Content uploaded by Ahona Ghosh
Author content
All content in this area was uploaded by Ahona Ghosh on Aug 26, 2020
Content may be subject to copyright.
23
Copyright © 2020, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Chapter 2
DOI: 10.4018/978-1-7998-5068-7.ch002
ABSTRACT
Wireless sensor networks consist of unattended small sensor nodes having low energy and low range of
communication. It has been observed that if there is any system to periodically start and stop the sensors
sensing activities, then it saves some energy, and thus, the network lifetime gets extended. According to
the current literature, security and energy efficiency are the two main concerns to improve the quality
of service during transmission of data in wireless sensor networks. Machine learning has proved its ef-
ficiency in developing efficient processes to handle complex problems in various network aspects. Routing
in wireless sensor network is the process of finding the route for transmitting data among different sen-
sor nodes according to the requirement. Machine learning has been used in a broad way for designing
energy efficient routing protocols, and this chapter reviews the existing works in the said domain, which
can be the guide to someone who wants to explore the area further.
Secured Energy-Efcient
Routing in Wireless Sensor
Networks Using Machine
Learning Algorithm:
Fundamentals and Applications
Ahona Ghosh
Brainware University, Kolkata, India
Chiung Ching Ho
Department of Computing and Information Systems, School of Science and Technology, Sunway
University, Malaysia
Robert Bestak
Department of Telecommunication Engineering, Czech Technical University in Prague, Czech
Republic
24
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
INTRODUCTION
The rise of artificial intelligence (AI) has influenced every field largely; from entertainment to education
or from agriculture to manufacturing. Healthcare and computer network are not outside the list which
have witnessed a great impact with the magical touch of AI and Machine Learning (ML). In computer
network, Wireless Sensor Network (WSN) denotes a collection of spatially isolated and dedicated sensors
for monitoring and recording the physical situations of the circumstances and shaping the collected data at
a central position. Apart from the different factors like time consumption, energy efficiency, security and
cost which contribute to the process of routing in WSN, the active research areas include different Quality
of service parameters like Packet Delivery Ratio (PDR), efficiency, robustness, reliability etc. The process
of data transmission between different sensor nodes and communication between them is called routing
and the goal always remains to reduce the energy consumption during this routing (Pathan et al., 2007) and
increase the lifetime of the sensor nodes as much as possible. In this chapter, the design issues of WSN have
been addressed first and then the applications of machine learning are described in the concerned domain.
In the next section, related background study and currently available applications of machine learn-
ing algorithms in the concerned domain are highlighted. In section 3, the advantages and drawbacks for
the existing ML approaches in Energy efficient routing scenario are discussed. Section 4 describes the
recent algorithms or techniques used to develop energy efficient routing in WSN and compares them
after performance evaluation. Segment 5 provides an overall discussion about the scope and limitation
of the chapter and future direction.
Motivation of the Chapter
This chapter collects views of different researchers worldwide from different perspectives. The survey
outline presented here is definitely going to help and guide the present researchers in the concerned
domain. WSNs can be used to track and monitor the dangerous and unreachable areas where exploration
of locations having irregular behaviours like volcanic eruption, forest fire etc. The initial configurations
should have the capability of changing its nature to adopt with circumstances, because anytime anything
can happen. Machine learning algorithms are capable of calibrating itself to newly acquired knowledge,
so application of machine learning in these types of systems will be really useful. The sensor devices
are often capable of collecting large data, but sometimes they cannot find the correlation between them.
Machine learning can be applied to them for exploring the correlation for better deployment and wide
area of coverage which is always desired for the systems. This chapter summarizes the existing systems
where limited resource and diversity in the learning patterns have been considered. However, areas like
development of distributed and lightweight message transmission system and using machine learning in
resource scheduling and management have still remained unexplored. Further experiments and researches
can be undertaken in the domain mentioned.
Background of WSN
Wireless Sensor Networks control and monitor rapidly changing environments efficiently. This dynamic
nature is sometimes due to some external factors affecting the system and sometimes it is due to some
requirement from designer’s perspective. To cope up with such ever changing scenarios, machine learn-
ing is applied to WSN implementation so that the network learns the nature and trend by its own.
25
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
Contribution of the Chapter
In this chapter, we have presented a review of the machine learning methods applied in various literatures
over the period 2004-2020 related to WSN, especially in security and energy efficient routing methods
of WSN. Various problems of the existing works have been addressed along with experimental finding
comparison and advantages of the methods. We have also provided a guidance to current and future
WSN designers for their concerned application challenges aiming to maximize resource utilization and
lifespan of sensor nodes.
Literature Survey
Exploration of machine learning in different WSN protocols and applications have become the reason
behind increasing attention of researchers in recent years. The set of rules applied to design a routing tech-
nique is called routing protocol and, in this context, different protocols (Osisanwo et al.,2017) have been
taken into account in different scenarios. Low Energy Adaptive Clustering Hierarchy (LEACH) (Dhawan
et al.,2014) is a traditional protocol which has taken energy efficiency as one of its considerable factors,
most of the machine learning based routing techniques have been compared with LEACH to evaluate
their performance in the concerned area of research. Pathan et al. have proposed an approach in (Pathan
et al.,2007) to design an energy efficient routing protocol where two security aspects, i.e. authenticity
and confidentiality have been considered by adopting one-way hash chain and preloaded shared secret
keys. Experimental results show that the method is promising one in its concerned domain, but here the
concept of optimization of interval value for network restructuring purpose is missing. Artificial neural
network, swarm intelligence and artificial immune system have been used to improve the existing rout-
ing processes of WSN in (Kumar et al.,2014) by making the network ‘cognitive’ in nature. The attacks
in wsn to capture nodes have been tried to reduce by introducing game theory to model the movements
of victim node and attacker node and the nature of the attacker has been learnt through neural network
approach. After reviewing the existing works, Bhanderi et al. in (Bhanderi et al., 2014) have concluded
that machine learning is an appropriate solution for the optimization problems in WSN and truthfully
utilize the complex characteristics of distributed systems. Two stage machine learning approach has been
designed by Vimalapriya et al. (Vimalapriya et al., 2019) to verify node adaptability depending on the
energy, delay and the capacity of accepting packets. The nodes with deficiency are identified and the
routes are refurbished after pruning with lesser delay and latency. The autonomy of sensor nodes results
to threats and malicious node injection in the network leads to performance degradation (Ishmanoy et
al., 2017), path loss etc. Contrary, Security, end to end latency, alive node, average energy utilization
and performance analysis show that the model can retain energy in an appreciable level and the active
nodes are capable of better survival than the traditional methods.
