This is a pre-print of an article published in Cognitive Computation. The final
authenticated version is available online at: https://doi.org/10.1007/s12559-018-9543-3
Cite this article as:
M. Mahmud, M.S. Kaiser, M.M. Rahman, M.A. Rahman, A. Shabut, S. Al-Mamun, A.
Hussain. (2018). A Brain-Inspired Trust Management Model to Assure Security in a
Cloud based IoT Framework for Neuroscience Applications. Cogn. Comput., doi:
10.1007/s12559-018-9543-3.
c
2018, Springer Nature holds the copyright of this article.
A Brain-Inspired Trust Management Model to Assure
Security in a Cloud based IoT Framework for Neuroscience
Applications
Mufti Mahmud
1,*
, M. Shamim Kaiser
2,*
, M. Mostafizur Rahman
3
, M. Arifur Rahman
4
,
Antesar Shabut
5
, Shamim Al-Mamun
6
, Amir Hussain
7
1
NeuroChip Lab, University of Padova, 35131 - Padova, Italy
2
IIT, Jahangirnagar University, Savar, 1342 - Dhaka, Bangladesh
3
American International University - Bangladesh, 1213 - Dhaka, Bangladesh
4
Department of Computer Science, University of Sheffield, Sheffield, S10 2TN, UK
5
Anglia Ruskin University, CM1 1SQ - Chelmsford, UK
6
Saitama University, Saitama, 338-8570, Japan
7
Division of Computing Science & Maths, University of Stirling, FK9 4LA Stirling, UK
*
Co-‘first and corresponding’ author. Emails: muftimahmud@gmail.com (M. Mahmud),
mskaiser@juniv.edu (M.S. Kaiser)
Abstract
Rapid popularity of Internet of Things (IoT) and cloud computing permits
neuroscientists to collect multilevel and multichannel brain data to better understand
brain functions, diagnose diseases, and devise treatments. To ensure secure and reliable
data communication between end-to-end (E2E) devices supported by current IoT and
cloud infrastructure, trust management is needed at the IoT and user ends. This paper
introduces a Neuro-Fuzzy based Brain-inspired trust management model (TMM) to
secure IoT devices and relay nodes, and to ensure data reliability. The proposed TMM
utilizes node behavioral trust and data trust estimated using Adaptive Neuro-Fuzzy
Inference System and weighted-additive methods respectively to assess the nodes
trustworthiness. In contrast to the existing fuzzy based TMMs, the NS2 simulation
results confirm the robustness and accuracy of the proposed TMM in identifying
malicious nodes in the communication network. With the growing usage of cloud based
IoT frameworks in Neuroscience research, integrating the proposed TMM into the
existing infrastructure will assure secure and reliable data communication among the
E2E devices.
1/17
Data
Logging
Procedure
Modeling
D
a
t
a
C
o
l
l
e
c
t
i
o
n
[
D
i
s
t
r
i
b
u
t
e
d
]
D
a
t
a
A
n
a
l
y
t
i
c
s
Clinical / Experimental Neuroscience
Computational Neuroscience
Data Science
E
x
p
e
r
i
m
e
n
t
a
l
D
e
s
i
g
n
Data
Figure 1. Cycle of modern Neuroscience research.
Introduction
In recent years biological data has grown significantly, thanks to the technological
developments, now scientists can acquire data simultaneously from multiple levels and
channels of a living system [1], and simulate large scale brain networks [2, 3]. One of the
major contributors to this biological big data is Neuroscience [4]. Brain signals, e.g.,
Electroencephalogram (EEG), Electrocorticogram (ECoG), Neuronal Spikes (AP),
Local Field Potentials (LFPs) along with brain imaging techniques, e.g.,
Magnetoencephalography (MEG), Magnetic Resonance Imaging (MRI), Functional MRI
(fMRI), Positron Emission Tomography (PET) have been extensively used in diagnosis
of neurodegenerative diseases [5, 6], neuropsychiatric disorders [7], and developmental
disorders such as Autism Spectrum Disorder [8]. Additionally, this data has been
effectively utilized in developing various data-driven disease models [9, 10].
