Cluster Computing

https://doi.org/10.1007/s10586-018-1770-6

Genetic algorithm based sensor node classifications in wireless body

area networks (WBAN)
K. Kalaiselvi1 · G. R. Suresh2 · V. Ravi3

Received: 30 October 2017 / Revised: 29 December 2017 / Accepted: 8 January 2018

© Springer Science+Business Media, LLC, part of Springer Nature 2018

Abstract
Wireless body area network (WBAN) is a promising methodology in present health care systems to monitor, detect, predict
and diagnose the disease in people. The performance of the WBAN network is affected by un-trusted nodes in WBAN
network. The un-trusted sensor nodes are formed in WBAN network due to the attackers from outside the world. In this
paper, sensor node classification algorithm is proposed which incorporates ANFIS classifier based trusted and un-trusted
sensor nodes detection and classification system is proposed inorder to improve the efficiency of the WBAN networks. This
proposed system constitutes feature extraction and classification modules. The trust features are extracted from sensor nodes
and these exracted features are optimized using genetic algorithm. The performance of the WBAN network is analyzed in
terms of classification rate, packet delivery ratio and latency.

Keywords WBAN networks · Sensor nodes · Trsut features · Optimization · Classification

1 Introduction WBAN innovation has made its initially strides in the ther-
apeutic recovery and observing of patients [8]. In any case,
With maturing of the population, existing medicinal assets basic innovation is still in an early advancement organized
can’t fulfill future human services requests of seniors and and normally in light of particular remote correspondences
patients. Assets are constrained and it is inconceivable for advances [9,10]. Patients might be easily checked at home
most patients to manage the cost of long haul clinic remains while completing their day by day exercises and restora-
because of monetary limitations, work, and different reasons, tive staff needs to screen numerous patients all the while
despite the fact that their wellbeing status must be observed [11]. The harmony between these by and large clashing ele-
in a continuous or short occasional time mode [1,2]. Subse- ments implies that exploration in this field is not wrapped
quently, remote observing therapeutic frameworks will turn up. Around there, information unwavering quality, control
out to be a piece of versatile medicinal services focuses with utilization, and little size are essential attributes to con-
constant checking later on [3,4]. Figure 1 shows the multiple sider while picking proper WBAN sensor hubs [12]. Many
sensor nodes connected in WBAN network. examinations have been centered on WBANs for restorative
In this unique circumstance, WBAN supporting human purposes [13]. Notwithstanding, few works have been wor-
services applications can offer important commitments to ried about a worldwide answer for tens or many patients, each
enhance quiet social insurance, including determination as of whom is fitted with numerous sensor hubs, and bound to a
well as therapeutics observing [5–7]. In a brief timeframe, moderately little condition like a clinic or a living or lounge
area of a doctor’s facility [14]. Up to this point, a portion of
B K. Kalaiselvi the examination and studies did for clinic situations have got-
[email protected] ten comes about for various hubs in a few trial subjects [15].
In conventional system, the security is the primary concern
1 Faculty of Computer Science and Engineering, Saveetha in WBAN network, due to the transfer of data of the patients
Engineering College, Chennai, India
[16]. Hence, the sensor nodes in WBAN network require high
2 Department of Electronics and Communication Engineering, level of security to ensure the data transfer between nodes
Rajalakshmi Institute of Technology, Chennai, India
and different WBAN networks. The present systems con-
3 Faculty of Electrical and Electronics Engineering, K.S.R centrate the performance improvement in WBAN networks
College of Engineering, Tiruchengode, Namakkal, India

123
 Cluster Computing

Fig. 1 WBAN network

in terms of energy consumption, packet delivery ratio and to electroencephalography (EEG) data and uses this sensi-
throughput [17]. The requirement of high level security is tivity to develop algorithms that minimizes energy usage in
not considered in present WBAN networks [18]. Hence, this EEG sampling and optimizes the utilization of limited data
paper proposes a novel methodology for the classification buffer. However, this work does not consider other WBAN
of trusted and un-trusted nodes in WBAN networks using sensors. Samanta and Misra [22] developed a methodology
classification approach, which improves the security of the to solve the throughput rate reduction crisis in wireless body
WBAN networks [19]. networks using low-order and static antiwindup control laws
Proposed MAC protocol which was based on Quasi- to improve the overall performance of the WBAN networks.
Sleep-Preempt-Supported algorithm to reduce the energy Boudargham et al. [23] considered the problem of charging
consumption in WBAN networks. Resource efficiency is one and rate-allocation scheme which was based on valuation of
of the most important factor that should be considered when utility function. It was shown that socially optimizing solu-
developing a MAC protocol for CPS like WBAN. The authors tions can be obtained for achieving user optimization for
analyzed their proposed methodology in terms of average MAC protocols [24,25].
delay and throughput. Energy consumption was reduced
using the sleep mode in sensor nodes of WBAN networks.
Sun et al. [20] proposed threshold Distance metric technique
to reduce the energy consumption during transmission and 2 Methodology
reception in WNAN networks. Control packet overhead of
communication, idle listening of nodes to receive expected The soft computing based Sensor Node Classification (SNC)
data packets, overhearing, and collision of data packets are in WBAN network is proposed in this paper using GA (SNC
the major sources of energy dissipation in WBANs. This Algorithm). This methodology can be operated in training
proposed technique improves energy efficiency by the rates which trains the input pattern and testing which is used to test
of 71.05 and 60.81% in both baseline and offline schemes. the individual nodes in WBAN network. Initially, the features
Used Priority-Based Data-Rate Tuning methodology for syn- are extracted from the trusted and un-trusted nodes of WBAN
chronizing all sensor nodes within WBAN networks. The network and trained by ANFIS classifier. Further, the features
authors also used cooperative game theoretic approach to are now extracted from the test WBAN networks which can
reduce the interference between sensor nodes in WBAN be classified using ANFIS classifier. The flow of proposed
networks. Stollenwerk et al. [21] studied the procedure of methodology for the classification of nodes in WBAN net-
learning the sensitivity of neurometric application fidelity work is illustrated in Fig. 2.