Protocol function-based routing is of five types according to Sharawi et al. (Sharawi et al.,2013) which
include negotiation-based routing, query-based routing, QoS based routing, coherent based routing.
Among different paradigms of soft computing, reinforcement learning has been the most appropriate as
its memory consumption is the lowest, flexibility is the highest and processing power consumption is the
lowest compared to the other methods, i.e. Swarm Intelligence, Evolutionary Algorithm, Fuzzy logic,
Neural network and artificial immune system. The hybrid computing paradigms which are inspired by
biology are also promising to solve optimization problems in WSN.
26
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
Alsheikh et al. have presented a literature review where the new application areas of WSN have been
explored, like cyber physical system, Internet of Things, machine to machine communication (Alsheikh
et al., 2017). With restricted human intervention, the machine learning algorithms extract the different
abstraction levels. Specialized machine learning based WSN systems like development of outlier detec-
tion system, task scheduling and optimal deployment systems have been also reviewed here. Apart from
different security aspects like anomaly and intrusion detection, Quality of Service enhancement and data
integrity have been considered as non-operational characteristics in (Khan et al., 2017) as because of
random network topologies, faulty and unreliable data in WSNs. The challenges in WSNs are most of
the time application specific like selection of sensors, management of resource where future trends and
estimations are required. Deep learning strategies have been very effective in existing unsophisticated
networking systems where intelligent and autonomous characteristics of network are required. Scattered
applications of deep learning in WSN have been summarized in (Fadlullah et al., 2017) where area of
prediction and classification of network flow and traffic, mobility prediction have been explored for
the newly emerging technology called deep learning. Open research challenges like smart network have
been discussed and one researcher can be guided well through this.
Machine Learning Methods Applied in WSN Routing
With the rapid increase in applications of machine learning, the area of computer network has also been
flourished with different formation techniques having flexible and localized communication mediums.
Machine learning methods can be broadly categorized into three types, i.e. supervised, unsupervised and
reinforcement learning. All three of these have been applied to design secured energy efficient routing
for WSN in existing literatures and comparison among the methods has been shown in Table 1. The most
popular methods include Support vector machine (SVM), K-Means Clustering, Decision tree (Zhong et
al.,2016), Ant colony optimization (Dorigo et al.,2006), Fuzzy logic (Zadeh et al.,1999), Evolutionary
algorithm (Vikhar et al.,2016), swarm intelligence (Kennedy et al.,2006), Artificial Neural Network
(ANN), Convolutional Neural Network (CNN) of Deep Learning as shown in Figure 1. In (Egrova et
al., 2006), the main contribution of Forster et al is in considering multiple destinations for the individual
sensor nodes instead of targeting only one destination for multiple sources in WSN. In each iteration the
route fitness feedback is provided and learning better routes is accomplished accordingly. Estimation of
link costs for routing optimization has been performed in (Singh et al.,2017). Particle swarm optimization-
based routing protocol by the authors has shown better performance in terms of consumed energy, end to
end delay and data transmitted than the traditional LEACH protocols designed for the purpose of routing
in WSN. The main advantage of LEACH as no need of maintaining routing information and forwarding
the collected data from every sensor node to its neighbour nodes and further forwarding of data to base
station node to its neighbour nodes using multi hop fashion has been discussed in (Ali et al.,2019). Data
redundancy, Route updating cost and end to end delay can be caused due to blackholes in network for
quick draining of energy of sensor nodes. To overcome these issues a cyclic neural network-based ap-
proach of data fusion has been proposed by the authors here. Fission fusion social structure-based spider
monkey optimization and Termite colony optimization-based clustering in WSN have been proposed in
(Gui et al.,2016) and several meta heuristics-based research efforts have been reviewed here. Improv-
ing SMO algorithms by reducing space and time complexity and development of SMO based routing
optimization technique can be done in future for better understanding of the concept.
27
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
Fuzzy logic is used to deal with uncertainty of some relative terms Arabi et al. have used fuzzy logic
in (Arabi et al.,2010) to select the cluster head based on some predefined condition and to shift between
two techniques, namely Earliest Fast tree and Source initiated dissemination for designing a hybrid
routing method having flexibility and network lifetime increase as its features. Transmission capability
of nodes, link quality within a range and battery capacity have been the three main considered factors
in the work of (Jaradat et al.,2013) to find out the next hop of relay node towards the target node where
the routing decision of next hop has been made by fuzzy logic controller.
Initial concept of artificial neural network comes from the working pattern of biological neurons. Some
nodes in the form of neurons are connected to each other and some function according to the application
area is invoked for assigning weights to the connections. The objective remains to adjust the weights
to get the output exactly as the mapping function. This technique has been used in different functional
aspects of Wireless sensor network in the recent literatures. Authors (Nehra et al.,2009) have presented
the problem of energy efficient routing as linear programming model with some specified restrictions.
Experimental results show that the number of active nodes and the residual energy are better than the
LEACH protocol in WSN. Authors (Zhao et al.,2009) have used self-organizing neural network on a
sensor node called MODABER to solve the problem of frequency interference. Authors (Bin et al.,2006)
have discussed a system where dynamic reduction of information to be forwarded has taken place to
reduce the energy consumption. Trunk of a genetic tree supported by the network structure gets updated
time to time for analysing the trade-off for the full system’s energy consumption and simulation results
show that the proposed method is promising one.
