ArticlePDF Available

Attack Resistance based Topology Robustness of Scale-free Internet of Things for Smart Cities

Authors:

Abstract and Figures

In Internet of Things (IoTs), the increase in the number of devices is directly proportional to the number of applications. The exponential growth of devices increases both the network complexity and risk against topology robustness. Moreover, the network is also prone to targeted and malicious attacks. In this paper, Enhanced Angle Sum Operation ROSE (EASO-ROSE), Enhanced ROSE, Adaptive Genetic Algorithm (AGA) and Cluster Adaptive Genetic Algorithm (CAGA) are proposed to cater the topology robustness issue for IoT enabled smart cities. In addition, the proposed solutions keep the nodes' initial degree distribution unchanged by maintaining the scale-free nature of the topology. Enhanced ROSE and EASO-ROSE significantly improve the topology robustness by calculating the difference in nodes' degree while rearranging the surrounding angles according to the highest degree node. CAGA and AGA also significantly improve the topology robustness by using adaptive probabilities of crossover and mutation that guide both algorithms to converge towards global optimum solution. Extensive simulations are preformed to evaluate the performance of the proposed strategies. Schneider R is used as a performance metric in the simulations. The results depict that the proposed algorithms perform 61.3%, 48.3%, 45.5% and 34.95%, better than simulating annealing algorithm.
Content may be subject to copyright.
Int. J. of Web and Grid Services, Vol. x, No. x, 202X 1
Attack Resistance based Topology Robustness of
Scale-free Internet of Things for Smart Cities
Talha Naeem Qureshi1, Nadeem Javaid,1,
Ahmad Almogren,2, Zain Abubaker1,
Hisham Almajed2, Irfan Mohiuddin2
1Department of Computer Science, COMSATS University Islamabad, Islamabad
44000, Pakistan
2Department of Computer Science, College of Computer and Information
Sciences, King Saud University, Riyadh 11633, Saudi Arabia
Correspondence:
Nadeem Javaid; (nadeemjavaidqau@gmail.com, www.njavaid.com),
Ahmad Almogren; (ahalmogren@ksu.edu.sa)
Abstract: In Internet of Things (IoTs), the increase in the number of devices is
directly proportional to the number of applications. The exponential growth of devices
increases both the network complexity and risk against topology robustness. Moreover,
the network is also prone to targeted and malicious attacks. In this paper, Enhanced Angle
Sum Operation ROSE (EASO-ROSE), Enhanced ROSE, Adaptive Genetic Algorithm
(AGA) and Cluster Adaptive Genetic Algorithm (CAGA) are proposed to cater the
topology robustness issue for IoT enabled smart cities. In addition, the proposed solutions
keep the nodes’ initial degree distribution unchanged by maintaining the scale-free
nature of the topology. Enhanced ROSE and EASO-ROSE significantly improve the
topology robustness by calculating the difference in nodes’ degree while rearranging
the surrounding angles according to the highest degree node. CAGA and AGA also
significantly improve the topology robustness by using adaptive probabilities of crossover
and mutation that guide both algorithms to converge towards global optimum solution.
Extensive simulations are preformed to evaluate the performance of the proposed
strategies. Schneider Ris used as a performance metric in the simulations. The results
depict that the proposed algorithms perform 61.3%, 48.3%, 45.5% and 34.95%, better
than simulating annealing algorithm.
Keywords: Internet of Things, topology robustness, malicious attacks, data driven, scale-
free.
1 Introduction
In the new era of automation, Internet is a global
network whose primary aim is to connect people and
smart devices together. Stupendous expansion of smart
devices emerges as Internet of Things (IoTs). Due to
massive increase in IoT devices, it is estimated that the
number of IoT devices will reach up to fifty million in
the near future S.E. Collier (2015). Different networks
are becoming a part of the IoT network, such as
fifth generation cellular network Tsai et al. (2015),
heterogeneous sensor networks T. N. Qureshi et al.
(2013), flying ad-hoc networks W. Zafar et al. (2016),
hybrid mobile networks Han et al. (2017), N. Javaid et
al. (2018), Waluyo et al. (2012), ad-hoc networks T. Qiu
et al. (2017), Nghiem et al. (2013), J. A and Stankovic
(2014), underwater Wireless Sensor Networks(WSNs) I.
Azam et al. (2017) and IoT-Enabled Smart Homes B.
Hussain et al. (2018). The IoT network becomes more
complex, complicated and vulnerable to fragmentation
by increasing the nodes’ density and network size.
Additionally, IoT nodes process a huge amount of
data that makes the network sensitive and prone to
serious challenges: topology robustness and delay. IoT
has a variety of applications, such as monitoring of
power networks, health care, transportation networks,
industrial and home automation systems, etc. These
applications result in excessive sensor nodes’ deployment
and conversion of a traditional city into a smart city S.E.
Collier (2015), T. Qiu et al. (2017) as described in Fig. 1.
Data is collected and sent to big data servers using the
sensors Taniar et al. (2017). Moreover, smart grids are
also part of the IoT network and they are considered as
one of the biggest applications of IoT S.E. Collier (2015).
Copyright ©201X Inderscience Enterprises Ltd.
2Talha Naeem Qureshi et al.
However, failure in the IoT network of synchronous grids
and power plants causes life threatening damage Taniar
et al. (2008).
IoT networks produce multiple types of data in
extensive quantity. In current era, topology robustness
of IoT networks (based on big data servers) has emerged
as a serious concern Wu et al. (2011). Small world and
scale-free are two commonly used models in complex
network theory. The main purpose of a scale-free model is
to model homogeneous network topologies. It has many
applications that include mobile networks N. Javaid et
al. (2017), cooperative networks N. Javaid et al. (2017),
social networks and transportation system networks
T. Qiu et al. (2017), Waluyo et al. (2009). A small
world model is generally used for heterogeneous network
topologies. It has a relatively high clustering coefficient
and a small average path length.
In IoT networks, the sensors have same
communication range and bandwidth. Therefore, scale-
free models are more suitable and applicable to the IoT
network than the small world model. In a scale-free
model, the degree distribution of nodes is based on power
law distribution resulting in several low degree nodes
T. Qiu et al. (2017). These nodes are often targeted by
random attacks and scale-free networks are proven to
be robust against these attacks J. Melthis et al. (2016),
I. Sohn et al. (2018). Whereas, high degree nodes are
highly vulnerable to intentional and targeted attacks R.
H. Li et al. (2012). Therefore, it is important to enhance
the network robustness and resilience of the scale-free
topologies against malicious attacks M. Mozafari et
al. (2019). Furthermore, scientists and researchers are
continuously finding ways to make robust scale-free
topologies against malicious, targeted, deliberate and
intentional attacks J. Liu et al. (2019), S. Wang et al.
(2019), S. Wang et al. (2019).
The main contributions of the proposed work are as
follows.
Multiple data driven strategies are proposed in
order to enhance the robustness of a scale-free IoT
network where deployed nodes are homogeneous in
nature.
Two enhancements of ROSE are proposed:
Enhanced ROSE and EASO-ROSE T. Qiu et
al. (2017). Both strategies increase topology
robustness of the scale-free IoT networks
without changing the nodes’ degree. Moreover,
the proposed work improves the topology by
making an onion-like structure. Furthermore, the
proposed algorithms outperform the conventional
techniques: ROSE and simulating annealing, in
terms of robustness against malicious attacks.
Schneider robustness metric Rbased adaptive
probabilities are proposed for crossover and
mutation in CAGA and AGA. Adaptive nature of
the probabilities ensures the convergence towards
a global optimal solution and retention of already
converged solutions.
CAGA integrates the clustering concept within
a solution space and results in fast convergence
towards global optima.
This work is an extension of T. N. Qureshi et al. (2019).
The rest of the paper is organized as follows. Detailed
related work is presented in Section 2. The proposed
models for the IoT network are given in Sections 3 and
4. Section 5 discusses the topology robustness metrices.
ROSE topology robustness algorithm is explained
in Section 6. The proposed robustness enhancement
schemes: EASO-ROSE, Enhanced ROSE, CAGA and
AGA are presented in Sections 7, 8, 9 and 10,
respectively. Simulation results of the proposed schemes
in comparison to previous schemes are given in Section
11. Finally, the paper is concluded in Section 12.
2 Related Work
Simulating annealing algorithm is used by P. Buesser
et al. (2011). The basic functionality of the simulating
annealing is similar to the Hill Climbing. However,
the algorithm includes edge swapping strategy, which
is based on probability that also records inferior
configurations in order to improve the topology
robustness. Optimization of a scale-free network is
performed using simulating annealing; however, this
algorithm is based on unnecessary comparison, which
results in decreased computational efficiency.
In A.-L. Barabsi et al. (1999), Barabasi-Albert (BA)
model is presented to generate the scale-free topologies.
The generated topologies are based on preferential
attachment concept where the nodes’ degree follows
power law distribution. The criteria for generating the
scale-free topologies in the BA model is given as follows.
The nodes are added in the network one by one
at each iteration, mNowhere mrepresents the
total number of existing nodes and Norepresents
the total number of nodes allowed in the network.
Whenever a new node is added to an existing
network, it selects the node iin order to connect.
Node is selected on the basis of probability, which
depends on its degree ki. It means that a node
with maximum edges has a high probability to
be selected as a next hop. This phenomenon is
known as preferential attachment and is described
as Matthew Effect.
Power law distribution is achieved by using the
above mentioned criteria. BA model proves that growing
network converges into a scale-free topology by power
law distribution, which is denoted as p(k)kγwhere
γ3. In the past few years, many researchers have
focused on multiple modeling strategies to enhance the
robustness and resilience of the scale-free networks.
Attack Resistance based Topology Robustness of Scale free IoT for Smart Cities 3
Figure 1: Proposed System Model for Enhanced ROSE and EASO-ROSE
Node
Node
Node
Decision
Processor
Modeling
Agriculture
Industry
Smart
Home
Transport
Security
Smart Cities Cloud
Big Data Server
Health
Care
Topology
Neighbour IDs
Neighbour X axis
Neighbour Y axis
Total Degree
Node List
In the traditional IoT networks, sensors have limited
communication range and some of them have insufficient
neighbors. It results in several preferential attachment
issues in the BA model. Due to these limitations, the
main focus of researchers is on the scale-free network
topologies. In the scale-free networks, there are few
high degree nodes, which are vulnerable to malicious,
intentional and targeted attacks. To tackle the high
degree nodes’ failure issue, a large number of edges are
removed, which causes the network fragmentation. In
order to cater this issue, additional edges are added
between the nodes. However, it requires additional
energy resources, wireless channels and bandwidth. The
structure of the scale-free network is changed that
increases the cost. Thus, there is a need to enhance
topology robustness of the scale-free networks by keeping
same initial degree distribution of the nodes.
In order to measure the robustness of the network
topology, C. M. Schneider et al. (2011), Adhinugraha et
al. (2020) present Schneider Rmetric. The metric refers
to the proportion of the largest connected cluster in
comparison to the total number of nodes. Schneider Ris
used to calculate robustness and resilience of the network
topology against intentional, targeted and malicious
attacks. It also considers the case in which the topology
suffers from a major loss without completely collapsing.