Modern day Neuroscience research is driven by data (see Fig. 1). Both clinical and
experimental neuroscience research generate huge amount of data [11] and analyzing
those data to draw meaningful conclusions is very challenging [12]. The extracted
knowledge from these data allow the development and refining of data-intensive models
and describe the underlying biological phenomena which in turn facilitate experimental
design [13]. The data analytics and modeling phases are computationally intensive, and
advancements in artificial intelligence [14] and cloud computing [15] allowed scientists to
perform these steps smoothly. The ‘cloudification’ greatly facilitated scientists by
providing ‘software as a service’ (e.g., service oriented architecture or SOA) instead of
running the data-intensive analyses and modeling locally in the computers. In other
words, cloud computing and big data paradigms converted context-aware research into
exhaustive, data-driven research.
Now, with the emergence of the Internet of Things (IoT), various sensors can be
connected to the cloud for seamless resource sharing. Such IoT-Cyber Physical Systems
(IoT-CPS) provide a platform to data-driven research and design appropriate medical
services for patients. The IoT-CPS tailored to patient monitoring and care are around
for a few years now and it allowed hospitals and healthcare processionals to seamlessly
exchange patients’ data even from remote locations. These data may represent a wide
range of healthcare parameters collected through the IoT for healthcare (IoHT) sensors.
One of the main challenges of this type of IoT-CPS is to ensure privacy and information
security. Thus, the trust management plays a vital role for the end users which act as a
2/17
first step of information security. Despite the fact that trust management is required for
all such frameworks dealing with biological data acquirable through the IoHT devices,
the Neuroscience data stands apart from the others and requires special attention due
to their high variability and spontaneity. While in many biosignals (e.g.,
Electrocardiogram, Electromyogram) periodicities and similarities have been noticed in
terms of frequency content, amplitude and shape, the Neuroscience data (e.g., EEG,
ECoG, LFPs, AP, etc.) have been known for their variabilities [16–18] making them
more prone to misidentification, misclassification and misinterpretation in cases when
the signals are unsupervisedly acquired without any experts. Therefore, to design
robust telemedicine systems using IoT-CPS targeting Neuroscience applications, extra
care must be taken to ensure the trustworthiness of the IoHT nodes.
Mahmud et al. introduced a service-oriented architecture for web based collaborative
biomedical signal analysis [19]. As an initial platform with three main components (i.e.,
users, contributors, and services), this model assumed the inherent security of the
internet and used certificate based security as authentication scheme for the
contributors and users to deploy and utilize services. The same architecture can be
extended by delegating the data coming from the IoT devices to the cloud for analysis.
Additionally, a cloud-based healthcare system was proposed in [20] to provide
convenient patient-centric healthcare services. In this model, the cloud performed the
big data analytics and the authors reported significant performance improvement in the
cloud-based system which too can be adapted to suit smart healthcare applications.
Also, biologically inspired cloud resource provisioning was proposed for optimal
handling of big healthcare data [21].
While the assumption of a secure cloud is appropriate in the context of currently
discussed communication models, discarding malicious transmission – identified by the
nodes profile information, behavior, and data similarity – is vital to ensure the
optimized performance, reliability, and robustness of a system. In the current scenario,
profile information is validated by the authentication services, and the nodes behavior
and data similarity are handled by a trust management system. To make a more
trustworthy system, Shabut et al. identified the malicious nodes based on their behavior
and improved packets delivery through a multi-hop relay network excluding those
misbehaving nodes [22]. Another work proposed a dynamic cluster based
recommendation model to minimize the data sparsity or cold start situations using
nodes behavior to improve quality of service (QoS) of end-to-end (E2E)
transmission [23]. Chen et al. proposed a Fuzzy reputation-based trust model (TRM)
for IoT-CPS which estimated the nodes trust from their behavior and showed an
improved performance in comparison to a communication system without trust [24]. An
ant colony based trust model was presented to determine the trust value of wireless
nodes which exhibited improved accuracy [25]. Context-aware multiservice trust
management systems were proposed in [26, 27] which filtered malicious nodes in the E2E
and heterogeneous IoT architectures with high accuracy. Another trust management
model (TMM) was proposed to evaluate the trustworthiness of nodes in the wireless
sensor network through beta distribution. The aggregated trust value from data and
energy was used in identifying the untrustworthy relay nodes to reduce the internal
threats [28]. Yet another trust management system, based on an agent’s trustworthiness
and confidence, was proposed to evaluate the trustworthiness of the IoT nodes [29].
Moreover, a joint social and QoS TMM was presented to find the trust level of wireless
nodes in a mobile adhoc network [30].