Natural evolution of adaptation to ever changing environment and existence capability is modelled
using one of the widely used machine learning approaches called evolutionary method. For solution
of routing in two tier sensor networks by the authors (Chakraborty et al.,2012) have proposed a dif-
Figure 1. Categories of Machine learning algorithms used in WSN
28
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
ferential evolution based memetic algorithm. Local search algorithm has been modified by combining
it with differential evolution to find the optimized path from every node to base station by achieving
energy efficiency and increased network lifetime. A multi objective differential evolution approach
has been introduced by the authors (Sharma et al.,2012) where latency for a linkage in communication
has been a factor in single and multipath routing problem. Enhancement of data quality has been the
major concern of data fusion problem addressed by the authors (Pinto et al.,2014) where different user
defined parameters have designed a trade-off in between using the genetic machine learning algorithm
of proposed approach. The ability to automatically adjust the communication rate of specific sensor
nodes according to the dynamic topology proves the novelty and significance of the work. The proposed
model’s application to optimization of other parameters like evolution and check point interval has not
been tested and thus missing here.
The uncontrolled systems are distributed by the concept of swarm intelligence (Okdem et al.,2009,
Saleem et al.,2019, Zungeru et al.,2012) based on some effective routing formula of shortest path in
ant’s colony (Sarangi et al.,2012). Basic Ant Based Routing (BABR), Flooded Forward Ant Routing
(FFAR), Sensor Driven and Cost Aware Ant Routing, Flooded Piggybacked Ant Routing are some of
the well-known techniques in this domain. The common concept behind all of these techniques is trans-
mission from one node to its neighbour nodes depends on the energy function of the neighbour nodes
and the present pheromone in the connection between them. When the data gets delivered to the proper
target node, a backward path tracking takes place for updating the node number and energy consumed.
A new idea of introducing dominant node has been expressed in (Venkataramana et al.,2019) where the
proposed approach has been compared with LEACH, LEACH-Two Level Group Head, Energy Efficient
Dynamic Clustering Algorithm (Sarella et al.,2017) and Energy-Efficient Level based and Time-based
Clustering algorithm (EELTC) (Tashtarian et al.,2007) and PEGASIS (Kim et al.,2006) for performance
evaluation and shown a better performance than the conventional ones.
Forster has reviewed presently used ML techniques in WSN and Mobile Ad Hoc Network (MANET)
(Forster et al.,2007). ML approaches have definitely performed better than the non-learning techniques
in WSN. Reinforcement learning has been the most effective one in mostly static WSN whereas swarm
intelligence has worked well in non-energy restricted, highly mobile MANET environment. Ant colony
optimization as a heuristic way for energy consumption reduction purpose has been proposed by (Yan
et al.,2011). The process of optimization here is divided into several stages, among which, the first one
is Ant System (AS), the second one is Elitist technique for the Ant System (EAS) and the later ones are
Ant System Rank, Max-Min Ant System, Ant Colony System (ACS). The simulation results shown in
six different scales, compare the performances of three algorithms, namely ACS, AS and an enhanced
AS called ASW and show that ACS consumes less energy than the other two.
Authors (Prajapati et al.,2018) have explored a new aspect of WSN where according to them, the
quantity of the sensor nodes should not affect Quality of Service of WSN as the traffic gets unbalanced
and packet transmission rate falls due to continuous change in the number of nodes for addition and
deletion of nodes. Artificial neural network methods like multilayer perceptron (MLP) have been very
useful in distributed and parallel network environment, whereas Naïve Bayes classifier has shown little
bit poor performance compared to SVM and MLP. Explicit programming of conventional designs of
WSN are not capable of handling its dynamic nature. Synchronization, congestion control and avoidance
have been discussed in (Kumar et al.,2019) apart from the common discussed factors of dealing with
WSN challenges using machine learning. Malicious nodes named as Hello flooding, selective forwarding
have been classified and modelled using Bayesian classifier and a malicious node detection system has
29
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
Table 1. Comparison of the existing literatures based on ML methods
Ref. Objective Methodology ML method Performance Drawback(s)
Supervised
(Khan et al.,
2016)
-To assign sensor nodes
to the nearest cluster
and balance the energy
dissipation in the
cluster heads
-Designing a
WSN routing
protocol for
clustering sensor
nodes
SVM
-Performance is better
than the LEACH by
comparing the PDR,
processing overhead,
energy dissipation &
memory overhead
--
(Wang et al.,
2006)
-To predict link quality
of wsn for optimization
of routing
-Feature
extraction & o/p
labelling
-Sample collection
-Offline training
Decision Tree
& Rule learner
-300% improved data
delivery rate evaluated
by MetricMap &
MintRoute
-Less complexity
-Reduction of
labelling cost not
considered
-No online
incremental training
approach considered
Unsupervised
(Preeth et al.,
2018)
-To design energy
efficient cluster &
immune inspired
routing protocol
-Cluster head
selection based
on QoS impact &
energy status
Adaptive
Fuzzy multi
criteria
decision
making
-OoS parameters, i.e.
