BRAIN-IoT Architecture and Platform for Building IoT Systems
Salim Chehida
1 a
, Saddek Bensalem
1 b
, Davide Conzon
2 c
, Enrico Ferrera
2 d
and Xu Tao
3
1
CNRS, VERIMAG, University of Grenoble Alpes, Grenoble, France
2
IoT and Robotics Area, LINKS Foundation, Turin, Italy
3
Computer Science Department, University of Kentucky, U.S.A.
Keywords:
IoT, Architecture, Model-based Design, Interoperability, Distributed Execution, Security and Privacy,
Monitoring, Resiliency, Cloud, Edge, Decentralized IoT Applications.
Abstract:
The integration of Internet of Things (IoT) for building complex and critical systems requires powerful plat-
forms enabling to deal with multiple issues, including modeling, monitoring, control, maintaining and man-
agement of IoT applications. In this work, the authors propose a new platform based on layered architecture
that integrates a set of assets for model-based development of IoT systems. This platform named BRAIN-IoT
aims to meet the new challenges of IoT applications and to reduce the effort for building and managing these
applications. It consists of three frameworks that allow building decentralized IoT applications with comput-
ing capacity at the edge in a computing continuum with the cloud. The modeling and validation framework
is used to design, develop, and validate IoT applications logic. The distributed execution framework provides
an autonomic distributed infrastructure for the dynamic deployment and execution of IoT services on a mixed
cloud-edge environment. The security framework enables access control, end-to-end security and privacy of
data collected using IoT devices. The BRAIN-IoT platform is mapped to a well-established IoT reference
architecture and experimented on two industrial use cases.
1 INTRODUCTION
In recent years, IoT has demonstrated to be able to of-
fer significant improvements to various domains, e.g.,
health, energy, hydraulics, and transport. Several ar-
chitectures and platforms (Adolphs and Epple, 2015;
Lin et al., 2019; R
¨
omer et al., 2020; Bauer et al., 2013;
IEEE, 2020) have been proposed for building IoT sys-
tems. However, the complexity and the dynamic of
these systems highlights the need for new solutions
that allow promoting automation, managing complex-
ity and increasing trust. To cope with these demand-
ing requirements, a multitude of novel technologies
such as Edge Computing, Artificial Intelligence (AI),
and Analytics, as well as Security, Privacy and Trust
schemes are being investigated to be adopted in cur-
rent IoT architectures standards.
The 3D IoT architecture (Vermesan and Bacquet,
2018) specified by Alliance for the Internet of Things
Innovation (AIOTI) (AIOTI, 2020) is high-level ar-
a
https://orcid.org/0000-0002-5070-2591
b
https://orcid.org/0000-0002-5753-2126
c
https://orcid.org/0000-0002-2962-8702
d
https://orcid.org/0000-0002-4671-3861
chitecture proposed for supporting the novel require-
ments that generic IoT applications have. It aims to
be one pioneer of the next-generation IoT paradigm,
establishing a novel generic reference for different
IoT/Industrial Internet of Things (IIoT) applications
from different domains. In this work, the authors
present a new platform founded on a layered concrete
architecture based on the 3D reference model. The
BRAIN-IoT platform integrates novel technologies to
address the following challenges that arise in recent
IoT applications:
C1) Facilitate IoT systems modeling through multi-
ple abstractions and enabling the integration of
Validation and Verification (V&V) techniques
and tools.
C2) Enable designing IoT applications involving sev-
eral heterogeneous platforms and Smart Things
interconnected to each other in a distributed en-
vironment.
C3) Support the analysis and exploitation of large
amount of raw data to extract and predict infor-
mation supporting system automaticity and au-
tonomicity.
Chehida, S., Bensalem, S., Conzon, D., Ferrera, E. and Tao, X.
BRAIN-IoT Architecture and Platform for Building IoT Systems.
DOI: 10.5220/0011086000003194
In Proceedings of the 7th International Conference on Internet of Things, Big Data and Security (IoTBDS 2022), pages 67-77
ISBN: 978-989-758-564-7; ISSN: 2184-4976
Copyright
c
2022 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
67
C4) Ensure the deployment of services in a dis-
tributed environment, dynamic resources alloca-
tion in response to environmental changes, and
autonomous dependency management to reduce
the system operational complexity.
C5) Ensure the security, privacy, and resiliency in
resource-constraint and distributed IoT environ-
ments.
In this work, the authors validate the BRAIN-IoT
platform through two use cases for warehouse logis-
tics and for water distribution management. The solu-
tion is evaluated using a standard well-known usabil-
ity questionnaire, the User Experience Questionnaire
(UEQ)
1
, which allows the end-users to evaluate their
feeling with the product in question.