Schneider Ris more appropriate for the scale-free IoT
networks rather than previously used metrics.
Hill Climbing algorithm is proposed by H. J.
Herrmann et al. (2011). It uses Schneider Rrobustness
metric for the validation. In the proposed scheme, Monte
Carlo method is used as a base. It includes swapping
of the edges to make an onion-like structure while
maintaining the same degree distribution of nodes. As
a result, robustness of the proposed algorithm increases.
However, the multi-model phenomenon is based on
randomization, which leads algorithm to get stuck into
local optima.
G. Zheng et al. (2012), G. Zheng et al. (2013) propose
two network topologies that are based on a linear growth
evolution concept with accelerated growth evolution
phenomenon for robustness of the scale-free networks. A
practical situation regarding reconstruction of edges, is
considered in both proposed topologies. Energy Aware
BA (EABA) is proposed by Y. Jian et al. (2013). In this
scheme, energy consumption of sensors and connectivity
ranges for scale-free network topology are tuned by
configurable coefficients. The proposed model considers
both energy consumption and transmission performance;
however, data overhead is increased during modeling
4Talha Naeem Qureshi et al.
process. Similarly, L. Liu et al. (2014) propose a complex
network topology on the basis of both small world and
scale-free networks. As, the IoT nodes have limited
communication range; therefore, the proposed model is
not suitable for the IoT networks. However, model can
be implemented on heterogeneous networks.
V. H. P. Louzada et al. (2013) propose a smart
rewiring solution. A two-fold selection is used to find
edges, needed to be compared. Topology robustness
is optimized by converting topology into modular
onion-like structure along with computational efficiency.
The model is applied to world air transportation
network. Multi-model phenomenon prevents solution
from escaping the local optima. However, proposed
algorithm does not consider the limited communication
range of the nodes. Thus, it is not applicable to IoT
networks. Memetic algorithm is proposed by M. Zhou
and J. Liu et al. (2014). The proposed algorithm
performs local and global search in solution space to
improve the robustness of topology without changing
degree distribution. Two channels’ operations make the
algorithm complex. The algorithm does not consider the
change in communication range of sensor node, which
makes it complex for the IoT enabled networks.
The authors in W.B. Du et al. (2013) introduce
the shortest path queuing strategy in the BA model.
The strategy is designed for the scale-free networks
and is applied on the transportation system by the
authors. T. Qiu et al. (2016) propose a scheme to
enhance robustness of topological structure of the IoT
networks. The importance is defined by two greedy
factors. Additionally, the concept of local importance
of nodes is defined. Moreover, topology robustness is
enhanced by adding new edges, which change the initial
degree distribution. Thus, it increases the cost of the
system. In addition, the designed scheme does not
consider the limited range of sensor nodes.
ROSE algorithm is proposed in T. Qiu et al. (2017).
The proposed strategy is applied on the scale-free WSNs.
As a result, network topology robustness is increased
using various operations to convert the topology into
onion-like structure. In this scheme, the first node
is used as a reference node in angle sum operation.
Nevertheless, in real scenario, onion-like structure is
referred by the highest degree node. The correct selection
of reference node makes the topology more robust.
In T. Qiu et al. (2017), Genetic Algorithm (GA) is
applied on the IoT network. The proposed scheme
searches for the possible solution sets by Schneider R.
Moreover, the fixed crossover and mutation probabilities
are used, which result in convergence towards local
optima. However, premature and slow convergence lead
to poor performance.
3 Scale-free IoT model for Enhanced ROSE
and EASO-ROSE
In the IoT network, sensors have limited energy
resources and communication range. These limitations
have different impacts on the construction of the scale-
free topologies. These impacts are as follows.
Sensors are unable to establish arbitrarily large
distance communication links.
They have limited connections; hence, they have
limited degree (to keep energy consumption
optimal).
Due to aforementioned constraints, it is difficult to
implement the BA model A.-L. Barabsi et al. (1999).
The proposed system model is motivated from T. Qiu
et al. (2017), T. N. Qureshi et al. (2019), which is
used for topology construction in Enhanced ROSE and
EASO-ROSE, as shown in fig. 1. In order to avoid
aforementioned limitations, we deploy a dense network
topology for the IoT network. Moreover, sensors enable
links with all neighbors within their communication
range. As in the era of developed technology, the
communication ranges of the sensors are increased at
minimum cost. This idea leads the network towards the
dense deployment. Therefore, to support a variety of
dynamic services, a large number of nodes are deployed
in residential areas, smart buildings, smart shopping
centers and smart hospitals to form the IoT network. In
addition, significant research is carried out to support
dense network W. P. Tay et al. (2008), M. Hefeeda and
M. Bagheri (2007). The reasons to induce the concept of
a dense network are as follows.
To make the topology robust.
To maintain the network coverage where nodes are
likely to be failed.
To achieve the scale-free network properties, sensors
should have communication links with 50% of the nodes
available in the network. To achieve this, sensors must
have enough communication radius; otherwise, high
degree nodes will scatter in the network and lead the
network topology towards fragmentation.
To cater the issues of fragmentation of the network
topology and high degree nodes scattering in the
network, the proposed system model treats every
fragmented segment as a cluster. The scale-free IoT
based topologies are created by transforming each cluster
into an onion-like structure. Afterwards, we fuse all the
clusters in the end. The following are the steps for scale-
free topology construction in the IoT networks.
There are asynchronous communication links
between nodes, which are in the same range.
All the edges of the IoT sensor nodes contribute
and form the local world of a respective sensor
node.
Attack Resistance based Topology Robustness of Scale free IoT for Smart Cities 5
Onion Like Structure
Maximum Degree Node
Ring of Second High Degree Nodes
Ring of Third High Degree Nodes
Ring of Forth High Degree Nodes
Ring of Fi h High Degree Nodes
Ring of Low Degree Nodes
The edge having 90 degree or maximum angle
between centroid and mid point. Edge is having
high probability of having equal degree nodes.
Centroid
Figure 2: Onion-like Robust structure against Malicious and Targeted Attacks
High degree sensor nodes consume high energy
as compared to the low degree nodes. Moreover,
due to the depletion of energy, high degree nodes
fail more quickly than low degree nodes. As a
result, edges between nodes break and topology
gets fragmented.
Any senor node with the highest edge limit is not
able to create any extra connection or edge more
than a defined limit W. Xiao et al. (2015).
High degree nodes are more preferable to be joined
by the newly incoming nodes.
Roulette method is opted by incoming nodes to
define the degree limit.
A node is connected to the neighbors within its
communication range and form its local world. Whereas,
all nodes form connections on the basis of connection
probability PConnect. Hence, we define PC onnect for
neighbor node ias follows:
PConnect(i)=Degree(i)/
n
X
j=1
Degree(j),(1)
where the total number of nodes that exists in
a network is denoted by nand Degree(i)refers to
the total number of edges of node i. When a new
node joins the network, it sends its location to big
data servers after obtaining coordinates via Global
Positioning System (GPS). Then, it requests the big data
servers to obtain the data of its neighbors. The big data
server provides the neighbors’ information to all sensor
nodes. After specific interval of time, all nodes broadcast
their locations to the big data server. Big data server
contains the location information of all nodes belonging
to different IoT networks.
4 Scale-free IoT model for CAGA and AGA
The system model for CAGA and AGA algorithms is
motivated from T. Qiu et al. (2017), T. N. Qureshi
et al. (2019), as shown in fig. 3. The sink nodes get
information from topology engine and direct the sensor
nodes regarding that information. Then, data is saved
in big data servers connected to smart cities’ cloud for
further processing. Topology engine and modeling engine
connected to big data servers are responsible to finalize
the topology of the IoT networks. Afterwards, decision
processor applies optimization algorithm over topology
adjacency matrix, which is obtained through modeling
engine and calculates Schneider R. After that, modeling
engine compares the existing topology with proposed
topology on the basis of Schneider R. Then, topology
engine sends topology change information to all sensors
through sink nodes. Moreover, modeling engine also
computes the adjacency matrix of the topology, as shown
in fig. 4. Then, adjacency matrix is further converted into
chromosomes. Finally, multiple chromosomes are stored
in the population set, referred as solution space.
5 IoT topology robustness metrics
The potential of a network to resist disasters and failures
is termed as robustness T. N. Qureshi et al. (2019). A
network’s robustness is the prime factor that assesses the
resilience of the network against attacks. Some edges or
nodes are destroyed in an IoT network due to energy
depletion or by deliberate malicious attacks. As a result,
network is fragmented. Moreover, preliminary topology
connections are broken down. These attacks are classified
into two categories.
Random Attacks: these attacks fragment the
network topology via random selection and
destruction of nodes.
6Talha Naeem Qureshi et al.
Topology
Topology Engine
Topology Extracted
Chromosome
i
j
l
i
0
1
1
j
1
0
1
k
1
0
1
l
1
1
0
0
1
1
1
0
0
1
0
1
0
Adjacency Matrix
AGA
Population
CAGA Clusters
Cluster 1
Cluster 2
Cluster 3
Cluster 4
Decision
Processor
Big Data Server
Modeling
Agriculture
Industry
Smart
Home
Transport
Security
Smart Cities Cloud
Health
Care
Chromosome
Figure 3: Proposed System Model for CAGA and AGA
IoT Topology Node l
Node k
Node i
Node j
i
j
k
l
i
0
1
1
1
j
1
0
0
1
k
1
0
0
1
l
1
1
1
0
0
1
1
1
0
0
1
0
1
0
Adjacency Matrix
Chromosome
Figure 4: Adjacency Matrix Conversion to Binary Chromosome
Malicious or Targeted Attacks: these are the
priority based attacks in which all nodes are
prioritized on the basis of their edge density.
Afterwards, nodes having the highest number of
edges are removed, resulting in several connections
breakage at the same time. This process is repeated
over and over again until the whole network is
fragmented. This strategy fragments the complete
topology by attacking few nodes having critical
importance.
All proposed schemes namely, EASO-ROSE, Enhanced
ROSE, CAGA and AGA, prioritize the nodes on the
basis of their degree distribution. The degrees of all
the nodes in IoT topology are extracted from big data
servers of the smart cities. Existing topology robustness
is calculated by malicious or targeted attacks. Prior
to calculate topology robustness, nodes are prioritized
on the basis of their number of edges. Then, nodes
are destroyed by starting from high priority node to
low priority node. The process of prioritized node’s
destruction continues until the complete topology is
fragmented and nodes are isolated. If a network has more
than one high degree nodes, then it randomly elects a
single node and eliminates it from the network.
In topology robustness, it is the proportion of attacks
or percolation threshold resulting in entire structure
crumbling. Besides percolation theory, there are various
robustness benchmarks that employ shortest path Frank
et al. (1970), Latora et al. (2001), Sydney et al. (2008)
and graph spectrum concepts Fiedler et al. (1973). A
threshold value for prioritized destruction of nodes is
denoted by qcthat is the proportion of amount of attacks
required to entirely paralyze and isolate the network
Holme et al. (2002). A condition may arise in which a
network may not completely collapse; however, it may
suffer from a huge damage.