channel load, PDR
99%, Jitter BER,
throughput 0.95Mbps,
network lifetime 5500
rounds
-No optimal route
with high residual
energy for increasing
the sensor lifetime is
searched
(Thangaramya
et al., 2019)
-Cluster formation
in WSN for energy
efficient routing of
packets
-Weight
adjustment using
CNN to prolong
network lifetime
-Neuro fuzzy
rule-based
clustering
-Better network lifetime
than the LEACH,
FLCFP &HEED
protocol
-Assumption of
every node to be
trustful
(Townsend,
2018)
-Node clustering
in near optimal
configuration for
energy efficiency
Two chromosome
repair methods
Arepair and
BFSrepair to
design gateway
nodes
-Genetic
algorithm
-4 selection
methods,
i.e. elite,
roulette wheel,
linear rank,
tournament
-Less time & energy
consumption than
traditional genetic
algorithm
-Consideration of
only single sink
-No variation
considered in
transmission range
-Consideration of
only static nodes in
WSN
Reinforcement
(Förster et
al., 2008)
-Routing on Real WSN
peripherals using
QLearning method of
Reinforcement learning
FROMS (multi-
source multicast
routing protocol)
& DD (Directed
Diffusion in
a test case of
ScatterWeb nodes
QLearning
-Improvement of
delivery rate in
different network
scenarios
-Small size of
testbed
-No consideration of
changing topologies,
new cost functions
and mobile sinks
(Kadam et
al., 2012)
-Energy efficient
routing in heterogenous
network
-Propagation of
energy beyond the
direct neighbours
Enhanced
version of
Qlearning
method in [57]
Improvements in n/w
lifetime after load
balancing also
--
(Liang et al.,
2008)
-Computation of
QoS route using
distributed value
function-distributed
reinforcement learning
method
-Identification of
optimal routing
policy through
previous rewards
& experience
Qlearning
algorithm
-Packet routing more
efficient than AODV
routing
-No energy
consumption
parameter is
considered for QoS
route determination
(Boyan et al.,
1994)
-To track which routing
decision leads to
minimum delivery time
-Shortest path
measurement by
Bellman Ford
algorithm
Qrouting
algorithm
Qrouting based
algorithms perform
better in the packet
routing domain
-Only table-based
representation -no
concept of function
approximator
(Dong et al.,
2007)
-To design a geographic
routing protocol for
providing suitable
localization and high
data rate at low cost in
WSN
-capable of
adapting energy
variation
Reinforcement
learning
-Robustness & network
lifetime is 75% to 213%
better than GPSR [67]
--
30
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
been proposed and simulated for performance evaluation in (Zou et al.,2012). The error rate of 0.6% in
classification proves the effectiveness and appropriateness of the method. Network health abnormality
prediction based on regression algorithm and network object detection and monitoring as well as state
transition analysis using pattern matching machine learning approach have been described in (Vijayaku-
mar et al,2019) where the need of revolutionary technology like Internet of Things in sensor and cloud
environment have been presented also apart from the WSN. Three stages, namely pre-processing, data
aggregation and inference of information processing in WSN have been highlighted in (Di et al.,2007)
where inference refers to the process of extracting hidden information from the aggregated data. Most
of the machine learning applications nowadays focus on classification of moving object in WSN and
can be integrated throughout these three steps for optimal information processing.
Supervised Learning Methods
Supervised learning methods (Asusanwo et al., 2017) are the tasks of learning where the inputs are
mapped to some outputs based on some predefined sample input output pairs. Analysis of training sample
is required to get the learning function and to trace the output of the testing sample. Supervised meth-
ods, like classification and regression have been applied in several aspects of Wireless Sensor Network
(Khan et al.,2016, Wang et al.,2006) where classification include SVM, artificial neural network, deep
learning, Random forest, Decision tree, Bayesian network and K-nearest neighbour as well. Decision tree
has been applied in (Rajagopalan et al.,2006) to choose the cluster head and important features like the
distance between decision tree node and the cluster centroid, the degree of movement, battery level and
identification of vulnerability have been considered during the iteration of input vector through decision
tree which results to better performance of the model than traditional LEACH protocol.
Janakiram et al. have proposed an outlier detecting scheme (Janakiram et al.,2006) using Bayesian
network where the first assumption is most of the neighbours of a particular node always have the similar
readings with the said node. Based on this phenomenon, the correlations or the dependencies among
several nodes are calculated and the outliers are identified accordingly. K-Nearest Neighbour has been
used in (Branch et al.,2013) where if there is any lost reading of some node, then it gets replaced by
the average value of readings provided by k number of nearest neighbours to that specific node, but the
limitation here is the large memory requirement for storing and monitoring each and every reading for
all the nodes of a network.
Support Vector Machine
Support vector machine is a widely popular binary classifier where data points belonging to two differ-
ent classes get separated by a line called, hyperplane. The working mechanism of SVM has been shown
in Figure 2.
Decision Tree
The classification and regression method works here in the form of a tree structure where the dataset is
broken down into sub datasets at first and at the same time an associated decision tree gets built. The
decision nodes contain two or more branches and finally the leaf nodes contain the decision or clas-
sification. The mechanism is shown in Figure 3.
31
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
Random Forest
The random forest is basically a collection of decision trees where ensemble method of classification
and regression is integrated. During the training time, it inputs some classes in the form of decision trees
and the output class becomes the mode of the classes in case of classification and mean prediction in
case of regression. The functionality of Random Forest algorithm is shown in Figure 4.
Naïve Bayes Classifier
Naïve bayes classifiers are simple families of probabilistic classifiers in machine learning which applies
Bayes theorem having strong (naïve) assumptions among the considered features. Its accuracy and speed
are good when it is applied on large datasets and the mechanism is shown in Figure 5.