Section 2 presents the existing IoT architectures
and platforms. Section 3 and 4 show the BRAIN-IoT
Architecture and its integrated components. In Sec-
tion 5, the authors map BRAIN-IoT components with
the 3D IoT reference architecture. Section 6 presents
the applications leveraged to validate the architecture.
Section 7 briefly explains the validation results. Fi-
nally, the authors draw conclusions in Section 8.
2 RELATED WORK
Nowadays, it is still not available an unique IoT refer-
ence architecture. Several different organizations and
consortia have tried to define it, but the scenario is still
fragmented. This section presents the available IoT
reference architectures, describing their main charac-
teristics.
The Reference Architecture Model Industrie 4.0
(RAMI 4.0) (Adolphs and Epple, 2015) is a refer-
ence architecture focused on the smart industry do-
main. The effort has been started in Germany, but
today is driven by several companies and foundations
in relevant industry sectors, i.e., the ones associated
in the Platform Industrie 4.0
2
. RAMI 4.0 is a Service
Oriented Architecture (SOA), which combines all el-
ements and Information technology (IT) components
in a layered and life cycle model. RAMI 4.0 breaks
down complex processes into easy-to-grasp packages,
including data privacy and IT security. Since the ar-
chitecture is focused on the industry domain, it does
not cover the requirements of generic IoT applica-
tions.
The Industrial Internet Reference Architecture
(IIRA) (Lin et al., 2019) is a common architecture
1
https://www.ueq-online.org/
2
https://www.plattform-i40.de/IP/Navigation/DE/
Home/home.html
framework defined by the Industry Internet of Things
Consortium (IIC), to develop interoperable IIoT sys-
tems for applications across a broad spectrum of in-
dustrial verticals, both public and private, to achieve
the IIoT goals. Also in this case, the architecture is in-
teresting, but it is focused only on industrial domain,
so it is not suitable to be applied in a more generic IoT
scenarios.
The Eclipse Arroehead (R
¨
omer et al., 2020) is a
SOA with a reference implementation for IoT interop-
erability that was originally developed as part of the
Arrowhead Tools European research project
3
, aligned
with the concept of RAMI 4.0 and IIRA. Also in this
case, the architecture is mainly focused on industry
4.0 applications, but it can be also adapted to smart
cities, e-mobility, energy, and buildings domains.
The Internet of Things Architecture (IoT-A)
(Bauer et al., 2013) is a reference architecture for IoT
applications, which has been defined as an outcome
of the IoT-A European Union (EU) project
4
, which
had the main objective to develop an architectural ref-
erence model for the interoperability of IoT systems.
The last version of the architecture has been defined in
2013 at the end of the project, but its use as reference
model is currently limited.
The Standard for an Architectural Framework for
the Internet of Things (IoT) (IEEE, 2020) has defined
an architecture framework description for IoT, which
conforms to the international standard International
Organization for Standardization (ISO)/International
Electrotechnical Commission (IEC)/Institute of Elec-
trical and Electronics Engineers (IEEE) 42010:2011.
The standard provides an architectural blueprint
for Smart City implementation, considering cross-
domain interaction and enabling semantic interoper-
ability among various domains and components of a
Smart City (e.g., mobility, healthcare). For this rea-
son, also if not specialized for a single domain, the
standard is limited to Smart City related applications.
The results of the newest research projects, es-
pecially in the EU, brought to the harmonization of
several standard approaches, which brought to the
3D IoT Architecture model(Vermesan and Bacquet,
2018) as shown in Figure 1. The AIOTI High Level
Architecture (HLA) (AIOTI, 2020) has been defined
by the AIOTI Working Group (WG) Standardization
for IoT to be applied by Large Scale Pilots (LSPs)
projects
5
. Capturing the commonalities shared among
Reference Architectures of the LSPs, the 3D Refer-
ence Architecture has been produced from the EU
3
https://www.eclipse.org/org/research/project/
4
https://cordis.europa.eu/project/id/257521/it
5
https://european-iot-pilots.eu/
IoTBDS 2022 - 7th International Conference on Internet of Things, Big Data and Security
68
Figure 1: 3D IoT architecture (Vermesan and Bacquet, 2018).
project CREATE-IoT
6
. The 3D architecture is generic
and offers a representation that can include the differ-
ent IoT/IIoT applications across different sector do-
mains (e.g., automated/autonomous vehicles, smart
farming, smart cities, energy, manufacturing, health).
The architecture includes the function by design con-
cept with end-to-end functions addressed across the 8
layers. This allows addressing the heterogeneous ap-
plications including different IoT platforms and pro-
cessing at the edge, fog, and cloud. The LSPs projects
are reflecting such a model in their architectures,
which completely represents this new level of com-
plexity.