Percolation theory based topology robustness metric
Schneider Ris proposed by C. M. Schneider et al. (2011).
Attack Resistance based Topology Robustness of Scale free IoT for Smart Cities 7
If a highest degree node of the network is attacked then
it is removed from the network. Then, Schneider Ris
used to calculate robustness. It is relatively simple in
terms of computations and bit complex in terms of the
highest degree node elimination. Schneider Rcertifies the
resilience of topology that after how many destructive
attacks on high degree nodes, the network still remains
operative. We employ this enhanced standard to evaluate
the proposed algorithms against intentional, deliberate,
malicious and targeted attacks. This metric calculates
the maximum degree of the nodes, sub graphs in
the topology and also estimates the magnitude of
impairment formed due to it. Schneider Ris described
as per equation below:
R= 1/N
N
X
Q=1
s(Q),(2)
where total number of nodes in the topology is
represented by N, the proportion of nodes to be deleted
is denoted by Q, fraction of sensor nodes in the largest
connecting portion after failing QN high degree sensor
nodes is denoted by s(Q). In order to evaluate the
networks of different sizes, normalization factor 1/N is
used. Schneider Rvaries from limit 1/N to 0.5. Where
the initial limit (1/N) represents the robustness of an
isolated graph and 0.5 represents the robustness of a full
mesh topology.
6 ROSE
ROSE T. Qiu et al. (2017) is designed to work in a
centralized system. All nodes share their coordinates
with centralized system. Then, neighbor list is calculated
and shared with all nodes. ROSE optimizes the
network topology by converging towards an onion-like
structure. According to C. M. Schneider et al. (2011),
nodes exhibiting onion-like formation are more resilient
against targeted and intentional attacks. The important
characteristic of an onion-like network is that it has high
degree nodes at the center. These nodes form a core
surrounded by low degree nodes attached together in
the form of a ring. The degrees of the nodes gradually
decrease from inner to outer side of onion. ROSE
transforms the topology into onion-like structure as
explained by Schneider (through tri-phase operation,
based on onion-like structure). In which, first phase is
the detection of independent edges followed by second
phase of degree difference operation and finally the third
stage of angle sum operation.
6.1 Independent edges
We relate the graph as G= (V , E) having Vvertices
and Eedges and form a scale-free network. The whole
network consists of Nnumber of nodes and Eedges
belonging to set M. The following criteria is the baseline
for the selection of independent edges:
All the nodes i, j, k and lincluding edges eij and
ekl should lie in the communication range of each
other.
Nodes iand jshould not have any edge with node k
and l, and vice versa. Fig. 5(a) shows the selection
of independent edges eij and ekl.
6.2 Degree difference operation
C. M. Schneider et al. (2011) and M. Zhou and J. Liu
et al. (2014) found that onion-like structures are robust
against intentional and targeted attacks. It is also noted
that nodes belonging to interconnected rings around the
core are of the same degree. If a high degree node fails
to perform its operation then another high degree node
performs its tasks to make the topology resilient. Then,
to achieve the goal, high degree nodes are connected
with other high degree nodes and low degree nodes are
connected with low degree nodes. As a result, degree
difference between two nodes in the same ring is reduced.
The basic aim of this operation is to keep the degree
difference between connecting nodes as low as possible.
After selection of independent edges, ROSE T. Qiu et al.
(2017) calculates the degree of all nodes from both edges
as di, dj, dkand dl. Further, degree differences SU B0,
SU B1and SU B2are calculated from equations 3 to 5
T. Qiu et al. (2017). All types of swaps between node i,
j,k,lare performed and recorded to calculate topology
robustness using Schneider R.
SU B0=|didj|+|dkdl|,(3)
SU B1=|didk|+|djdl|,(4)
SU B2=|didl|+|djdk|.(5)
After the calculation of Rfor all connection options (as
shown in fig. 5), the connection method with maximum
Ris selected. Probability pis calculated to control the
degree difference operation T. Qiu et al. (2017). Where,
value of plies between range (0,1). Probability pis
calculated as equation 6.
p=max(|didk|+|djdl|
|didj|+|dkdl|,|didl|+|djdk|
|didj|+|dkdl|)(6)
pacts as a controlling threshold for swapping. If the
selected edges have low degree differences as compared
to the previous scenario then the value of Ris checked
for all connection options, as illustrated in fig. 5. The
option meeting the criteria of palong with increasing
Schneider Ris finally selected.
6.3 Angle sum operation
Onion-like topology is the alignment of nodes and edges
in a ring like pattern. In parallel to the ring pattern
nodes, some horizontal edges also exist with respect
to the core of the onion. ROSE T. Qiu et al. (2017)
takes advantage of these edges and further enhances the
8Talha Naeem Qureshi et al.
Algorithm 1 Degree Difference Operation
1: Input: V,E,listiand listnodedegr ee
2: Collect neighbor list (listi) and degree distribution list (listnodedegr ee) for nodes (N) from big data servers of
smart city
3: Plot topology G= (V, E) based on data driven from listi
4: Randomly select two edges eij and ekl
5: if eij and ekl are independent edges then
6: Calculate the number of neighbors nifrom listifor node i,j,kand l
7: Degree diSum of all neighbors ni
8: Degree djSum of all neighbors nj
9: Degree dkSum of all neighbors nk
10: Degree dlSum of all neighbors nl
11: SU B0=|didj|+|dkdl|,SU B1=|didk|+|djdl|,SU B2=|didl|+|djdk|
12: SU Bmin=min(SU B0,SU B1,S U B2)
13: end if
14: if SU Bmin =SU B0then
15: No change in topology G
16: end if
17: if SU Bmin =SU B1then
18: Remove edges eij,ekl and Add edges eik ,ej l
19: GG0
20: Calculate R(G0)
21: if R(G0)> R(G), G=G0then
22: end if
23: end if
24: if SU Bmin =SU B2then
25: Remove edges eij,ekl and Add edges eil ,ejk
26: GG0
27: Calculate R(G0)
28: if R(G0)> R(G)then
29: G=G0
30: end if
31: end if
32: Output: Topology=G
robustness by performing angle sum operation over these
edges. The operation further transforms the topology
into onion-like structure by utilizing the effect of these
horizontal edges. To start with this operation, ROSE
selects the central point of topology as centroid. After
selection of centroid, a random edge eij is elected.
Surrounding angle between centroid and edge eij is
calculated as per equation 7 T. Qiu et al. (2017). The
calculation of angle starts after calculating a center point
of the edge eij and then by drawing a line from center
point to the centroid. The angle formed between drawn
line and edge eij is denoted as αand β. The smaller
angle is recorded as α; whereas, βrepresents the larger
angle. Edge eij is taken as lencfor calculation of αas
shown in fig. 6(a). Virtual lines between centroid to node
iis drawn and recorded as lenb.lenais the line between
centroid and node j. The whole graphical explanation is
shown in fig. 6(a).
α=arccos(len2
b+len2
clen2
a
2lenblenc
).(7)
Angle αwill be subtracted from 180oto calculate
angle βas per equation 8:
β= 180oα. (8)
From above equations, αand βare calculated for edge
ekl. All possible swaps are performed as per fig. 6(a),
(b) and (c). Angles αand βare calculated in the same
manner for all possible swaps. For assurance, Schneider
Ris again calculated as per algorithm 1.
7 EASO-ROSE
ROSE T. Qiu et al. (2017) is developed to handle the
centralized systems. However, unlike ROSE, proposed
EASO-ROSE extracts nodes’ geographic location and
calculates neighbor information from big data servers.
All the nodes broadcast their location obtained through
built-in GPS to the big data server in smart cities for
localization purposes.
EASO-ROSE performs the degree difference
operation the same as ROSE and calculates the degree
Attack Resistance based Topology Robustness of Scale free IoT for Smart Cities 9
Algorithm 2 EASO
1: Randomly select two independent edges eij and ekl
2: if eij and ekl are independent edges then
3: Calculate number of neighbors nifrom listifor node i,j,kand l
4: end if
5: Calculate highest degree node, Highestmax =max(listnodedeg ree)
6: Select Centroid =Highestmax
7: for Edges eij and ekl do
8: Surrounding angle α with respect to Centroid for edge eij
9: Surrounding angle β with respect to Centroid for edge ekl
10: SU M1=α+β
11: end for
12: Remove edges eij ,ekl and Add edges eik,ej l
13: for Edges eik and ejl do
14: Surrounding angle α with respect to Centroid for edge eik
15: Surrounding angle β with respect to Centroid for edge ejl
16: SU M2=α+β
17: end for
18: Remove edges eij,ekl and Add edges eil ,ejk
19: for Edges eil and ejk do
20: Surrounding angle α with respect to Centroid for edge eil
21: Surrounding angle β with respect to Centroid for edge ejk
22: SU M3=α+β
23: end for
24: Calculate Maximum Sum SU Mmax =max(SUM1, SU M2, SU M3)
25: then Swap edges with respect to SU Mmax GG0
26: Calculate R(G0)
27: if R(G0)> R(G) , G=G0then
28: end if
29: Output: Topology=G
(a) (b) (c)
ij
k
l
ij
k
l
ij
k
l
Figure 5: Exchange Options of Independent Edges, (a) Original Topology, (b) Exchange Option 1, (c) Exchange
Option 2
difference SU B0, SU B1and SU B2for all possible
swap combinations. Swap combinations with the lowest
degree difference are selected. After that Schneider Ris
calculated to assure the increase in topology robustness
as per algorithm 2. ROSE considers the very first node
of topology as reference and centroid for angle sum
operation. Afterward, it converts the topology to onion-
like structure. According to the C. M. Schneider et al.
(2011) and M. Zhou and J. Liu et al. (2014), core or
high degree nodes are interconnected with each other
and form a ring like structure to be resilient against
network attacks. As we move from inner to outer side of
the onion-like structure, high degree nodes are further
surrounded by rings of low degree nodes. Mostly, the
rings of the nodes are not symmetric. It means that
core of high degree nodes is highly probable, not to
be situated at exact center of the onion. To cater
the mentioned issue, proposed scheme EASO-ROSE
10 Talha Naeem Qureshi et al.
(b) (c) (d)
Len a
(a)
c
c
c c
ij
k
Len b
Len c
i
ii
j
m
m1
m2
m
m2 m2
m1
Figure 6: (a) Geometric Details of Angle Sum Operation (b) First Edge Swap Possibility, (c) Second Edge Swap
Possibility, (d) Third Edge Swap Possibility
requests the node’s degree information from big data
servers and calculates the coordinates of high degree
nodes. Afterwards, these core high degree nodes are
taken as centroid to calculate αand β. Equation 9 is
used for calculation of the highest degree node.