Figure 2. Working mechanism of SVM
Figure 3. Working mechanism of Decision tree
32
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
Unsupervised Learning Methods
Unsupervised learning approaches namely clustering, dimensionality reduction have been widely used
(Preeth et al.,2018, Thangaramya et al.,2019, Townsend et al.,2018, Fakhet et al.,2017) in Wireless
Sensor Network. Clustering is a technique of data analysis where an intuition about the structure of a
collection of data points is analysed and groups/clusters of similar data points based on some application
specific conditions like Euclidean distance or correlation-based distance etc are formed. It is inefficient
to transmit each and every data directly to sink in case of a large energy restricted network, if data from
every node gets transmitted to its local aggregator (also known as cluster head) and the aggregator after
Figure 4. Working mechanism of Random forest
Figure 5. Working mechanism of Naïve Bayes Classifier
33
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
aggregating all the data received from its cluster nodes, passes it to the sink finally, here lies the concept
of clustering the similar nodes in WSN. The popular clustering algorithms which have been applied in
WSN are k-means clustering, hierarchical clustering and Fuzzy C-means clustering, whereas dimen-
sionality reduction algorithms like principal component analysis, independent component analysis and
singular value decomposition have been widely used in various aspects of WSN. Principal Component
Analysis (PCA) is a technique to extract the most important features from some large feature set and
present them as a set of orthogonal variables called principal components. In WSN, the PCA reduces
the data amount among sensor nodes by eliminating the un correlated data from the nodes and keeps
only the important portion of data to reduce the overall complexity of the system. Two data aggregation
techniques (Li et al.,2004, Masiero et al.,2009, Mekua et al.,2010, Rajagopalan et al.,2006), namely
Compressive Sensing (CS) (Duarte et al.,2011) and Expectation Maximization (EM) (Dempster et
al.,1977) are combinedly used with Principal component analysis to achieve a better performance with
enhanced data aggregation in WSN.
K-Means Clustering
K means clustering is the most well-known unsupervised learning method which has been applied in
several literatures for determining the wireless sensor nodes cluster and in an efficient way to make the
system effective and reliable to the network user. The working mechanism of K-Means clustering has
been shown in Figure 6. The steps followed in K-means clustering are:
Step 1: Choose k number of nodes randomly to be the initial centroids of a cluster
Step 2: Mark each node using the nearest centroid measured using distance function
Step 3: Re-evaluate the centroid using current node’s membership
Step 4: Stop when the predefined convergence condition gets satisfied
The addition of the distances between every node and its corresponding centroid is compared with
a pre-defined threshold value every time and if the condition is satisfied then the problem converges,
otherwise the flow goes back to step 2. A cluster is a collection of similar data points aggregated together.
Typically unsupervised methods create inferences from the concerned datasets using only input vectors
without having any known or labelled output.
Figure 6. Working mechanism of K means clustering
34
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
Hierarchical Clustering
In this method (also called hierarchical cluster analysis) similar features are grouped with each other based
on some pre-defined condition and the output given is a set of distinct clusters where the objects within
each cluster are similar to each other. The working mechanism (Rani et al.,2013) is shown in Figure 7.
Fuzzy Clustering
Fuzzy clustering is also known as soft clustering and soft k means (Yang et al.,1993). The speciality
of this method is here one data point can belong to more than one clusters. The points in a class are
similar to each other and in different classes, the points are dissimilar to among them. The data points
are assigned some values between 0 and 1 according to their degree of membership which avoids the
distortions in the final solution of regular clustering algorithms. The working mechanism of Fuzzy
clustering is shown in Figure 8.
Principal Component Analysis (PCA)
It is a linear transformation process for reduction of dimensionality for a large dataset having datapoints
at two, three or more than that dimensional space. A line which is having the least average squared dis-
tance from a point is said to be best fitted and the next best fitted line can be defined as the line which
Figure 7. Working mechanism of Hierarchical clustering algorithm
35
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
is having perpendicular directions from the first. By repetition of this process, some orthogonal basis
gets formed where the data are not correlated, these vectors are called as principal components and the
process is called principal component analysis. The working mechanism of PCA is shown in Figure 9.
Reinforcement Learning Methods
Reinforcement learning methods have strong contributions (Forster et al.,2010, Kadam et al.,2012,
Boyan et al.,1994, Dong et al.,2007) in the domain of computer network and especially in wireless sen-
sor network. The working mechanism of Reinforcement learning as shown in Figure 10 describes the
process of learning what to do next and mapping process of different scenarios to actions for maximizing
the reward signal for acting perfectly upon corresponding environment. The decision of which action
to take is not available to the learner at the very beginning, by time, he/she discovers the actions which
will result to the most rewarding statements and these actions affect not only the next step, but also all
the subsequent stages of rewards. Existing approaches to Q-learning method of WSN sometimes face
difficulties in convergence, to overcome this problem, Sharma et al. have (Sharma et al.,2012) proposed
Figure 8. Working mechanism of Fuzzy clustering
Figure 9. Working mechanism of PCA
36
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
an enhanced model of Q-learning method, but Network Simulator version 2 provided the simulation
result of enhanced model as same as the traditional model. Authors (Dorigo et al.,2006) have presented
an approach to manage power dynamically by using a learning algorithm which does not have any prior
model, application of neural network having multi-layer provides the workload information and helps
to take improved decision depending on the varying workload. Based on the estimated workload, the
learning method is formed to use optimized time out value and idle time is minimized.
Reinforcement learning is also used in scheduling, service provisioning, medium access control apart
from the routing in WSN.
CONCLUSION
This chapter explains various techniques applied to design secured energy efficient routing schemes for
wireless sensor networks using one of the most emerging technologies in current scenario, called ma-
chine learning. To meet the demands of the latest technologies and the research trend, this chapter will
be able to provide guidance to those who are interested to conduct research in the concerned domain.
The machine learning approach helps to enhance the overall performance by making the network intel-
ligent and effective. Collaboration with Internet of Things (IoT) based technology and environment will
bring more advanced and real-world features in WSN to proceed further. Future research may lead to a
solution of more matured protocol and applications of it in practical scenario.
REFERENCES
Ahmed, G., Khan, N. M., Khalid, Z., & Ramer, R. (2008, December). Cluster head selection using deci-
sion trees for wireless sensor networks. In 2008 International Conference on Intelligent Sensors, Sensor
Networks and Information Processing (pp. 173-178). IEEE. 10.1109/ISSNIP.2008.4761982
Ali, B., Mahmood, T., Abbas, M., Hussain, M., Ullah, H., Sarker, A., & Khan, A. (2019). LEACH Robust
Routing Approach Applying Machine Learning. IJCSNS, 19(6), 18–26.