In this work, we present a new concrete architec-
ture platform based on the 3D model. The BRAIN-
IoT platform is developed in an EU project
7
. It pro-
vides a set of integrated assets to build IoT applica-
tions from different domains. Assets can be used in
the different stages of the IoT development process,
starting from modeling to deployment and mainte-
nance. They provide the key functions defined by the
3D reference model to address next-generation IoT
challenges. The next two sections will introduce the
BRAIN-IoT architecture, which then will be mapped
in the 3D reference model.
6
https://european-iot-pilots.eu/project/create-iot/
7
https://www.brain-iot.eu/
3 BRAIN-IoT ARCHITECTURE
Figure 2 shows the BRAIN-IoT architecture, which
includes several assets structured in six layers.
The application layer describes the applications
that are built. As mentioned earlier, the BRAIN-IoT
platform can be applied for generic IoT applications.
In this work, two scenarios will be presented : Service
Robotics for Warehouse Logistics (SRWL) and Criti-
cal Infrastructure for Water Distribution Management
(CIWDM). SRWL is a special logistic service involv-
ing several robotic platforms from Robotnik Automa-
tion company
8
, which need to collaborate to scan a
given warehouse and to assist humans in a logistic
domain. A fleet of robots supports the movement of
different loads in a warehouse. CIWDM is a scenario
related to the management of water distribution net-
work in collaboration with EMALCSA company
9
. It
focuses on monitoring and control the management of
the water urban cycle in metropolitan environment of
the city of la Coruna in Spain. More details on the
applications realized by the BRAIN-IoT platform are
described in Section 6.
The modeling and validation layer provides sev-
eral functions, e.g., design of complex IoT systems
considering several abstraction levels, simulation and
8
https://robotnik.eu/
9
https://www.emalcsa.es/index.php/es/
BRAIN-IoT Architecture and Platform for Building IoT Systems
69
Figure 2: BRAIN-IoT architecture.
validation of systems behavior before their real de-
ployment, automatic code generation from the mod-
els, ensuring online intelligent services for prediction
and anomaly detection based on data, monitoring and
controlling IoT applications, modeling IoT devices
and checking the correctness of systems before its de-
ployment in the physical world.
The distributed execution layer is responsible of
dynamic deployment and execution of IoT applica-
tions. It allows assembling, configuring and maintain-
ing runtime applications. It contains required IoT ser-
vices as well as functionalities for discovery, look-up,
and name resolution of IoT services. This layer is also
responsible for operational management and monitor-
ing of specific devices via the gateways deployed to
each BRAIN-IoT Edge Node.
The security layer provides multiple services for
protecting IoT systems such as security risk assess-
ment of IoT systems (Chehida et al., 2021), ensuring
end-to-end security from IoT devices/Cyber Physical
System (CPS) to application (Maillet-Contoz et al.,
2020), ensuring identity and distributed access con-
trol management for IoT devices/CPS and users, and
the implementation of privacy features for the IoT ap-
plications (Rashid et al., 2019).
The bottom layers represent connectivity proto-
cols and Application Programming Interface (API)
such as Long Range (LoRa) and Robot Operating
System (ROS), and physical IoT devices, e.g., actu-
ators, sensors.
4 BRAIN-IoT INTEGRATED
PLATFORM
Figure 3 shows the BRAIN-IoT integrated assets
available in the Eclipse Research Lab
10
, which re-
sponds to challenges C1, C2, C3, C4 and C5 pre-
sented in Section 1.
The BRAIN-IoT Services Development Toolkit
adopts a methodology and toolset for modeling dif-
ferent abstraction levels of IoT systems (C1). It is re-
sponsible for designing the logic of the IoT Applica-
tion, which composes the services provided by avail-
able devices (i.e., sensors, actuators, CPSs) and exter-
nal services (e.g., weather forecast, open data, third-
party IoT platforms, databases) based on their inter-
actions. The application logic is modelled along with
the relevant IoT environment using IoT-ML through
BRAIN-IoT Modeling Tool. The IoT-ML Modeling
Tool defines a Domain-Specific Modelling Language
(DSL), based on Unified Modeling Language (UML)
(UML2, 2017), to describe system-level components
choices and their dependencies, also IoT devices ca-
pabilities. Then, the IoT-ML model is refined to Be-
havior, Interaction, and Priority (BIP) model repre-
senting formally the components behavior and their
possible interactions using automata-based models
expressed in the BIP language (Basu et al., 2011).