Highesti=max(Listnodedegr ee).(9)
Centroid is elected on the base of the highest degree
node using equation 10:
Centroid =Highesti.(10)
Where, surrounding angles (α) are calculated on the base
of Highestias follows:
α=arccos len2
b(Highesti)+len2
c(Highesti)len2
a(Highesti)
2lenb(Hig hesti)lenc(H ighesti)
.(11)
8 Enhanced ROSE
Enhanced ROSE is based on data driven approach. In
which all the nodes send the geographic location to
a big data server using GPS sensor. Big data servers
calculate the neighbor list of each sensor on the base of
their communication range. Afterwards, big data server
calculates the degree of each node and maintain a list.
Initial steps of Enhanced ROSE and ROSE are
same, i.e., “Selection of Independent Edges” and “Degree
Difference Operation”. SUB0, S UB1and SU B2are
calculated in degree difference operation and the one
with the lowest value is selected to swapping. Swap
is re-validated by finding Schneider R. If the value of
Rincreases then the swap is accepted; otherwise, it is
reverted back.
Enhanced ROSE and arccos function calculate
the surrounding angle αsame as ROSE. After that
Enhanced ROSE transforms the topology into onion-
like structure using algorithm 3. At the final step, two
independent edges are selected from the topology as evx
and eyz . Then, degree information for nodes v, x, y and z
is extracted from big data server’s database. Afterward,
Enhanced ROSE lists the degrees in descending order
and name them as Enhanced1, E nhanced2, Enhanced3
and Enhanced4. Where, Enhanced ROSE utilizes the
degrees in the formula below to obtain SU BE nhanced.
SU BEnhanced1=|Enhanced1Enhanced2|+
|Enhanced3Enhanced4|(12)
Then another degree difference SU BEnhanced2is based
on following formula:
SU BEnhanced2=|DegreevDegreex|+
|DegreeyDegreez|(13)
Ratio PEnhanced is based on SU BE nhanced1and
SU BEnhanced2as per formula below.
PEnhanced =SU BE nhanced1
SU BEnhanced2
(14)
We use a swap threshold Sthreshold to limit the number
of swaps. The range of swap threshold lies between
(0,1). Then, we compare the PEnhanced with Sthreshold .
If PEnhanced is less than Sthreshold , then evx and ey z
edges are swaped, accordingly.
9 Adaptive Genetic Algorithm
Dual goals of preserving diversity in population and
maintaining the convergence ability of GA are achieved
through adaptive probabilities of crossover and mutation
M. Srinivas et al. (1994). Ability to locate and
converge towards the global optimum solution in
multi-model landscape is the robustness attribute of
Attack Resistance based Topology Robustness of Scale free IoT for Smart Cities 11
Algorithm 3 Enhanced Convergence Operation
1: Randomly select two independent edges eij and ekl
2: if eij and ekl are independent edges then
3: Calculate number of neighbors (ni) from listifor node i,j,kand l
4: end if
5: highest degree node, Highestmax =max(listnodedeg ree)
6: Select Centroid =Highestmax
7: for Edges eij and ekl do
8: Surrounding angle α with respect to Centroid for edge eij
9: Surrounding angle β with respect to Centroid for edge ekl
10: SU M1=α+β
11: end for
12: Remove edges eij,ekl and add edges eik ,ej l
13: for Edges eik and ejl do
14: Surrounding angle α with respect to Centroid for edge eik
15: Surrounding angle β with respect to Centroid for edge ejl
16: SU M2=α+β
17: end for
18: Remove edges eij,ekl and add edges eil ,ejk
19: for Edges eil and ejk do
20: Surrounding angle α with respect to Centroid for edge eil
21: Surrounding angle β with respect to Centroid for edge ejk
22: SU M3=α+β
23: end for
24: Calculate Maximum Sum SU Mmax =max(SUM1, SU M2, SU M3)
25: then Swap edges with respect to SU Mmax GG0,Calculate R(G0)
26: if R(G0)> R(G)G=G0then
27: end if
28: Randomly select two independent edges evx and eyz then Calculate degrees of nodes v, x, y and zthrough
listnodedegree
29: Randomly select two independent edges evx and eyz then Calculate Degree of nodes v, x, y and zas through
listnodedegree
30: Arrange Degrees of node v, x, y , z in descending order Enhanced1, E nhanced2, Enhanced3and Enhanced4
31: Calculate SU BEnhanced1=|Enhanced1Enhanced2|+|Enhanced3Enhanced4|,SUBE nhanced2=
|DegreevDegreex|+|DegreeyDegreez|and PE nhanced =SU BEnhanced 1
SU BEnhanced2
32: if PEnhanced < Sthreshold swap edges and new edges are evy and exz then
33: end if
GA. Probabilities of crossover Pcand mutation Pm
are tuned according to fitness values of the result.
The solution with increased robustness is protected
while solutions with low robustness value are entirely
discontinued. Therefore, to locate global optimum
solutions, classical GA uses random search in crossover
and mutation processes. Where crossover occurs with
some probability Pc. AGA operation is described
in algorithm 4. Moreover, adaptive probabilities of
crossover and mutation enable multi-goal achievement
by GA. It enables diversity in the solution space or
population along with enhanced convergence capacity
of GA (towards global optima.) Adaptive Pcand Pm
relief the system from manual and optimized selection of
probabilities. Rmax Rdecreases with the IoT topology
while converging towards the local optima. Where Rmax
is the maximum fitness value of the population and Ris
the average fitness value before and after the crossover
operation.
In AGA, adaptability varies with the values of Pc
and Pmaccording to the fitness function represented as
Schneider R. When solution stucks into local optima,
proposed algorithm increases the probabilities of Pc
and Pm. Moreover, when solution lies near to global
optima, it decreases the probabilities of Pcand Pm.
Concept is achieved by calculating the fitness function
for first population. Then fitness value is compared with
maximum fitness value Rmax of the population that is
0.5. Ris the average fitness value after and before the
crossover operation Pcand Pmwhile converging the
solution towards the local optima. Therefore, Pcand Pm
change inversely when Rmax Rdecreases. The Pcand
Pmat this stage are calculated using equation 15 and 16.
Both equations are derived by taking motivation from
M. Srinivas et al. (1994).
Pc=C1
Rmax R,(15)
12 Talha Naeem Qureshi et al.
Algorithm 4 AGA Mechanism
1: Input: V,E,listi,listnodedegr ee and φ[0,1]
2: Collect listineighbor list and listnodedegr ee degree distribution list for Nnodes from big data servers of smart
city
3: Plot topology G= (V, E) based on data driven from listi
4: Calculate Adjacency Matrix = Adjmatrix
5: Convert Adjmatrix into chromosome and add to population P opi
6: Select parent chromosomes Chromosomeparent by crossover probability Pc
7: Calculate Schneider R(Chromosomepar ent1&Chromosomeparent2[P opi])
8: Calculate exclusive eij |eij Chromosomepar ent1&eij 6∈ Chromosomeparent2
9: Calculate exclusive ekl |ekl Chromosomeparent2&ekl 6∈ C hromosomeparent1
10: if eij and ekl are exclusive then
11: Calculate tempneighbours1, neighbors nl|nlChromosomepar ent1&nl6∈ nk
12: Calculate tempneighbours2, neighbors nj|njChromosomepar ent1&nj6∈ ni
13: Calculate distances of all tempneighbours1&tempneighbour s2
14: sort according to distances
15: end if
16: Candidate1= least distance node from nktempneighbours1
17: Candidate2= least distance node from njtempneighbours2
18: Calculate tempneighbours3, neighbors Candidate1|Candidate1Chromosomeparent1&Candidate16∈ nl
19: Calculate tempneighbours4, neighbors Candidate2|Candidate2Chromosomeparent2&Candidate26∈ nj
20: Calculate distances of all tempneighbours3&tempneighbour s4
21: sort according to distances
22: Remove edge el,Candidate1,ej,Candidate2,ek,C andidate3and ei,C andidate4
23: Create edge (ekl &eCandidate1,C andidate3)Chromosomepar ent1& (eij &eCandidate2,C andidate4)
Chromosomeparent1
24: Randomly select two independent edges eij and ekl
25: if eij and ekl are independent edges then
26: Calculate number of neighbors nifrom listifor node i,j,kand l
27: degree diSum of all neighbors ni
28: degree djSum of all neighbors nj
29: degree dkSum of all neighbors nk
30: degree dlSum of all neighbors nl
31: Sort degree (di, dj, dkand dl) in descending order as (d1< d2< d3< d4)
32: end if
33: Calculate sl= (d1d2)+(d3d4), s2= (didj)+(dkdl) and p= (s1/s2)
34: if p<φthen
35: Exchange edges as per (d1, d2, d3and d4)
36: end if
37: Calculate Schneider R0
38: if if R(G0)> R(G)then
39: G=G0
40: end if
41: Note: we have used the symbols “& and |” for “logical AND and OR”.
Pm=C2
Rmax R.(16)
High values to Pcand Pmeffect the already
converged solutions. To deal with this, Pcand Pmshould
vary directly as Rmax R0where R0is the fitness value
of majority of the chromosomes. Now, equation 17 and
18 are used to find values of Pcand Pm, derived by
motivating from M. Srinivas et al. (1994).
Pc=C1Rmax R0
Rmax R,(17)
Pm=C2Rmax R0
Rmax R,(18)
where, C1and C2are constants.
9.1 Crossover in IoT
In GA, a large population space consists of solutions
obtained by applying crossover and mutation operations
on parent chromosomes (Chromosomeparent ) to form
child chromosomes by retaining a part of parent genes.
Where the best solution is required in large solution
space. In proposed algorithms, degree distribution of
Attack Resistance based Topology Robustness of Scale free IoT for Smart Cities 13
Algorithm 5 CAGA Mechanism
1: Input: V,E,listi,listnodedegr ee,C lustersiz e,P opf iltered and φ[0,1]
2: Collect listineighbor list and listnodedegr ee degree distribution list for Nnodes from big data servers of smart
city
3: Plot topology G= (V, E) based on data driven from listi
4: Calculate Adjacency Matrix = Adjmatrix , convert into chromosome and add to population P opi
5: Create clusters Cluster1to Clusternas per C lustersiz e factor defined
6: for each Clusterido
7: Calculate crossover probability Pcon basis of Schneider Rof cluster 1
8: Select parent chromosomes Chromosomeparent cluster i
9: Crossover parent chromosomes Chromosomeparent cluster i
10: Mutation parent chromosomes Chromosomeparent cluster i
11: Calculate Schneider R0If R(G0)> R(G)then G=G0,Best Chromosomecluster i =G
12: end for
13: Elect the Best Chromosome and sent to Filtered population P opf iltered
nodes is kept unchanged as described by T. Qiu et al.
(2017). Parent topologies perform crossover operation
with each other to form new child topologies. Where
parent topologies are selected on the basis of adaptive
crossover probability Pc.
Crossover process starts with the extraction of
exclusive edges from both parents. The term exclusive
edges represent the edges, which are only present in
one parent. Topology of Chromosomepar ent1forms an
exclusive edge with Chromosomepar ent2and similarly
Chromosomeparent1as described by T. Qiu et al. (2017).
Complete operation is given in algorithm 4 T. Qiu et al.
(2017).