Figure 10. Working mechanism of Reinforcement Learning
37
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
Alsheikh, M. A., Lin, S., Niyato, D., & Tan, H. P. (2014). Machine learning in wireless sensor networks:
Algorithms, strategies, and applications. IEEE Communications Surveys and Tutorials, 16(4), 1996–2018.
doi:10.1109/COMST.2014.2320099
Arabi, Z. (2010, June). HERF: A hybrid energy efficient routing using a fuzzy method in wireless sen-
sor networks. In 2010 International Conference on Intelligent and Advanced Systems (pp. 1-6). IEEE.
10.1109/ICIAS.2010.5716145
Beyens, P., Peeters, M., Steenhaut, K., & Nowe, A. (2005). Routing with compression in wireless sen-
sor networks: a q-learning approach. In Fifth European workshop on adaptive agents and multi-agent
systems (AAMAS 05) (Vol. 8). Academic Press.
Bhanderi, M. V., & Shah, H. B. (2014). Machine Learning for Wireless Sensor Network: A Review,
Challenges and Applications. Adv. Electron. Electr. Eng, 4, 475–486.
Bin, G., Zhe, L., & Ze-Jun, W. (2005, September). A dynamic-cluster energy-aware routing algorithm
based on neural structure in the wireless sensor networks. In The Fifth International Conference on
Computer and Information Technology (CIT’05) (pp. 401-405). IEEE. 10.1109/CIT.2005.8
Boyan, J. A., & Littman, M. L. (1994). Packet routing in dynamically changing networks: A reinforcement
learning approach. In Advances in neural information processing systems (pp. 671-678). Academic Press.
Branch, J. W., Giannella, C., Szymanski, B., Wolff, R., & Kargupta, H. (2013). In-network outlier detec-
tion in wireless sensor networks. Knowledge and Information Systems, 34(1), 23–54. doi:10.100710115-
011-0474-5
Chakraborty, U. K., Das, S. K., & Abbott, T. E. (2012, June). Energy-efficient routing in hierarchical
wireless sensor networks using differential-evolution-based memetic algorithm. In 2012 IEEE Congress
on Evolutionary Computation (pp. 1-8). IEEE. 10.1109/CEC.2012.6252985
Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977). Maximum likelihood from incomplete data via
the EM algorithm. Journal of the Royal Statistical Society. Series B. Methodological, 39(1), 1–22.
doi:10.1111/j.2517-6161.1977.tb01600.x
Dhawan, H., & Waraich, S. (2014). A comparative study on LEACH routing protocol and its variants
in wireless sensor networks: A survey. International Journal of Computers and Applications, 95(8).
Di, M., & Joo, E. M. (2007, December). A survey of machine learning in wireless sensor networks from
networking and application perspectives. In 2007 6th international conference on information, commu-
nications & signal processing (pp. 1-5). IEEE.
Dong, S., Agrawal, P., & Sivalingam, K. (2007, November). Reinforcement learning based geographic
routing protocol for UWB wireless sensor network. In IEEE GLOBECOM 2007-IEEE Global Telecom-
munications Conference (pp. 652-656). IEEE. 10.1109/GLOCOM.2007.127
Dorigo, M., Birattari, M., & Stutzle, T. (2006). Ant colony optimization. IEEE Computational Intel-
ligence Magazine, 1(4), 28–39. doi:10.1109/MCI.2006.329691
Duarte, M. F., & Eldar, Y. C. (2011). Structured compressed sensing: From theory to applications. IEEE
Transactions on Signal Processing, 59(9), 4053–4085. doi:10.1109/TSP.2011.2161982
38
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
Egorova-Förster, A., & Murphy, A. L. (2006, May). A feedback-enhanced learning approach for routing
in WSN. In Communication in Distributed Systems-15. ITG/GI Symposium (pp. 1-12). VDE.
Fadlullah, Z. M., Tang, F., Mao, B., Kato, N., Akashi, O., Inoue, T., & Mizutani, K. (2017). State-of-the-art
deep learning: Evolving machine intelligence toward tomorrow’s intelligent network traffic control systems.
IEEE Communications Surveys and Tutorials, 19(4), 2432–2455. doi:10.1109/COMST.2017.2707140
Fakhet, W., El Khediri, S., Dallali, A., & Kachouri, A. (2017, October). New K-means algorithm for
clustering in wireless sensor networks. In 2017 International Conference on Internet of Things, Embed-
ded Systems and Communications (IINTEC) (pp. 67-71). IEEE. 10.1109/IINTEC.2017.8325915
Forster, A. (2007, December). Machine learning techniques applied to wireless ad-hoc networks: Guide
and survey. In 2007 3rd international conference on intelligent sensors, sensor networks and informa-
tion (pp. 365-370). IEEE.
Förster, A., Murphy, A. L., Schiller, J., & Terfloth, K. (2008, October). An efficient implementation of
reinforcement learning based routing on real WSN hardware. In 2008 IEEE International Conference
on Wireless and Mobile Computing, Networking and Communications (pp. 247-252). IEEE. 10.1109/
WiMob.2008.99
Gui, T., Ma, C., Wang, F., & Wilkins, D. E. (2016, March). Survey on swarm intelligence based routing
protocols for wireless sensor networks: An extensive study. In 2016 IEEE international conference on
industrial technology (ICIT) (pp. 1944-1949). IEEE.