The BIP language integrated in BRAIN-IoT platform
10
https://github.com/eclipse-researchlabs/brain-iot
IoTBDS 2022 - 7th International Conference on Internet of Things, Big Data and Security
70
Figure 3: BRAIN-IoT Integrated Assets.
gives a precise semantics to the system models ex-
pressed by IoT-ML. Also, the BIP Modelling and
Verification Tool provides efficient tool for verifica-
tion and analysis of the BIP models based on Statis-
tical Model Checking (SMC) technique. SMC uses
simulation-based approach to reason about formal re-
quirements expressed in temporal logic properties.
After simulation and verification, the BIP model is
converted in Java source code as application arte-
facts through the BIP Code Generator. The gener-
ated artifacts are then released and stored in BRAIN-
IoT Services Repository, to be deployed and executed
by the distributed execution framework. The BRAIN-
IoT modeling and validation framework also offers
the ability to use development-time models to super-
vise running execution platform states using sensi-
Nact Studio. This solution enables monitoring of
BRAIN-IoT services and starting or stopping them
from the BRAIN-IoT Services Development Toolkit.
In addition, the generated IoT applications, named
as BRAIN-IoT services, can be validated leverag-
ing IoT physical device models with Physical Layer
Modelling Tool based on SystemC-TLM language
(Ghenassia, 2006) to validate the correctness of the
system behavior, before deploying it in the physical
world. Thanks to the Physical Layer Modelling Tool,
the system reliability is increased, as the validation
strategy of the end-device is strengthened by adding
scenarios focusing on device robustness considering
its interaction with system environment. Neverthe-
less, at the system level modelling, it is able to de-
scribe the services provided by the existing IoT de-
vices with Web of Things (WoT) Thing Description
11
,
so that enforces the services composition in the sys-
11
https://www.w3.org/TR/wot-thing-description/
tem models and provides the inputs to the edge node
at runtime to communicate with the external IoT de-
vice using the communication protocol it supports.
Finally, an AI approach supported by the sOnar tool is
provided to model and implement AI modules, which
can be composed with the system behavior models
(C3). The sOnar tool provides the online data analy-
sis service for anomaly detection and prediction func-
tionality using machine learning techniques. It acts
as an intelligent computing unit to enable autonomic
functionalities for the Runtime Infrastructure through
the deployment and execution of reusable software
AI/ML algorithms. It sends notifications to the execu-
tion framework to trigger a prompt reaction that miti-
gates or avoids negative consequences to the IoT/CPS
systems.
The BRAIN-IoT distributed execution frame-
work provides an autonomic distributed infrastructure
(BRAIN-IoT Fabric) for the dynamic deployment, dis-
covery and execution of BRAIN-IoT services on a
mixed cloud-edge environment (C4). The BRAIN-
IoT Fabric is combined with the Paremus Service
Fabric to manage communications between the devel-
oped applications from BRAIN-IoT Services Reposi-
tory through asynchronous events using the BRAIN-
IoT EventBus. By using BRAIN-IoT EventBus, the
multiple deployed applications are completely decou-
pled and the failure of one application has no impact
on other running applications. The distributed exe-
cution framework also provides the interoperability
for the external IoT devices/platforms, using WoT-
enabled Edge Nodes based on a World Wide Web
Consortium (W3C) WoT Thing Description of the
communications interface and sensiNact Edge Nodes
(G
¨
urgen et al., 2016) based on Eclipse sensiNact gate-
BRAIN-IoT Architecture and Platform for Building IoT Systems
71
way
12
, which can be deployed in a distributed and
bulk manner (C2). The sensiNact-enabled Edge Node
provides connectivity, interoperability, and data pro-
cessing to various IoT devices by its capability to in-
teract with a wide variety of equipment and protocols,
as well as its extensibility mechanisms. The WoT-
enabled Edge Node implements an approach to en-
able the interoperability between ROS-based CPS ap-
plications and other heterogeneous IoT platforms in a
sophisticated IoT software ecosystem.
The BRAIN-IoT security framework provides a
Security Module and Gateway, a Message Integrity
Service (MIS), and a distributed Authentication, Au-
thorization, and Accounting (AAA) Server to ensure
data confidentiality, integrity, availability, and authen-
tication for the BRAIN-IoT platform (C5). The se-
curity module authenticates and encrypts data sent
over the network at the application level with reduced
energy consumption for IoT sensors and actuators.