9.2 Mutation in IoT
The goal of mutation is to introduce new individuals in
GA and to reproduce lost fitness of GA. It also avoids
the solution to stuck into local optima and to stop
from premature convergence. It also helps to find the
global fittest solution from solution space and to move
towards the global optima. In proposed AGA, mutation
is performed with adaptive probability Pmand rest of
the process occurs as described by T. Qiu et al. (2017);
however, instead of static probability, it uses adaptive
probability. Mutation phase starts with the selection of
independent edges as explained in previous section 6.1,
graphically explained in fig. 5 and as per criteria are
given below:
eij =(eij Chromosomeparent1,
eij 6∈ Chromosomeparent2,(19)
ekl =(ekl Chromosomeparent2,
ekl 6∈ Chromosomeparent1.(20)
After selection of independent edges eij and ekl, the
degrees for nodei,nodej,nodekand nodelare sorted
in descending order. We call them node1,node2,node3
and node4. We calculate p,s1and s2as per below
expressions, described by T. Qiu et al. (2017):
s1= (d1d2)+(d3d4),(21)
s2= (didj)+(dkdl),(22)
p=s1/s2.(23)
After calculating p, we compare it with threshold φthat
ranges between (0,1). If p<φ then we exchange edges
according to descending order as (d1, d2, d3and d4).
Mutation operation is stated in algorithm 4.
10 Cluster Adaptive Genetic Algorithm
In CAGA, the whole population is divided into optimum
sized clusters J. Zhang et al. (2007). The increase
in the number of clusters leads to increase in both
computational and time complexity. To cater the
aforementioned issues, optimal number of clusters are
found and fixed as per equation 24.
Clustersize =T otal chromosomes in population
4.(24)
All the clusters act independently. Then, values of Pcand
Pmare calculated adaptability using equations 17 and
18. Adaptive probabilities are calculated for each cluster,
which are independent and different. After crossover and
mutation process, best chromosome of each cluster is
extracted on the basis of the highest level of Schneider
Rvalue. All the best chromosomes of each cluster are
compared on the basis of Schneider R. The chromosomes
with the highest Schneider Rvalue are finally exported
to filtered population P opf iltered . Parallel processing in
all clusters leads towards fast convergence of algorithm
and better results. Complete operations of CAGA are
given in algorithm 5 and graphically illustrated in fig. 7.
14 Talha Naeem Qureshi et al.
0
1
1
1
0
0
1
0
1
0
0
1
0
0
0
1
1
1
0
1
1
1
1
0
1
0
1
1
0
0
0
1
1
1
0
0
1
0
1
0
1
0
1
0
1
0
1
0
1
1
0
1
1
1
0
0
1
0
1
0
0
1
0
0
0
1
1
1
0
1
1
1
1
0
1
0
1
1
0
0
0
1
1
1
0
0
1
0
1
0
1
0
1
0
1
0
1
0
1
1
0
1
1
1
0
0
1
0
1
0
0
1
0
0
0
1
1
1
0
1
0
1
1
0
1
0
1
1
0
0
0
1
1
1
0
0
1
0
1
0
0
1
0
0
0
1
1
1
0
1
1
0
1
0
1
0
1
0
1
1
Cluster 1
Cluster 2
Cluster 3
Cluster 4
Best Chromosome ex Cluster 1
Best Chromosome ex Cluster 2
Best Chromosome ex Cluster 3
Best Chromosome ex Cluster 4
0
1
1
0
1
0
1
1
0
1
0
0
1
0
1
1
1
1
0
0
0
1
0
0
1
0
1
1
0
1
0
0
1
0
1
1
1
1
0
0
0
1
1
0
1
0
1
1
0
1
Comparing Fitness Value of
Chromosomes Filtered Population
Figure 7: Population Conversion to Clusters and Election of Best Chromosome
11 Simulations and results
After extracting nodes’ deployment locations from smart
cities, IoT topology is simulated using MATLAB. It
helps the cities to become smart and economical by
adopting a circular design S. Prendeville et al. (2018).
Hence, in order to simulate a smart city, the circular
region is selected with 500 meter diameter. In this
regard, a network of 100 IoT sensors is deployed in the
restricted circular diameter. Moreover, to validate the
concept of scale-free topology, communication range of
each IoT sensor is restricted to 200 meters. Furthermore,
for Enhanced ROSE and EASO-ROSE, edge density is
restricted to 40% of the total nodes in the network.
Whereas, for CAGA and AGA, edge density is fixed to
60%.
11.1 Comparison of Enhanced ROSE and
EASO-ROSE with Existing Algorithms
Enhanced ROSE and EASO-ROSE are compared
with ROSE and simulating annealing algorithm. 100
iterations are performed for all algorithms under the
same environment. The results are averaged over
20 simulations. Enhanced ROSE and EASO-ROSE
outperformed ROSE and simulating annealing. The
best results (Schneider Rvalue) obtained by Enhanced
ROSE, EASO-ROSE, ROSE and simulating annealing
are 0.150, 0.126, 1.120 and 0.077, respectively, as shown
in fig. 8.
EASO-ROSE performed better than ROSE because
of modification in the angle sum operation phase of
ROSE. Original ROSE considers 1st node centroid and
calculates the surrounding angles. However, in original
onion-like topology, high degree nodes interconnected
with each other form a core, which is surrounded by
rings of low degree interconnected nodes. Unlike ROSE,
EASO-ROSE uses the highest degree node centroid.
Enhanced ROSE outperforms its counter schemes in
terms of efficiency. The strategy uses the highest degree
node as centroid. Afterwards, similar degree nodes
tend to connect with each other, which converts the
topology into onion-like structure. It leads to the increase
in topology robustness against malicious attacks. All
algorithms are tested over a network of 100 to 300
nodes. Enhanced ROSE has proven to be the most
efficient in converging the topology to robust onion-like
structure. Enhanced ROSE has 17.4%, 22.2% and 61.3%
better efficiency as compared to EASO-ROSE, ROSE
and simulating annealing , respectively. Whereas, EASO-
ROSE shows 4.9% and 48.3% better results as compared
to ROSE and simulating annealing. The results are
shown in fig. 9.
0 10 20 30 40 50 60 70 80 90 100
Number of Iterations
0.05
0.1
0.15
Schneider R
Enhanced ROSE
EASO-ROSE
ROSE
Simulating Annealing
Figure 8: Performance comparison between Enhanced
ROSE, EASO-ROSE and other algorithms
100 150 200 250 300
Number of Nodes
0.05
0.1
0.15
0.2
0.25
0.3
Schneider R
Enhanced ROSE
EASO-ROSE
ROSE
Simulating Annealing
Figure 9: Bar comparison between Enhanced ROSE,
EASO-ROSE and other Algorithms with Different Sizes
of Network Topologies in IoT
Attack Resistance based Topology Robustness of Scale free IoT for Smart Cities 15
11.2 Comparison of CAGA and AGA with
Existing Algorithms
In the proposed system, the nodes’ geographic
information obtained from big data server is converted
into adjacency matrix for further processing of CAGA
and AGA. Few parameters are adjusted after numerous
independent simulations including population size,
which is based on the number of chromosomes in
a population. 100 iterations are performed after
considering numerous factors including time complexity
and energy consumption. It is observed that the
proposed algorithms are converged to optimal results
within 100 iterations. Few parameters including Pc
and Pmare adaptively set by the proposed algorithms
depending on the fitness of the population.
For comparison, neighourlimit is set to 20. All the
results are the average of 20 independent simulations.
In fig. 10, y-axis shows the improvement in Schneider
Rfor all compared algorithms during each iteration. In
CAGA, the value of Schneider Ris minimum in the start.
Then, till 22nd iteration, there is a rapid improvement
in the value of Schneider Rdue to conversion of the
population into independent clusters and treating each
cluster with different Pcand Pm. The best chromosome
of each population is extracted during each iteration
and transferred to filtered population. All transferred
chromosomes from clusters are evaluated using fitness
function and only the best one is retained. Till 22nd
iteration, parallel processing of crossover and mutation
phases results in rapid increase in Schneider Rin
comparison to AGA, simulating annealing and hill
climbing algorithms. Overall, CAGA performs better
as compared to simulating annealing and hill climbing.
After 22nd iteration, the value of Schneider Rincreases
slowly, this is because the value is already increased
to a considerably high level. It means that each
chromosome in each cluster is at considerably good
fitness level. CAGA and AGA are kept on increasing
till 100th iteration. It is important to be noted that
after 50 iterations, the graph of simulating annealing
and hill climbing are depicting a straight line. This is
because both algorithms are trapped in local optima.
Whereas, adaptive mutation probability Pmmakes
CAGA and AGA to come out of local optima and
move towards global optima. AGA performs better
than simulating annealing and hill climbing but not
well in comparison to CAGA. AGA also keeps on
increasing till end and jumps out from local optima
due to adaptive mutation criteria. Simulating annealing
performs better than hill climbing because simulating
annealing also accepts inferior configurations with some
positive probability that is discarded in hill climbing.
However, performance of simulating annealing and hill
climbing is significantly low in comparison to CAGA and
AGA. CAGA shows 10.94%, 36.3% and 45.5% better
performance as compared to AGA, simulating annealing
and hill climbing algorithms. Whereas, AGA performs
25.6% and 34.95% better than simulating annealing and
hill climbing.
0 10 20 30 40 50 60 70 80 90 100
Number of Iterations
0
0.05
0.1
0.15
0.2
0.25
0.3
Schineder R
CAGA
AGA
SIMULATING ANNEALING
HILL CLIMBING
Figure 10: Performance Comparison between CAGA,
AGA and other Algorithms
For comparison, the value of neighourlimit is set to
20 for 100 nodes, it increases by 5 points when 50 new
nodes are added in the network on each iteration. A
scale-free topology is validated by varying the number
of nodes from 100 to 300 while neighourlimit is kept the
same for all algorithms. Fig. 11 shows the performance
comparison of the proposed algorithms CAGA and AGA
with simulating annealing and hill climbing algorithms
by averaging 20 independent simulations. It can be
graphically observed that value of Rdecreases for all
algorithms with increase in IoT enabled nodes’ density.
For large networks fragmentation resistance is more
complicated as compared to small networks. CAGA
and AGA show better robustness as compared to other
two algorithms. Scalability analysis also validates the
efficacy of proposed CAGA and AGA. Moreover, all
IoT topologies remain connected after implementing the
optimization algorithms.
100
150
200
250
300
Number of Nodes
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
Schineder R
CAGA
AGA
SIMULATING ANNEALING
HILL CLIMBING
Figure 11: Bar Comparison between CAGA, AGA
and other Algorithms with Different Sizes of Network
Topologies in IoT
12 Conclusion
This paper improves the topology robustness of a
complex IoT network using two proposed models. The
first model consists of Enhanced ROSE and EASO-
ROSE strategies while the second model contains
16 Talha Naeem Qureshi et al.
CAGA and AGA algorithms. In the former, the nodes’
geographical location and information are provided by
the smart cities’ cloud server to get the exact coordinates
of the nodes. Also, the selection of the highest degree
node enhances the reliance and topology robustness
by converting the network into an onion-like structure.