Ishmanov, F., & Bin Zikria, Y. (2017). Trust mechanisms to secure routing in wireless sensor net-
works: Current state of the research and open research issues. Journal of Sensors, 2017, 2017.
doi:10.1155/2017/4724852
Janakiram, D., Reddy, V. A., & Kumar, A. P. (2006, January). Outlier detection in wireless sensor net-
works using Bayesian belief networks. In 2006 1st International Conference on Communication Systems
Software & Middleware (pp. 1-6). IEEE. 10.1109/COMSWA.2006.1665221
Jaradat, T., Benhaddou, D., Balakrishnan, M., & Al-Fuqaha, A. (2013, July). Energy efficient cross-
layer routing protocol in wireless sensor networks based on fuzzy logic. In 2013 9th International
Wireless Communications and Mobile Computing Conference (IWCMC) (pp. 177-182). IEEE. 10.1109/
IWCMC.2013.6583555
Kadam, K., & Srivastava, N. (2012, March). Application of machine learning (reinforcement learning)
for routing in Wireless Sensor Networks (WSNs). In 2012 1st International Symposium on Physics and
Technology of Sensors (ISPTS-1) (pp. 349-352). IEEE.
Karp, B., & Kung, H. T. (2000, August). GPSR: Greedy perimeter stateless routing for wireless networks.
In Proceedings of the 6th annual international conference on Mobile computing and networking (pp.
243-254). 10.1145/345910.345953
Kennedy, J. (2006). Swarm intelligence. In Handbook of nature-inspired and innovative computing (pp.
187–219). Springer. doi:10.1007/0-387-27705-6_6
39
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
Khan, F., Memon, S., & Jokhio, S. H. (2016, November). Support vector machine based energy aware
routing in wireless sensor networks. In 2016 2nd International Conference on Robotics and Artificial
Intelligence (ICRAI) (pp. 1-4). IEEE. 10.1109/ICRAI.2016.7791218
Khan, Z. A., & Samad, A. (2017). A study of machine learning in wireless sensor network. Int. J. Com-
put. Netw. Appl, 4(4), 105–112. doi:10.22247/ijcna/2017/49122
Kim, D. S., & Chung, Y. J. (2006, June). Self-organization routing protocol supporting mobile nodes
for wireless sensor network. In First International Multi-Symposiums on Computer and Computational
Sciences (IMSCCS’06) (Vol. 2, pp. 622-626). IEEE. doi:10.1109/IMSCCS.2006.265
Kumar, D. P., Amgoth, T., & Annavarapu, C. S. R. (2019). Machine learning algorithms for wireless
sensor networks: A survey. Information Fusion, 49, 1–25. doi:10.1016/j.inffus.2018.09.013
Kumar, S. (2014). Improving WSN Routing and Security with an Artificial Intelligence Approach. In
DWAI@ AI* IA (pp. 62-71). Academic Press.
Lee, S., & Chung, T. (2004, October). Data aggregation for wireless sensor networks using self-organizing
map. In International Conference on AI, Simulation, and Planning in High Autonomy Systems (pp. 508-
517). Springer.
Liang, X., Balasingham, I., & Byun, S. S. (2008, October). A multi-agent reinforcement learning based
routing protocol for wireless sensor networks. In 2008 IEEE International Symposium on Wireless Com-
munication Systems (pp. 552-557). IEEE. 10.1109/ISWCS.2008.4726117
Macua, S. V., Belanovic, P., & Zazo, S. (2010, June). Consensus-based distributed principal component
analysis in wireless sensor networks. In 2010 IEEE 11th International Workshop on Signal Processing
Advances in Wireless Communications (SPAWC) (pp. 1-5). IEEE.
Masiero, R., Quer, G., Munaretto, D., Rossi, M., Widmer, J., & Zorzi, M. (2009, November). Data
acquisition through joint compressive sensing and principal component analysis. In GLOBECOM 2009-
2009 IEEE Global Telecommunications Conference (pp. 1-6). IEEE. 10.1109/GLOCOM.2009.5425458
Nehra, N. K., Kumar, M., & Patel, R. B. (2009, December). Neural network based energy efficient
clustering and routing in wireless sensor networks. In 2009 First International Conference on Networks
& Communications (pp. 34-39). IEEE. 10.1109/NetCoM.2009.56
Okdem, S., & Karaboga, D. (2009). Routing in wireless sensor networks using an ant colony optimiza-
tion (ACO) router chip. Sensors (Basel), 9(2), 909–921. doi:10.339090200909
Osisanwo, F. Y., Akinsola, J. E. T., Awodele, O., Hinmikaiye, J. O., Olakanmi, O., & Akinjobi, J. (2017).
Supervised machine learning algorithms: Classification and comparison. International Journal of Com-
puter Trends and Technology, 48(3), 128–138. doi:10.14445/22312803/IJCTT-V48P126
Pathan, A. S. K., & Hong, C. S. (2007, August). A secure energy-efficient routing protocol for WSN.
In International symposium on parallel and distributed processing and applications (pp. 407-418).
Springer. 10.1007/978-3-540-74742-0_38
40
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
Patil, M., & Biradar, R. C. (2012, December). A survey on routing protocols in wireless sensor net-
works. In 2012 18th IEEE International Conference on Networks (ICON) (pp. 86-91). IEEE. 10.1109/
ICON.2012.6506539
Pinto, A. R., Montez, C., Araújo, G., Vasques, F., & Portugal, P. (2014). An approach to implement data-
fusion techniques in wireless sensor networks using genetic machine learning algorithms. Information
Fusion, 15, 90–101. doi:10.1016/j.inffus.2013.05.003
Prajapati, J., & Jain, S. C. (2018, April). Machine learning techniques and challenges in wireless sensor
networks. In 2018 Second International Conference on Inventive Communication and Computational
Technologies (ICICCT) (pp. 233-238). IEEE. 10.1109/ICICCT.2018.8473187
Preeth, S. S. L., Dhanalakshmi, R., Kumar, R., & Shakeel, P. M. (2018). An adaptive fuzzy rule based
energy efficient clustering and immune-inspired routing protocol for WSN-assisted IoT system. Journal
of Ambient Intelligence and Humanized Computing, 1–13. doi:10.100712652-018-1154-z
Rajagopalan, R., & Varshney, P. K. (2006). Data aggregation techniques in sensor networks: A survey.