The security gateway checks the sender’s authentica-
tion before decrypting the data. The distributed AAA
server manages identity and rights from users and IoT
devices. The MIS signs the data event before sending
it and verifies the message integrity and sender au-
thentication when the event is received by a node. The
Attack-Defense Strategies Exploration Tool (Chehida
et al., 2020) is a decision-supporting tool which gives
the suggestions for the security considerations by pro-
viding insightful information that allows the security
manager to evaluate system vulnerabilities and to de-
sign reliable security policies. The tool identifies the
potential attack actions that are most likely to succeed
such as network attacks and data manipulation known
as False Data Injection Attacks (FDIAs). It also se-
lects high impactful defense actions that make the sys-
tem harder to attack while finding a balance between
the attack cost and its probability of success. Further-
more, the Privacy Control System provides the policy-
based mechanism for the IoT users to protect the per-
sonal data collected using IoT devices. It is based on a
Policy Enforcement Point (PEP) that applies the poli-
cies and controls access to the data by available ser-
vices. It attaches the policies to the data event and
delivers it along with the data over the EventBus. The
Privacy Control System aims allowing a Privacy-as-a-
Service approach and facilitating the adoption of pri-
vacy policies control in IoT environments based on
SOA.
12
https://projects.eclipse.org/projects/technology.
sensinact
5 MAPPING BRAIN-IoT ASSETS
WITHIN THE 3D MODEL
The BRAIN-IoT architecture previously presented is
based on the 3D architecture introduced by AIOTI. To
better explain the link between the two architectures,
this section presents the mapping of the BRAIN-IoT
main components in the 3D architecture as shown in
Figure 4.
Specifically, the Physical layer includes actuators,
sensors and CPSs providing the connectivity cross-
cutting function and useful to support integrability.
On the Network Connectivity Layer there are the End-
to-end Security Module and Gateway providing secu-
rity and integrability as well as a set of different APIs
providing connectivity and integrability. On the Pro-
cessing Layer there are the Privacy Control System
for privacy control and dependability, as well as the
sensiNact Edge Nodes and the WoT Enabled Edge
nodes that provide connectibity and interoperability;
at the same layer the AAA server and the MIS provid-
ing reliability and dependability. The Storage Layer
includes the BRAIN-IoT Services Repository provid-
ing resilience and availability. On the Service Layer
there are BRAIN-IoT Fabric that provides connec-
tivity and composability; the BRAIN-IoT Moelling
Tool to provide manageability; the IoT-ML that pro-
vides interoperability; the BIP Code Generator, which
provides dependability; sOnar providing intelligence;
Brain-IoT Physical Layer Modeling Language to pro-
vide connectivity and integrability. Finally on the
Application Layer there are the BRAIN-IoT Attack-
Defense Strategies Exploration Tool providing secu-
rity and availability as well as the sensiNact Studio
that provides connectivity and manageability.
This mapping shows that the BRAIN-IoT compo-
nents are able to provide the main IoT cross-cutting
functions and IoT system properties needed to satisfy
the requirements of next-generation IoT challenges.
6 APPLICATIONS
As said in Section 3, the BRAIN-IoT platform has
been validated through two main scenarios: Service
Robotics for Warehouse Logistics (SRWL) and Criti-
cal Infrastructure for Water Distribution Management
(CIWDM).
In the first scenario, the BRAIN-IoT platform sup-
ported the implementation of a multi-agent system for
the distributed self-organized management of a fleet
of robots to move loads among the different areas of
a warehouse. The movement of these loads does not
require any operator to control the fleet. Robots are
IoTBDS 2022 - 7th International Conference on Internet of Things, Big Data and Security
72
Figure 4: Mapping BRAIN-IoT assets with 3D IoT architecture.
expected to empty continuously the “inbound area”
of the warehouse, where the loads are temporarily po-
sitioned when items are delivered in for storage, mov-
ing them to a specific place in the ”storage area”.
When triggered by the human operator, each robot
autonomously moves from their base station to the in-
bound area and patrols looking for the loads and iden-
tifying them through their barcode. Consequently, the
robot establishes a communication with the backend
system of the warehouse to ask which cell of the ”stor-
age area” the load must be placed in. When the back-
end system return back the coordinates of the storage
cell, the robot picks the load up and moves it through
the warehouse to the right position in the storage area.
The navigation of the robots across the warehouse is
designed to be safe, avoiding obstacles along the path,
and adaptable to the specific environment where the
fleet of robots is deployed. For instance, the same
warehouse logistics application has been deployed in
a scenario similar to a supermarket, where the items
to store can also be perishable, e.g. like fruits and
vegetables or fish and meat. In such scenario, the
storage area must be kept at a specific environment
temperature to not ruin the food, and it is separated
from the the other warehouse areas by a controllable
door. In such scenario, the robots are able to dis-
cover the door as a new controllable service available
on the network. Consequently, through the BRAIN-
IoT platform it is able to trigger the deployment at
run time of a new driver which enables the interac-
tion with the automated door. Afterwards, the robots
capabilities are then augmented allowing them to nav-
igate across the warehouse also opening and closing
the door when needed. In this scenario, the main ob-
jectives of the application of the BRAIN-IoT platform
were to reduce the development time for multi-agent
robot applications, establishing a secure end-to-end
interaction with robotic platforms (more specifically
ROS-based
13
robots) as well as legacy backend sys-
tems and external devices within the surrounding en-
vironment. Moreover, the use-case allowed also to
demonstrate and evaluate the dynamic autonomous
reconfiguration of the system for self-adaptation to
the environment layout and anomaly situations.