On the other hand, in the latter, AGA uses adaptive
probabilities and maintains a balance between local and
global optima. It prevents the optimal solution from
getting stuck in the local optima and moves it towards
the global optima. Whereas, CAGA uses the concept
of multiple clusters and provides fast convergence by
performing parallel operations. Additionally, the initial
degrees of nodes are kept unchanged. To determine
the topology robustness, Schneider Rmetric is used.
Besides, the high value of Schneider Rmeans that the
topology is robust. Extensive simulations are performed
to validate the effectiveness of the proposed models
by comparing with existing models, such as ROSE,
simulating annealing and hill climbing. From the results,
it is obvious that the proposed Enhanced ROSE
and EASO-ROSE outperform ROSE and simulating
annealing while CAGA and AGA perform better than
simulating annealing and hill-climbing in terms of high
Schneider R.
Acknowledgement
This work was supported by the Deanship of Scientific
Research at King Saud University under Grant RG-1437-
035.
References
S.E. Collier, “The Emerging Enernet: Convergence of
the Smart Grid with the Internet of Things”, Rural
Electric Power Conference (REPC), 2015 IEEE, 2015
Tsai, C. Wei, H. H. Cho, T. K. Shih, J. S. Pan, and J.
J. Rodrigues. “Metaheuristics for the Deployment of
5G.” IEEE Wireless Communications 22, no. 6 (2015):
40-46.
T. N. Qureshi, N. Javaid, A. H. Khan, A. Iqbal, E.
Akhtar and M. Ishfaq, “Balanced Energy Efficient
Network Integrated Super Heterogenous Protocol for
Wireless Sensor Networks”, International Workshop
on Body Area Sensor Networks (BASNet), 2013
W. Zafar, B.M. Khan, “Flying Ad-Hoc Networks:
Technological and Social Implications”, IEEE
Technology and Society Magazine, 2016.
Han, Guangjie, L. Liu, S. Chan, R. Yu, and Y. Yang.
“HySense: A hybrid mobile crowdsensing framework
for sensing opportunities compensation under
dynamic coverage constraint.” IEEE Communications
Magazine 55, no. 3 (2017): 93-99.
N. Javaid, A. Sher, H. Nasir, and N. Guizani,
“Intelligence in IoT based 5G Networks: Opportunities
and Challenges”, IEEE Communications Magazine,
2018.
Waluyo, Agustinus Borgy, David Taniar, Wenny
Rahayu, Ailixier Aikebaier, Makoto Takizawa, and
Bala Srinivasan. “Trustworthy-based efficient data
broadcast model for P2P interaction in resource-
constrained wireless environments.” Journal of
Computer and System Sciences 78, no. 6 (2012):
1716-1736.
T. Qiu et al., “Heterogeneous Ad Hoc Networks:
Architectures, Advances and Challenges,” Ad Hoc
Networks, vol. 55, 2, Feb. 2017, pp. 14352.
Nghiem, Thao P., Agustinus Borgy Waluyo, and David
Taniar. “A pure peer-to-peer approach for kNN query
processing in mobile ad hoc networks.” Personal and
Ubiquitous Computing 17, no. 5 (2013): 973-985.
J. A and Stankovic, “Research Directions for the Internet
of Things”, IEEE Internet of Things J., vol. 1, no. 1,
Feb. 2014, pp. 39.
I. Azam, N. Javaid, A. Ahmad, W. Abdul, A. Almogren
and A. Alamri, 2017. “Balanced load distribution with
energy hole avoidance in underwater WSNs.” IEEE
Access, 5, pp.15206-15221.
B. Hussain, Q.U. Hasan, N. Javaid, M. Guizani, A.
Almogren and A. Alamri, 2018. “An Innovative
Heuristic Algorithm for IoT-Enabled Smart Homes
for Developing Countries.” IEEE Access, 6, pp.15550-
15575.
T. Qiu, J. Liu, W. Si, M. Han, H. Ning, M.
Atiquzzaman, “A Data-Driven Robustness Algorithm
for the Internet of Things in Smart Cities”, IEEE
Communications Magazine, 2017.
Taniar, David. “Big data is all about data that we
don’t have.” In 2017 International Conference on
Advanced Computer Science and Information Systems
(ICACSIS), pp. 1-8. IEEE, 2017.
Taniar, David, Clement HC Leung, Wenny Rahayu, and
Sushant Goel. “ High-performance parallel database
processing and grid databases.” Vol. 67. John Wiley
& Sons, 2008.
Wu, Jun, M. Barahona, Y. J. Tan, and H. Z.
Deng. “Spectral Measure of Structural Robustness in
Complex Networks.” IEEE Transactions on Systems,
Man, and Cybernetics-Part A: Systems and Humans
41, no. 6 (2011): 1244-1252.
N. Javaid, M. Ejaz, W. Abdul, A. Alamri, A. Almogren,
I.A. Niaz and N. Guizani, 2017. “Cooperative position
aware mobility pattern of AUVs for avoiding void
zones in underwater WSNs.” Sensors, 17(3), p.580.
Attack Resistance based Topology Robustness of Scale free IoT for Smart Cities 17
N. Javaid, A. Sher, W. Abdul, I.A Niaz, A. Almogren
and A. Alamri, 2017. “Cooperative opportunistic
pressure based routing for underwater wireless sensor
networks.” Sensors, 17(3), p.629.
T. Qiu, A. Zhao, F. Xia, W. Si, D. Wu, “ROSE:
Robustness Strategy for Scale-free Wireless Sensor
Networks”, IEEE/ACM Transactions on Networking,
2017.
Waluyo, Agustinus Borgy, David Taniar, Wenny
Rahayu, and Bala Srinivasan. “Mobile service oriented
architectures for NN-queries.” Journal of Network and
Computer Applications 32, no. 2 (2009): 434-447.
J. Melthis, S. Tang, P. Yang, M. Hanneghan, C. Carter,
“Topologies for Combining the Internet of Things
and Serious Games”, Journal of Intelligent and Fuzzy
Systems, 2016.
I. Sohn, “Robustness Enhancement of Complex
Networks via No-Regret Learning”, ICT Express,
2018.
R. H. Li, J. X. Yu, X. Huang, H. Cheng, and Z. Shang,
“Measuring Robustness of Complex Networks under
MVC Attack”, in Proc. 21st ACM Int. Conf. Inf.
Knowl. Manage., New York, NY, USA, Oct./Nov.
2012, pp. 15121516.
M. Mozafari M. Khansari, “Improving the Robustness
of Scale-free Networks by Maintaining Community
Structure”, Journal of Complex Networks, 2019.
J. Liu, H. A. Abbass, K. C. Tan, “Evolving
Robust Networks Using Evolutionary Algorithms”,
Evolutionary Computation and Complex Networks,
Springer, 2019.
S. Wang, J. Liu, “Designing Comprehensively Robust
Networks Against Intentional Attacks and Cascading
Failures”, Information Sciences, 2019.
S. Wang, W. Lv, L. Zhao, S. Nie, H. E. Stanley,
“Structural and Functional Robustness of Networked
Critical Infrastructure Systems under Different Failure
Scenarios”, Physica A: Statistical Mechanics and its
Applications, 2019.
T. N. Qureshi, N. Javaid, S. A. Butt, W. Khan, S. A.
Changazi and M. M. Iqbal, “Enhanced Robustness
Strategy for IoT in Smart Cities Based on Data
Driven Approach”, 33nd International Conference on
Advanced Information Networking and Applications
AINA, 2019.
P. Buesser, F. Daolio, and M. Tomassini, “Optimizing
the Robustness of Scale-free Networks with Simulated
Annealing”, in Proc. 10th Int. Conf. Adapt.
Natural Comput. Algorithms ICAN NGA, Ljubljana,
Slovenia, Apr. 2011, pp. 167176.
A.-L. Barabsi and R. Albert, “Emergence of Scaling in
Random Networks.” Science, vol. 286, no. 5439, pp.
509512, 1999.
C. M. Schneider, A. A. Moreira, J. S. Andrade, Jr., S.
Havlin, and H. J. Herrmann, “Mitigation of Malicious
Attacks on Networks,” Proc. Nat. Acad. Sci. USA, vol.
108, no. 10, pp. 38383841, 2011.
Adhinugraha, Kiki, Wenny Rahayu, Takahiro Hara, and
David Taniar. “On Internet-of-Things (IoT) gateway
coverage expansion.” Future Generation Computer
Systems 107 (2020): 578-587.
H. J. Herrmann, C. M. Schneider, A. A. Moreira, J.
S. Andrade, Jr., and S. Havlin, “Onion-like Network
Topology Enhances Robustness Against Malicious
Attacks,” J. Statist. Mech., Theory Experim., vol.
2011, no. 1, pp. 19, 2011.
G. Zheng, S. Liu, and X. Qi, “Scale-free Topology
Evolution for Wireless Sensor Networks with
Reconstruction Mechanism”, Comput. Elect. Eng.,
vol. 38, no. 3, pp. 643651, 2012.
G. Zheng and Q. Liu, “Scale-free Topology Evolution for
Wireless Sensor Network”, Comput. Elect. Eng., vol.
39, no. 6, pp. 17791788, 2013.
Y. Jian, E. Liu, Y. Wang, Z. Zhang, and C. Lin,
“Scale-free Model for Wireless Sensor Networks”, in
Proc. IEEE Wireless Commun. Netw. Conf. (WCNC),
Shanghai, China, Apr. 2013, pp. 23292332.
L. Liu, X. Qi, J. Xue, and M. Xie, “A Topology
Construct and Control Model with Small-world
and Scale-free Concepts for Heterogeneous Sensor
Networks”, Int. J. Distrib. Sens. Netw., vol. 10, no. 3,
2014, Art. no. 374251.
V. H. P. Louzada, F. Daolio, H. J. Herrmann,
and M. Tomassini, “Smart Rewiring for Network
Robustness”, J. Complex Netw., vol. 1, no. 2, pp.
150159, 2013.
M. Zhou and J. Liu, “A Memetic Algorithm for
Enhancing the Robustness of Scale-free Networks
Against Malicious Attacks”, Phys. A, Statist. Mech.
Appl., vol. 410, pp. 131143, Sep. 2014.
W.B. Du, Z.-X. Wu, and K.-Q. Cai, “Effective Usage of
Shortest Paths Promotes Transportation Efficiency on
Scale-free Networks”, Phys. A, Statist. Mech. Appl.,
vol. 392, no. 17, pp. 35053512, 2013.
T. Qiu, D. Luo, F. Xia, N. Deonauth, W. Si, A. Tolba,
“Greedy Model with Small World for Improving the
Robustness of Heterogeneous Internet of Things”,
Journal of Computer Networks, Volume 101, 4 June
2016, Pages 127-143
18 Talha Naeem Qureshi et al.
W. P. Tay, J. N. Tsitsiklis, and M. Z.Win, “On
the Impact of node Failures and Unreliable
Communications in Dense Sensor Networks,” IEEE
Trans. Signal Process., vol. 56, no. 6, pp. 25352546,
Jun. 2008.