Academic Press.
Rani, Y., & Rohil, H. (2013). A study of hierarchical clustering algorithm. ter S & on Te SIT-2, 113.
Saleem, M., Di Caro, G. A., & Farooq, M. (2011). Swarm intelligence based routing protocol for wireless
sensor networks: Survey and future directions. Information Sciences, 181(20), 4597–4624. doi:10.1016/j.
ins.2010.07.005
Sarangi, S., & Thankchan, B. (2012). A novel routing algorithm for wireless sensor network using particle
swarm optimization. IOSR Journal of Computer Engineering, 4(1), 26–30. doi:10.9790/0661-0412630
Sarella, V. R., Reddy, P. P., Rao, S. K., & Padala, P. (2017). EEECARP: Efficient Energy Clustering
Adaptive Routing Procedure for Wireless Sensor Networks. Journal of Global Information Management,
25(4), 125–138. doi:10.4018/JGIM.2017100108
Sharawi, M., Saroit, I. A., El-Mahdy, H., & Emary, E. (2013). Routing wireless sensor networks based
on soft computing paradigms: Survey. International Journal on Soft Computing Artificial Intelligence
and Applications (Commerce, Calif.), 2(4), 21–36.
Sharma, V. K., Shukla, S. S. P., & Singh, V. (2012, December). A tailored Q-Learning for routing in
wireless sensor networks. In 2012 2nd IEEE International Conference on Parallel, Distributed and Grid
Computing (pp. 663-668). IEEE. 10.1109/PDGC.2012.6449899
Singh, A., Rathkanthiwar, S., & Kakde, S. (2016, April). Energy efficient routing of WSN using particle
swarm optimization and V-LEACH protocol. In 2016 International Conference on Communication and
Signal Processing (ICCSP) (pp. 2078-2082). IEEE. 10.1109/ICCSP.2016.7754544
Singh, K., & Kaur, J. (2017). Machine learning based link cost estimation for routing optimization in
wireless sensor networks. Adv. Wirel. Mob. Commun, 10, 39–49.
Tashtarian, F., Haghighat, A. T., Honary, M. T., & Shokrzadeh, H. (2007, September). A new energy-
efficient clustering algorithm for wireless sensor networks. In 2007 15th International Conference on
Software, Telecommunications and Computer Networks (pp. 1-6). IEEE.
41
Secured Energy-Ecient Routing in Wireless Sensor Networks Using Machine Learning Algorithm
Thangaramya, K., Kulothungan, K., Logambigai, R., Selvi, M., Ganapathy, S., & Kannan, A. (2019).
Energy aware cluster and neuro-fuzzy based routing algorithm for wireless sensor networks in IoT.
Computer Networks, 151, 211–223. doi:10.1016/j.comnet.2019.01.024
Townsend, L. (2018). Wireless Sensor Network Clustering with Machine Learning. Academic Press.
Venkataramana, S., Sekhar, B. V. D. S., Deshai, N., Chakravarthy, V. V. S. S. S., & Rao, S. K. (2019,
May). Efficient time reducing and energy saving routing algorithm for wireless sensor network. Journal
of Physics: Conference Series, 1228(1), 012002. doi:10.1088/1742-6596/1228/1/012002
Vikhar, P. A. (2016, December). Evolutionary algorithms: A critical review and its future prospects. In
2016 International conference on global trends in signal processing, information computing and com-
munication (ICGTSPICC) (pp. 261-265). IEEE. 10.1109/ICGTSPICC.2016.7955308
Vimalapriya, M. D., Vignesh, B. S., & Sandhya, S. (2019). Energy-Centric Route Planning using Ma-
chine Learning Algorithm for Data Intensive Secure Multi-Sink Sensor Networks. International Journal
of Innovative Technology and Exploring Engineering. IJITEE.
Wang, Y., Martonosi, M., & Peh, L. S. (2006, May). A supervised learning approach for routing opti-
mizations in wireless sensor networks. In Proceedings of the 2nd international workshop on Multi-hop
ad hoc networks: from theory to reality (pp. 79-86). 10.1145/1132983.1132997
Yan, J. F., Gao, Y., & Yang, L. (2011, July). Ant colony optimization for wireless sensor networks rout-
ing. In 2011 International Conference on Machine Learning and Cybernetics (Vol. 1, pp. 400-403).
IEEE. 10.1109/ICMLC.2011.6016670
Yang, M. S. (1993). A survey of fuzzy clustering. Mathematical and Computer Modelling, 18(11), 1–16.
doi:10.1016/0895-7177(93)90202-A
Zadeh, L. A. (1999). Fuzzy logic= computing with words. In Computing with Words in Information/
Intelligent Systems 1 (pp. 3–23). Physica. doi:10.1007/978-3-7908-1873-4_1
Zhao, W., Liu, D., & Jiang, Y. (2009, January). Distributed neural network routing algorithm based on
global information of wireless sensor network. In 2009 WRI International Conference on Communica-
tions and Mobile Computing (Vol. 1, pp. 552-555). IEEE. 10.1109/CMC.2009.103
Zhong, Y. (2016, August). The analysis of cases based on decision tree. In 2016 7th IEEE International
Conference on Software Engineering and Service Science (ICSESS) (pp. 142-147). IEEE.
Zou, K., Ouyang, Y., Niu, C., & Zou, Y. (2012, September). Simulation of Malicious Nodes Detection
Based on Machine Learing for WSN. In International Conference on Information Computing and Ap-
plications (pp. 492-499). Springer. 10.1007/978-3-642-34038-3_68
Zungeru, A. M., Ang, L. M., & Seng, K. P. (2012). Performance evaluation of ant-based routing pro-
tocols for wireless sensor networks. arXiv preprint arXiv:1206.5938