In the second scenario, the BRAIN-IoT platform
is adopted to implement a safe autonomous control
for the management of a critical infrastructure like
the one distributing the water in the city of A Coru
˜
na,
in Spain. The objective is to collect and exploit data
coming from geographically distributed and hetero-
13
https://www.ros.org/
BRAIN-IoT Architecture and Platform for Building IoT Systems
73
geneous devices and platforms, used for water level
monitoring and treatment, with the aim of early de-
tecting abnormal and critical situations that may hin-
der the correct functioning of the infrastructure or
even damage it. The good quality of the data is
fundamental for creating accurate indicators for deci-
sion making and for real-time adaptive control pro-
cedures. For this reason, the protection of the nu-
merous resource-constrained sensors and meters is
extremely important to guarantee the dependability
of the measured data. Also, it is very important to
guarantee that all the different data sources are au-
thorized to provide data, and no malicious injection
of fake information are present. More specifically,
the needed control consists in the following: when
the level of water reaches a specific threshold, the
spillgates shall be opened. Normally, the spillgates
opening is manually performed. In this use case,
BRAIN-IoT platform controls automatically the spill-
gates opening, based on the measures of the water
flows. Such control strategies are modelled and devel-
oped with the BRAIN-IoT Modelling Tool. A model-
based approach allows to modify in an agile way the
control strategies whenever, in the future, modifica-
tions or improvements will be needed. Furthermore,
the BRAIN-IoT platform is used for collecting data
from different water flow meters placed in different
locations and analyzing them to detect possible de-
viations w.r.t. the normal behaviour. In fact, possible
data outliers (e.g. due to device obsolescence, miscal-
ibration, external attacks) may be extremely risky for
the water infrastructure because data are used to con-
trol the spillgates for let the water flowing correctly
in the pipes and deliver the water correctly to the cus-
tomers. According to the outcomes of the analysis,
the BRAIN-IoT platform must autonomously recon-
figure the spillgates control strategy in such a way to
cut off the ones which are directly affected by the me-
ter’s misbehaviors. This approach allows to: enhance
the resiliency of the water infrastructure from failures
in order to provide a service which has reduced possi-
bilities of discontinuity; identify in real-time the pos-
sible problems with the meters and sensors, allowing
a prompt reaction for mitigating the issues and pro-
vide a better service experience to the customers; col-
lect as a whole the data coming from heterogeneous
meters and sensors which are spread all over the city.
7 VALIDATION RESULTS
The BRAIN-IoT platform has been evaluated through
workshops conducted carrying out a usability assess-
ment performing several teleconferences with rele-
vant stakeholders and were oriented to the compila-
tion of a standard well-known usability questionnaire,
which allows end-user of a product to evaluate their
feeling with the product in question. The results of
the inquiry demonstrated to be very valuable to mea-
sure the interest from the stakeholders for the pro-
vided functionalities and their degree of satisfaction
with the overall BRAIN-IoT Platform.
Using the UEQ, the items are scaled from -3 to
+3, where -3 represents the most negative answer, 0
a neutral answer, and +3 the most positive answer.
The items are categorised into six dimensions each
of which consists of 4-6 items on the UEQ which
thus describes a distinct quality aspect of an interac-
tive product. Attractiveness: general impression to-
wards the product. Efficiency: is it possible to use
the product fast and efficient? Perspicuity: is it easy
to understand how to use the product? Dependabil-
ity: does the user feel in control of the interaction?
Stimulation: is it interesting and exciting to use the
product? Novelty: Is the design of the product in-
novative and creative? The results are also evaluated
using a benchmark. For this analysis, the measured
scale means are set in relation to existing values from
a benchmark data set, which contains data from 20190
persons from 452 studies concerning different prod-
ucts (business software, web pages, web shops, social
networks). The benchmark classifies a product into 5
categories (per scale). Excellent: in the range of the
10% best results. Good: 10% of the results in the
benchmark data set are better and 75% of the results
are worse. Above average: 25% of the results in the
benchmark are better than the result for the evaluated
product, 50% of the results are worse. Below aver-
age: 50% of the results in the benchmark are better
than the result for the evaluated product, 25% of the
results are worse. Bad: in the range of the 25% worst
results.