M. Hefeeda and M. Bagheri, “Randomized k-coverage
Algorithms for Dense Sensor Networks,” in Proc. 26th
IEEE Int. Conf. Comput. Commun. (INFOCOM),
May 2007, pp. 23762380.
W. Xiao, L. Lin, and G. Chen, “Vertex-Degree Sequences
in Complex Networks: New Characteristics and
Applications,” Physica A: Statistical Mechanics and
Its Applications, vol. 437, no. 11, Nov. 2015, pp. 43741.
Frank, Howard, and I. Frisch. “Analysis and Design
of Survivable Networks.” IEEE Transactions on
Communication Technology 18, no. 5 (1970): 501-519.
Latora, Vito, and M. Marchiori. “Efficient Behavior of
Small-world Networks.” Physical review letters 87, no.
19 (2001): 198701.
Sydney, Ali, C. Scoglio, M. Youssef, and P. Schumm.
“Characterizing the Robustness of Complex
Networks.” arXiv preprint arXiv:0811.3272 (2008).
Fiedler, Miroslav. “Algebraic Connectivity of Graphs.”
Czechoslovak mathematical journal 23, no. 2 (1973):
298-305.
Holme, Petter, B. J. K., C. N. Yoon, and S. K. Han.
“Attack Vulnerability of Complex Networks.” Physical
review E 65, no. 5 (2002): 056109.
M. Srinivas, L.M. Patnaik,“Adaptive Probabilities of
Crossover and Mutation in Genetic Algorithms”,
IEEE Transactions on Systems, Man and Cybernetics,
1994.
J. Zhang, H. S. Chung, ,“Clustering-Based Adaptive
Crossover and Mutation Probabilities for Genetic
Algorithms”, IEEE Transactions on Evolutionary
Computation, 2007.
S. Prendeville, E. Cherim, N. Bocken“Circular Cities:
Mapping Six Cities in Transition”, Environmental
Innovation and Societal Transitions, Volume 26,
March 2018, Pages 171-194.
... The network fragments into multiple subgraphs because of important nodes removal. The MCS is used to calculate the robustness [14] and a large number of nodes are required to make the network robust. The sensor nodes have limited energy resources, therefore, the SFNs robustness decreases due to the nodes' failure. ...
... Critical networks, including healthcare, military, and Internet, etc., have scale-free nature. These networks should be robust against attacks however, the existing algorithms [68,70,138] have high cost, therefore, self-optimization is used in [14]. Still, the problem of network robustness is not solved against malicious attacks and these attacks make the network vulnerable. ...
... If the obtain edges are independent then the degree of each edge is calculated (Lines 7-10). Moreover, for all the possible edges the difference of edge degree is calculated (Lines [11][12][13][14] and the minimum degree difference is selected (Line 15). The edges are swapped according to the minimum difference in edge degree and after each swap, the robustness is calculated (Lines [16][17][18][19][20][21][22]. ...
Thesis
Full-text available
During the past few decades, the Internet of Things (IoT) has made remarkable progress in many real-world applications including healthcare, military, transportation, etc. Multiple sensor nodes are deployed in these _elds to get the required data. Different network topologies are used in IoT and scale-free is one of them. It is mostly preferred due to its robust behavior against random node removal, however, the network collapsed because of malicious attacks. Therefore, in this thesis, robustness of the scale-free networks is enhanced against malicious attacks through optimization. To achieve this, the edge's degree and nodes' distance based edge swap operations are used in the proposed Improved Scale-Free Networks (ISFNs) scheme. In the edge's degree based operation, nodes of similar degrees are linked. Moreover, the connections of the nearest nodes are made in distance based edge swap. These operations help to achieve a better onion-like structure without changing the degree distribution of the network. Therefore, the network becomes robust against malicious attacks. Moreover, no new links or nodes are added in the optimization process, therefore, no extra cost is incurred. Furthermore, to make the network more robust against realistic attacks, the variable attacks are considered. Simulation results of the proposed scheme are compared with ROSE and Simulated Annealing (SA) for different number of nodes. The proposed scheme outperforms the existing techniques for different numbers of nodes and against the low degree, high degree and random attacks. Moreover, ISFNs has 13% and 23% better network robustness as compared to ROSE and SA, respectively. Network Topology Evolution Scheme (NTES) is proposed to prevent the scale-free networks from random and malicious attacks. In this scheme, the network field is divided into two parts with uniformly distributed nodes. After the network's evolution, the nodes are linked with each other through one-to-many correspondence. The division of the network field is made by considering that a network is robust if its size is small. Moreover, to study the hierarchical changes in the degree of nodes, k-core decomposition is used. In addition, nodes' degrees and core based attacks are performed on the network to evaluate the performance of the proposed scheme. Furthermore, the network robustness is analyzed using three optimization techniques: Artificial Bee Colony (ABC), Bacterial Foraging Optimization (BFO) and Genetic Algorithm (GA). The techniques are compared with each other and a technique that efficiently optimizes the network to increase the robustness is selected. In the optimization process, we make use of three edge swap methods. Due to the edge swap, the network robustness is enhanced without changing the degree distribution, so the addition of nodes/links is not required to increase the robustness. Furthermore, NTES is compared with Barabasi Albert (BA) model and Hill Climbing (HC) algorithm against random and malicious attacks. The simulation results show that the proposed NTES optimized using GA outperforms BA and HC by 46.90% and 57.08%, respectively, in terms of robustness. In addition, the network robustness of Scale Free Networks (SFNs) is enhanced against the malicious attacks. For that purpose, initially, a parameterless optimization algorithm JAYA is used because it requires less computational efforts as compared to the heuristic techniques. Then, as the edge swap plays an important role to enhance the robustness of SFNs, therefore, the edge swaps are classified into three categories. For each category, effects on the network's topological parameters such as average shortest path length, assortativity and clustering coefficient are analyzed. Next, the robustness is enhanced with the addition of nodes in the maximum connected subgraphs and the protection of bridge edges maintain the network connectivity. Moreover, optimized network is analyzed for different attack strengths. In simulations, the comparison of JAYA is made with two existing algorithms: ROSE and Simulated Annealing (SA). The network optimized by JAYA has a better robustness against random and malicious attacks, as compared to the existing algorithms. Furthermore, among the edge swap categories, the degree dependent edge swap is better to increase the robustness of SFNs. Moreover, the addition of nodes into the maximum connected subgraphs enhances the robustness and the protection of bridge edges ensures the network connectivity in all the algorithms. Furthermore, the robustness against different attack strengths are analyzed and the results show that high attacks strength paralyzed the network more efficiently.
... The network fragments into multiple subgraphs because of important nodes removal. The Maximum Connected Subgraph (MCS) is used to calculate the robustness [14] and a large number of nodes are required to make the network robust. The sensor nodes have limited energy resources, therefore, the Scale-Free Networks (SFNs) robustness decreases due to the nodes' failure. ...
... Critical networks, including healthcare, military, and Internet, etc., have scale-free nature. These networks should be robust against attacks however, the existing algorithms [68,70,138] have high cost, therefore, self-optimization is used in [14]. Still, the problem of network robustness is not solved against malicious attacks and these attacks make the network vulnerable. ...
Research Proposal
Full-text available
In this synopsis, robustness of the Scale-Free Networks (SFNs) is enhanced against malicious attacks through optimization. To achieve this, the edge’s degree and nodes’ distance based edge swap operations are used in the proposed Improved Scale-Free Networks (ISFNs) scheme. In the edge’s degree based operation, nodes of similar degrees are linked. Moreover, connections of the nearest nodes are made in distance based edge swap. These operations help to achieve a better onion-like structure without changing the degree distribution of the network. Therefore, the network becomes robust against malicious attacks. Furthermore, to make the network robust against realistic attacks, the variable attacks are considered. Apart from that, a Network Topology Evolution Scheme (NTES) is proposed to prevent SFNs from random and malicious attacks. In this scheme, the network field is divided into two parts with uniformly distributed nodes. After the network’s evolution, the nodes are linked with each other through one-to-many correspondence. The division of the network field is made by considering that a network is robust if its size is small. Moreover, to study the hierarchical changes in the degree of nodes, k-core decomposition is used. In addition, nodes’ degrees and core based attacks are performed on the network to evaluate the performance of the proposed scheme. Furthermore, the network robustness is analyzed using three optimization techniques: Artificial Bee Colony (ABC), Bacterial Foraging Optimization (BFO) and Genetic Algorithm (GA). The techniques are compared with each other and a technique that efficiently optimizes the network to increase the robustness is selected. In the optimization process, we make use of three edge swap methods. Due to the edge swap, the network robustness is enhanced without changing the degree distribution, so the addition of nodes/links is not required to increase the robustness. In addition, the network robustness of SFNs is enhanced against the malicious attacks. For that purpose, initially, a parameterless optimization algorithm JAYA is used because it requires less computational efforts as compared to the heuristic techniques. Then, as the edge swap plays an important role to enhance the robustness of SFNs, therefore, the edge swaps are classified into three categories. For each category, effects on the network’s topological parameters such as average shortest path length, assortativity and clustering coefficient are analyzed. Next, the robustness is enhanced with the addition of nodes in the maximum connected subgraphs and the protection of bridge edges maintain the network connectivity. Moreover, optimized network is analyzed for different attack strengths.
... However, IoT devices face many issues in their communication. Network security and data integrity are the major problems faced by IoT network because the number of connected devices is increasing day by day [3]. These networks also face the issue of a single point of failure [2] and are not secure because of a third party involvement. ...
Article
Full-text available
In this paper, Internet of Things (IoTs) devices are used for sensing the data through which the device owners earn revenue. Interested users can purchase data from IoT device owners, according to their demands. However, users are not confident about the quality of data they are purchasing. Moreover , the users do not rely on the device owner and are not willing to initiate data trading. Currently, data trading systems have many drawbacks, as they involve a third party, security and reputation mechanisms. Therefore, in this paper, IoTs and BlockChain (BC) are integrated to monetize IoT's data and provide trustful data trading. A BC based review system to monetize IoT's data trading is developed through Ethereum smart contracts. The review system encourages the owners to provide authentic data and solves the issues regarding data integrity, fake reviews and conflicts between entities. Reviews and ratings are stored in the BC database for providing a guarantee about the data quality to users. To maintain data integrity, we use an Advanced Encryption Standard (AES)-256 encryption technique to encrypt data. Moreover , an arbitrator entity is responsible to resolve conflicts between data owner and users. The incentive is provided to the users and arbitrators to increase user participation and honesty. Simulations are performed for the validation of our system. We examine the proposed model using three parameters: gas consumption, mining time and encryption time.
... It is also forecasted that by 2025, there will be 30 billion IoT connections [1]. Moreover, IoT has a wide range of applications in different domains like industrial IoT [2], smart cities [3], agriculture food chain [4], etc. The IoT networks usually operate in open-access environments, such as smart cities, food production and energy supply. ...
