functionalities on their own. The user engagement is important as
well [
The aim of this model is to give users a way to assess one or more di erent IoT devices. By the assessment the end user can detect the right trade o between 84
all the characteristics of the devices and chose which one suits better to his/her
needs. The domain experts take advantage of the model by having under control
all the devices characteristics at a glance, even if they are not technical experts.
This is an evolution of the work of presented in [
In this section we propose a comparison between two devices: the Sono SC ambient analyzer and an home-made air and light analyzer. The rst one is an all-in-one device, which has all the required components inside the plastic case. It has an external power supply (an USB power plug) and a slot for a memory card to store all the data collected by the sensors.
The home-made air and light analyzer is composed by two parts: a Raspberry PI 3 and an Arduino board with many sensors (such as humidity, CO2, light sensors) connected to it. The communication between the Raspberry and the Arduino uses the Bluetooth 4 technology. The Raspberry can formally be connected with other Arduino boards and act as an external \sink", where all the data get collected and then sent to the external web server.
To assess all the device characteristics, we propose a score (s) from 0 to n (with n > 0) to assess the elements and the (optional) sub-elements of the model as follows: { s = 0: the characteristic is not supported by the device; { [0 < s < n]: the device partially supports the characteristic (e.g.: it adopts a technology, but a new and more performing technology is available); { s = n: the device fully supports the characteristic in terms of both functionalities and recent technology.
In this assessment, for the sake of simplicity, we set n = 2.
Table 1 shows the assessment for the Sono SC ambient analyzer, while Table shows 2 the assessment of the home-made air and light analyzer has been shown. For the sake of simplicity, each element and sub-element has been named according to a \level", which is the distance from the center of the star. At a rst sight, the two devices are very similar in terms of performances, price and technology. Both of them can be used in a personal and semi-professional environmental setting.
The assessment of the Communication element is similar. The home-made device, supporting the sink node, performs slightly better than the Sono and the assessment of the Target element is almost the same as the previous one. Thanks to the Bluetooth 4 technology, the home-made device has a better assessment than the Sono .
The Data Manipulation is performed better by the Sono device, thanks to the possibility to bu erize the data and the absence of any LED or display that shows the data sent. Similarly, the Sono device allows a little bit of personalization of the behavior of the IoT device: thanks to the graphical programming application, it is possible to de ne rules and send alerts to the users, if needed.
The overall results from the assessment of those two devices shows that they are almost equivalent in all the elds. The home-made device is more industry oriented, while the Sono is better to be used in an home environment. They 86 resides in the End Users element because the Sono SC device gives users the possibility to program alerts and functionalities. 4
In this paper we have presented a model to support the assessment of a smart devices according to di erent dimensions. The model, has four \macro" elements: Communication, Target, Data Manipulation and Development. Each of them is then further speci ed. By ranking (using a discrete scale) each element, it is possible to get an overall assessment of a given IoT device. Its suitability for a speci c domain is then evaluated comparing its assessed characteristics to the domain needs. We provided an example that assesses and compares two similar IoT devices that can be used at home.
In order to formally evaluate the model we are currently planning a test study with speech-therapist physicians and IoT devices.
87