The results of the evaluation of the BRAIN-IoT
platform show that the stakeholders perceived it as an
innovative solution. The discussion during the work-
shops let the self-administrative capabilities emerged
as the main innovative functionality, along with the
capacity of distribute edge intelligence in a decentral-
ized environment. The good score for the depend-
ability of the platform is due to the good comments
about the security perspective, more specifically for
the capability of protecting data sourced by resource
constrained devices, as well as the capacity to pro-
vide the data owner the control of the access poli-
cies of their data. Despite one of the objectives of
BRAIN-IoT is to relief the IoT developers and sys-
tem operators from the burden of implementing and
operating IoT applications involving multiple hetero-
IoTBDS 2022 - 7th International Conference on Internet of Things, Big Data and Security
74
Figure 5: Validation results of the BRAIN-IoT platform.
Figure 6: Validation results of the BRAIN-IoT platform applied to the SRWL scenario.
Figure 7: Validation results of the BRAIN-IoT platform applied to the CIWDM scenario.
geneous existing platforms, the perspicuity score is
not very high. The comments have been that the plat-
form is considered very helpful once it has been con-
figured and executed in the production environment,
but the configuration phase seems to be harder than
expected, because of the difficult of the modelling
phase and the setup of the BRAIN-IoT Fabric. The
low score for the perspicuity also driven the low at-
tractiveness, which is affected by the perceived dif-
ficulty for configuration and setup of the whole plat-
form. It is however important to notice the most part
of the BRAIN-IoT Platform components have been
developed from scratch or starting from low Technol-
ogy Readiness Level (TRL) background assets. The
BRAIN-IoT Architecture and Platform for Building IoT Systems
75
consequence is that some more effort will need to be
invested in order to improve the usability in terms of
system setup. The good achievement is that the main
novel characteristics of the platform have been well
received and considered relevant and useful by the ex-
perts.
Figure 5 shows the positioning of the BRAIN-IoT
Platform relative to the benchmark. The results show
the good feeling w.r.t. the Novelty and Dependabil-
ity, for which the score is Excellent. Considering the
benchmark refers to solutions in every Information
and Communication Technologies (ICT) field, meant
to be commercial products, the curve of the BRAIN-
IoT platform is in line with the expectation of a re-
search and innovation action, where the maturity of
the developed solution is still about around TRL 6,
but the innovative aspects are well evident.
A similar evaluation has been performed to evalu-
ate the software within the scope of the two applica-
tion scenarios. When applied to the Service Robotics
scenario, despite the perspicuity remains in line with
the values determined before, the attractiveness is
much higher. This is because the perception may
change a lot when you think about the technology ap-
plied to a specific context, and the benefits that the
platform could bring are much more evident. Basi-
cally, the stakeholder but himself in the shoes of the
person who will benefit from the execution of the plat-
form, instead of the IT manager who is supposed to
install and setup the system. This means that, despite
the BRAIN-IoT Platform continues to be evaluated as
complicate to setup, it is considered good for manag-
ing swarm robotics applications.
Figure 6 shows the positioning of the BRAIN-IoT
Platform relative to the benchmark. It shows even
clearly how the BRAIN-IoT Platform is considered
above the average in a context of SRWL. Instead,
the dependability is decreased because of the secu-
rity perspective in this context is more in line with
the state of the art technologies. As for the SRWL
scenario, a similar analysis could be made for the CI-
WDM. However, differently from the other scenario,
here the perspicuity is higher: while for the SRWL
scenario, the robotic applications developer cannot ig-
nore the system setup phase, in the case of the man-
agement of the CIWDM, the platform is evaluated
from the perspective of the pure end-user. Figure
7 shows the positioning of the BRAIN-IoT Platform
relative to the benchmark. In this case, the platform
has been evaluated as beneficial for managing a criti-
cal infrastructure, especially considering the security
features and the resiliency capabilities.
8 CONCLUSION
BRAIN-IoT platform aims to pave the way of the
research around the strict requirements of the next-
generation IoT paradigm(AIOTI, 2020). In other
words BRAIN-IoT platform is a first implementa-
tion of a meta operating system for the IoT domain,
which facilitates the implementation of secure and
self-adaptive IoT applications. The platform has been
tested and applied to robotics scenarios in which
BRAIN-IoT enables the self-controlled robots inter-
act with the environment and adopt their behaviors
correspondingly. In addition, BRAIN-IoT platform
demonstrated its advanced features in industry and
agile critical infrastructures management especially
for the physical infrastructure monitoring, and abnor-
mal behavior detection and prediction. Such meta op-
erating system maps toward the 3D architecture de-
fined by AIOTI but some functionalities still need to
be either covered or enhanced, such as safety, trust-
worthiness. As part of the future works, the authors
plan to extend the BRAIN-IoT platform implement-
ing these functionalities with the aim to implement a
full featured meta operating system for the IoT.