Article
Full-text available
Internet of Things (IoT) is an emerging domain in which different devices communicate with each other through minimum human intervention. IoT devices are usually operated in hostile and unattended environments. Moreover, routing in current IoT architecture becomes inefficient due to malicious and unauthenticated nodes' existence, minimum network lifetime, insecure routing, etc. This paper proposes a lightweight blockchain based authentication mechanism where ordinary sensors' credentials are stored. As IoT nodes have a short lifespan due to energy depletion, few credentials are stored in the blockchain to achieve lightweight authentication. Moreover, the route calculation is performed by a genetic algorithm enabled software defined network controller, which is also used for on-demand routing to optimize the energy consumption of the nodes in the IoT network. Furthermore, a route correctness mechanism is proposed to check the existence of malicious nodes in the calculated route. Moreover, a detection mechanism is proposed to restrict the malicious nodes' activities, while a malicious node's list is maintained in the blockchain, which is used in the route correctness mechanism. The proposed model is evaluated by performing intensive simulations. The effectiveness of the proposed model is depicted in terms of gas consumption, which shows the optimized utilization of resources. The residual energy of the network shows optimized route calculation, while the malicious node detection method shows the number of packets dropped.
... In random attacks, a node is removed randomly from the network. However, malicious attacks happen on the important nodes [15]. The importance of the nodes is usually measured based on their degree and the attack on them termed as High Degree Adaptive (HDA) attacks. ...
Conference Paper
Full-text available
Internet of Things (IoT) is rapidly increasing day by day due to its involvement in many applications such as electric grids, biological networks, transport networks, etc. In complex network theory, the model based on Scale Free Networks (SFNs) is more suitable for IoT. The SFNs are robust against random attacks; however, vulnerable to malicious attacks. Furthermore, as the size of a network increases, its robustness decreases. Therefore, in this paper, we propose a novel topology evolution approach to enhance the robustness of SFNs. Initially, we divide the network area into upper and lower parts. The nodes are deployed equally in both parts and connected via one-to-many correspondence. The distribution is made because small sized networks are more robust against malicious attacks. Moreover, we use k-core decomposition to calculate the hierarchical changes in the nodes' degree. In addition, the core-based and degree-based attacks are performed to analyze the robustness of SFNs. For the network optimization, we compare the Genetic Algorithm (GA) with Artificial Bee Colony (ABC) and Bacterial Foraging Algorithm (BFA). In the optimization process, the node's distance based edge swap is performed to draw long links in the network because these links make the network more robust.
Chapter
In this paper, we aim to increase the robustness of Scale-Free Networks (SFNs) that are robust against random attacks; however, they are fragile to malicious attacks. When the highest-degree nodes are removed from the SFNs, they greatly affect the connectivity of the remaining nodes. In the proposed method, we perform two attacks, namely, the high degree link attacks and the recalculated high degree link attacks. These link attacks affect the core part of the network. To measure the extent of damaging the network at a greater portion, we use the closeness centrality and the eigenvector centrality. The goal is to construct a topology that shows better robustness against different malicious attacks with a minimum computational cost. The Pearson’s correlation coefficient is used to select the measures that are positively correlated with each other. Based on these centrality measures, we optimize the network robustness for both node and link attacks simultaneously. The optimization is performed based on two efficient measures, which are computationally less expensive to increase the network robustness.
Article
Full-text available
Optimizing complex networks to be resilient against various attack models has been an important problem that is actively studied in the academia. In the proposed optimization method, individual node degrees are balanced iteratively based on the No-Regret learning algorithm, resulting in a robust network topology with increased resilience against outside attacks. Through simulation results, we show that the proposed robustness enhanced networks perform well under targeted attacks compared to the conventional optimized networks. © 2018 The Korean Institute of Communications and Information Sciences (KICS)
Article
Full-text available
The Internet of Things has been applied in many fields, especially in smart cities. The failure of nodes brings a significant challenge to the robustness of topologies. The IoT of smart cities is increasingly producing a vast amount different types of data, which includes the node's geographic information, neighbor list, sensing data, and so on. Thus, how to improve the robustness of topology against malicious attacks based on big data of smart cities becomes a critical issue. To tackle this problem, this article proposes an approach to improve the robustness of network topology based on a multi-population genetic algorithm (MPGA). First, the geographic information and neighbor list of nodes are extracted from a big data server. Then a novel MPGA with a crossover operator and a mutation operator is proposed to optimize the robustness of topology. Our algorithm keeps the initial degree of each node unchanged such that the optimized topology will not increase the energy cost of adding edges. The extensive experiment results show that our algorithm can significantly improve the robustness of topologies against malicious attacks.
Article
Full-text available
Due to the recent proliferation of cyber-attacks, improving the robustness of wireless sensor networks (WSNs), so that they can withstand node failures has become a critical issue. Scale-free WSNs are important, because they tolerate random attacks very well; however, they can be vulnerable to malicious attacks, which particularly target certain important nodes. To address this shortcoming, this paper first presents a new modeling strategy to generate scale-free network topologies, which considers the constraints in WSNs, such as the communication range and the threshold on the maximum node degree. Then, ROSE, a novel robustness enhancing algorithm for scale-free WSNs, is proposed. Given a scale-free topology, ROSE exploits the position and degree information of nodes to rearrange the edges to resemble an onion-like structure, which has been proven to be robust against malicious attacks. Meanwhile, ROSE keeps the degree of each node in the topology unchanged such that the resulting topology remains scale-free. The extensive experimental results verify that our new modeling strategy indeed generates scale-free network topologies for WSNs, and ROSE can significantly improve the robustness of the network topologies generated by our modeling strategy. Moreover, we compare ROSE with two existing robustness enhancing algorithms, showing that ROSE outperforms both.
Article
The Internet of Things (IoT) has developed rapidly in recent years where significant numbers of devices have been connected to the network and this will increase over the ensuing years. There is also a trend to shift the IoT topology from cloud computing to fog computing where computing logic is brought as nearest as possible to the sensors. One of the topologies that is commonly used in fog computing is the mesh network. In a mesh network, end point nodes/sensors are connected to other nodes with routing capabilities called routers, and these nodes are connected to gateways where the mesh can communicate with other meshes or clouds. In multi-gateway mesh networks, each gateway may have set of routers that is fully depended on the gateway to forward the data to the clouds. Mesh is a complex network structure and the overall performance of the networks can be affected by several issues, such as overload gateways, network latency and gateway failover. In this paper, we compare the ordinary Network Voronoi Diagram (NVD) with Hops Voronoi diagram (HVD) to distribute the gateway workload based on network hops; and extend these methods to identify the overlapped routers for gateway failover.
Article
Scale-free networks, which play an important role in modelling human activities, are always suffering from intentional attacks which have serious consequences on their functionality. Degree distribution and community structure are two distinguishing characteristics of these networks considered in optimizing network robustness process recently. Since community structure is known as functional modules in some networks, modifying them during the improving network robustness process may affect network performance. We propose a preferential rewiring method to improve network robustness which not only keeps degree distribution unchanged but also preserves community structure and decreases the number of rewired edges at the same time. At first, the robustness of each community is improved by applying a smart rewiring method based on the neighbourhood of nodes. Then, relations between communities are gotten more robust with a preferential rewiring based on degree and betweenness hybrid centrality of nodes. This method was applied to several types of networks including Dolphins, WU-PowerGrid and US-Airline as real-world networks and Lancichinetti–Fortunato–Radicchi benchmark model as an artificial network with the scale-free property. The results show that the proposed method enhances the robustness of all networks against degree centrality attacks between 50% and 80% and betweenness centrality attacks between 30% and 70%. Whereas, in all cases, community structure preserved more than 50%. In comparison with previous studies, the proposed method can improve network robustness more significantly and decrease the number of rewires. It also is not dependent on the attack strategy.
Article
We analyze the structural and functional robustness of networked critical infrastructure systems (CISs). We propose a structural and functional robustness model of a typical complex network and take into account the corresponding measuring metrics and cascading processes to assess the impact of different hazard modes on robustness. We analyze the robustness of the shanghai subway network and the central China power grid to demonstrate an application of the model. We find that both are strongly robust to random failure but extremely vulnerable to targeted attack. We identify the critical areas most likely to be attack targets and use a functional perspective to identify vulnerabilities. Our proposed method can be applied to other CISs and aid in understanding the mechanisms of network robustness.
Article
The requirement of high data rates, low latency, efficient use of spectrum and coexistence of different network technologies are major considerations in Internet of Things (IoT) based Fifth Generation (5G) networks. To achieve the above requirements, the incorporation of Artificial Intelligence (AI) is required to make efficient decisions based on the massive data generated by the large number of IoT devices. AI methods analyse the data to extract the patterns and make sense of the data to prescribe action to the end devices. In this work, we first give an overview, discussing the challenges and relevant solutions of the 5G and IoT technologies including the IoT based 5G enabling technologies. Second, we discuss the need for AI in future IoT based 5G networks in the perspective of Kipling's method. In addition, we review the intelligent use of spectrum through full duplex and cognitive radio technologies.
Article
Over the past few years, active research on algorithm development for the optimal operations of HEMSs has been performed. The objective is to compute optimized schedules for shiftable home appliances (SHAs). This is based on the demand response (DR) synergized with renewable energy sources (RESs) and ESS optimal dispatch (DRSREOD). An improved algorithm for a DRSREOD-based HEMS is proposed in this paper. This heuristic-based algorithm considers DR, photovoltaic (PV) availability, the state of charge and charge/discharge rates of the storage battery (SB) and the sharing-based parallel operation of more than one power source to supply the required load. The HEMS problem has been solved to minimize the cost of energy (CE) and time-based discomfort (TBD) with conflicting tradeoffs. The mixed scheduling (MS) of appliances (delayed scheduling (DS) for some appliances and advanced scheduling (AS) for others) is introduced to improve the CE and TBD performance parameters. An inclining block rate (IBR) scheme is also incorporated to reduce the peak load. A set of optimized tradeoffs between CE and TBD has been computed to address multi-objectivity using a multi-objective GA (MOGA) with Pareto optimization (PO) to perform the tradeoff analysis and to enable consumers to select the most feasible solution. Due to the rapid increase in demand for electricity, developing countries are facing large-scale load shedding (LS). An innovative algorithm is also proposed for the optimal sizing of a dispatchable generator (DG) that can supply a DRSREOD-based HEMS during LS hours to ensure an uninterrupted supply of power. The proposed MOGA/PO-based algorithm enables consumers to select a DG of the optimal size from among a number of optimal choices based on tradeoffs between the DG size, CE and TBD.
Article
Urbanisation and climate change are urging cities to chart novel paths towards sustainable futures. Many cities are turning to the alluring ‘circular economy’ (CE) concept to guide this redirection. The CE concept re-imagines how flows of resources moving through economies might be ‘closed’. Here, we explore this new ‘circular city’ agenda by asking: How are cities adopting CE as a strategy?. We found that political leadership, building adaptable future visions, using experimental approaches (such as living labs), developing contextual knowledge about resource use, and engaging with diverse stakeholders to be important. However, we also expose that there is a lack of consensus on what a circular city constitutes and a need to further untangle the how and why of the circular city concept. The research contributes to the field by outlining emergent cases, identifying a set of common policy strategies, conceptualising a circular city and identifying areas for future research.