However, identifying the malicious transmission using only nodes behavior isn’t
enough to ensure reliable communication. It is important to guarantee that the data
generated by the nodes are error-free – which is a big challenge – and a TMM that takes
into account both nodes behavior and data similarity can be a solution to confirm nodes
3/17
reliability.
This paper presents an Adaptive Neuro-Fuzzy based Brain-inspired TMM targeting
cloud based IoT architecture to determine data trust and behavioral trust for all IoT
devices and relay nodes to ensure reliable data communication between E2E devices.
This work also investigates the effects of trust management on the QoS issues of the
cloud based IoT architecture suitable for neuroscience applications.
IoT End
Perception
Connectivity
Access and Network
Processing and Analysis
Cloud (or Server) and
Application Platform
User End
Visualization
Secure Cloud based IoT Framework for Neuroscience Applications
Authentication
Approval
Cloud
Gateway
Storage
Real-time Stream Processing
Trust Engine
(Global)
Trust Engine
(Local)
Software Defined
Network
Architecture
Devices
Security Protocols
(HTTPS, CoAP, REST, MQTT, XMPP)
User Registry and
Device State
Application API
(with App. Mgmnt)
Researcher
Caregiver
Doctor
EEG
ECoG
MRI
PET
Models
IoT Router
or Gateway
Figure 2.
Cloud based IoT Architecture for Neuroscience Applications. All the IoT
sensor nodes are deployed in the perception layer (IoT site).
1 Cloud based IoT Architecture
The big data and cloud are two paramount elements for creating collaborative
frameworks to analyze brain signals (e.g., EEG, ECoG, AP, LFPs, etc.) and brain
images (e.g., MEG, MRI, fMRI, PET, etc.) and to perform data-driven modeling [19].
Due to the wide range of advantages offered by such architectures, they have become
the trend in recent years [31].
Focusing on applications related to Neuroscience, Fig. 2 illustrates a cloud based IoT
framework which consists of three main components, i.e., the IoT end (contains the data
generating devices), the cloud component (provides the access and connectivity, and
processing and analysis of data), and the user end (provides the analyzed and processed
data to the users, e.g., doctors, caregivers, and researchers). In this framework, the data
from various Neurotechnology empowered devices are collected for the development of
state-of-the-art techniques pertaining to intelligent healthcare and advancement of
Neuroscience research. At the IoT end, also known as perception layer, various data
generating devices are connected to respective transceiver devices to forward the data to
the cloud through the IoT gateway either for data analytics or simply for storage.
Additionally, the brain signals generated at the IoT end are also used in operating
various medical and assistive devices (e.g., automatic wheelchair, robotic arm,
4/17
IoT Devices
Users
Authentication
Server
Trust
Relationship
1. Contributor Info
2. Security Approval
3. Security Approval
4. Certificate
5. Certificate
6. Service Access
1. User Info
3. Security Approval
2. Security Approval
4. Certificate
5.Certificate
6. Service Request
7. Service Output
Cloud
Web
Server
Application
Server
Security
Server
Figure 3. Cloud authentication model (adapted from [19]).
etc.) [31, 32] to provide the better monitoring and improve the quality of life. The cloud
is used for defining the access and the network and perform data storage and analytics.
Extending the work of Mahmud et al. [19], in our framework, we consider the cloud to
be secure through existing certification and authentication models (see Fig. 3). Finally,
at the user end, the service consumers can access and visualize the processed data based
on granted rights and privileges.
Trust
Propagation
Trust
Level
Trust
Properties
Trust
Aggregation
Trust
Update
Figure 4. Block diagram showing various steps of a trust evaluation process.
In the cloud based IoT architectures, the IoT devices or nodes generate data owing
to various Neuroscience applications. Like human relationships, these nodes collaborate
with each other through certain predefined social properties, and these properties are
the ‘Trust Compositions’ (see section 2). The values of these social properties are
propagated on the IoT and user ends (known as ‘Trust Propagation’). During direct or
indirect interactions, the trust metrics of each node are aggregated through static
weighted sum, neuro-fuzzy method, and Bayesian inference (known as ‘Trust
Aggregation’). The trust value of each node is then updated when an interaction is
completed (known as ‘Trust Update’). This update can also be done periodically for
energy efficiency. The block diagram of the trust management steps is illustrated in Fig.
4.
5/17