ACKNOWLEDGMENT
The research leading to these results has been sup-
ported by the European Union through the BRAIN-
IoT project H2020-EU.2.1.1. Grant agreement ID:
780089.
REFERENCES
Adolphs, P. and Epple, U. (2015). Reference architecture
model industrie 4.0 (rami4.0). Standard, ZVEI – Ger-
man Electrical and Electronic Manufacturers’ Associ-
ation.
AIOTI (2020). High Level Architecture (HLA).
Standard, AIOTI WG Standardisation.
https://aioti.eu/wp-content/uploads/2020/12/
AIOTI HLA R5 201221 Published.pdf.
Basu, A., Bensalem, S., Bozga, M., Combaz, J., Jaber,
M., Nguyen, T.-H., and Sifakis, J. (2011). Rigor-
ous Component-Based System Design Using the BIP
Framework. IEEE Software, 28(3):41–48.
Bauer, M., Boussard, M., Bui, N., Carrez, F., Jardak, C.,
Loof, D., Magerkurth, C., Meissner, S., Nettstr
¨
ater,
A., Olivereau, A., Thoma, M., Walewski, Stefa, J.,
and Salinas, A. (2013). Internet of things architec-
ture iot-a deliverable d1.5 final architectural refer-
ence model for the iot v3.0. Technical report.
Chehida, S., Baouya, A., Alonso, D. F., Brun, P.-E., Mas-
sot, G., Bozga, M., and Bensalem, S. (2021). Asset-
IoTBDS 2022 - 7th International Conference on Internet of Things, Big Data and Security
76
driven approach for security risk assessment in iot sys-
tems. In Garcia-Alfaro, J., Leneutre, J., Cuppens, N.,
and Yaich, R., editors, Risks and Security of Internet
and Systems, pages 149–163, Cham. Springer Interna-
tional Publishing.
Chehida, S., Baouya, A., Bozga, M., and Bensalem, S.
(2020). Exploration of impactful countermeasures on
iot attacks. In 2020 9th Mediterranean Conference on
Embedded Computing (MECO), pages 1–4.
Ghenassia, F. (2006). Transaction-Level Modeling with Sys-
temc: Tlm Concepts and Applications for Embedded
Systems. Springer-Verlag, Berlin, Heidelberg.
G
¨
urgen, L., Munilla, C., Druilhe, R., Gandrille, E., and
Nascimento, J. (2016). sensiNact IoT Platform as a
Service, pages 127–147.
IEEE (2020). Ieee standard for an architectural framework
for the internet of things (iot). IEEE Std 2413-2019,
pages 1–269.
Lin, S.-W., Miller, B., Durand, J., Bleakley, G., Chigani,
A., Martin, R., Murphy, B., and Crawford, M. (2019).
The industrial internet of things volume g1: Refer-
ence architecture. Standard, Industry IoT Consortium.
https://www.iiconsortium.org/pdf/IIRA-v1.9.pdf.
Maillet-Contoz, L., Michel, E., Nava, M. D., Brun, P.-E.,
Lepr
ˆ
etre, K., and Massot, G. (2020). End-to-end secu-
rity validation of iot systems based on digital twins of
end-devices. In 2020 Global Internet of Things Sum-
mit (GIoTS), pages 1–6.
Rashid, M. R. A., Conzon, D., Tao, X., and Ferrera,
E. (2019). Privacy Awareness for IoT Platforms:
BRAIN-IoT Approach. In Ramos, J. L. H. and
Skarmeta, A. F., editors, Security and Privacy in the
Internet of Things: Challenges and Solutions, vol-
ume 27 of Ambient Intelligence and Smart Environ-
ments, pages 24–43. IOS Press.
R
¨
omer, L., Jeroschewski, S. E., and Kristan, J. (2020).
Leveraging eclipse iot in the arrowhead framework.
In NOMS 2020 - 2020 IEEE/IFIP Network Operations
and Management Symposium, pages 1–6.
UML2 (2017). Unified Modeling Language (Version 2.5.1).
Object Management Group.
Vermesan, O. and Bacquet, J. (2018). The Next Genera-
tion Internet of Things Hyperconnectivity and Em-
bedded Intelligence at the Edge, pages 19–102. River
Publisher.
BRAIN-IoT Architecture and Platform for Building IoT Systems
77