123
Cluster Computing

Fig. 2 Proposed method (a)

training mode, (b) testing mode

2.1 Feature extraction where as, T11 is alpha factor and T21 is beta factor between
nodes I and k.
The trust features are used in this work to discriminate the Initially, T1 is 1 and T2 is 0 which leads to obtain the value
normal trust node in WBAN network from the un-trust node. of Trust1 is 1 for obtaining best feature value. When T1 is 1
The trust values are based on the estimation of trusts between and T2 is 1, the trust value leads to 0.5 for obtaining worst
two distinct nodes in WBAN networks. The trust values of feature value.
each node in WBAN network can be based on alpha factor (α) The accumulated trust value on node ‘I’ is given in the
and beta factor (β). The number of correctly sent packets is following equation as,
represented by alpha factor and number of degraded packets
is represented by beta factor. Trust = Weight (I,j) ∗ Trust1 + Weight (I,k) ∗ Trust2
The trust between node’I’I and node’ j’ which is centered
on node ‘I’ is,
The weight between nodes ‘i’ and ‘j’ is represented by
T1 weight (i,j) and weight between nodes ‘i’ and ‘k’ is repre-
Trust1 = sented by weight (i,k) and they are described in the following
T 1 + T2
equations as,
where as, T1 is alpha factor and T2 is beta factor between 
nodes I and j. b N 2 × |yd2 − y | + bmax , xd2 − x < 0
The trust between node’I’I and node’ k’ which is centered wdi (i, j) = (1)
bN 2, else
on node ‘I’ is,

T 11 b N 1 × (yd2 − y) + bmax , yd2 − y > 0
Trust2 = win (i, j) = (2)
T 11 + T21 bN 1, else

123
 Cluster Computing

where, b N 1 and b N 2 represents the bandwidth of the sensor performs node function of the extracted feature set which are
node 1 and sensor node 2, respectively. (xd2 , yd2 ) represents optimized by GA methodology. The internal functionalities
the coordinates of the sensor node 2 and (x,y) represents the of Layer1 are stated as,
coordinates of the sensor node 1. The maximum bandwidth
of the sensor node is noted as bmax . Oi1 = μ Ai (x) i = 1, 2
Oi1 = μ Bi−2 (y) i = 3, 4
2.2 Genetic algorithm
The membership functions used in this paper are Trian-
gular, Generalized bel and Gaussian membership functions.
The optimization techniques are implemented for selecting
This ANFIS architecture adopts any membership functions
the optimum number of features from the extracted feature
based on the optimized set of features.
set. Particle Swarm Optimization (PSO) algorithm [ ] was
Layer2 performs the firing strength of nodes in the internal
used as conventional optimization technique, in which fea-
architecture.
tures were considered as particles. Though this technique
provided optimum features for classifications, it consumed
ok2 = wk = μ Ai (x) μ Bi (y) , i = 1, 2; k = 2 (i − 1) + j
high latency for optimum feature selection. In this paper,
GA is used as optimization algorithm in order to overcome
Layer3 computes the normalized firing strengths of the
the drawbacks of the PSO optimization technique. The GA
nodes and these are stated as,
methodology consists of the modules as selection, cross over,
mutation and fitness. The procedure for this optimization wi
technique is explained in the following steps as, oi3 = wi = , I = 1, 2, 3, 4
w1 + w2 + w3 + w24
Step 1: Initial step for GA technique is selection process
Layer4 is called as consequent layer which computes the
of extracted features. It constitutes two chromosomes and
consequent vectors of the nodes in ANFIS architecture and
length of each chromosome is dependent of the extracted
it is represented as,
features which are to be optimized. At this stage, first ten
vectors from extracted features are loaded into chromosome1
oi4 = wi z i = wi ( pi x + qi y + ri ) , i = 1, 2, 3, 4
and next ten vectors from the extracted features are loaded
into chromosome2.
Layer5 performs summation of all nodes in ANFIS architec-
Step 2: Crossover between these selected chromosomes is ture and it is stated as,
performed by implementing single stage cross over method
between these chromosomes. w1 z 1+ w1 z 2 + w3 z 3+ w4 z 4
oi5 = wi = , I = 1, 2, 3, 4
w1 + w2 + w3 + w4
Step 3: Mutation between this two chromosomes is now
carried out by implementing single stage mutation process
between these chromosomes. 3 Results and discussion
Step 4: Compute Euclidean distance of all the vector ele-
ments in muted chromosomes. In this paper, 50 sensor nodes are considered and data rate
of each sensor node is 512 B per second. The transmission
Step 5: Keep the chromosome which has higher Euclidean
delay between one cycles of data between two consecutive
distances and discard the vector elements in another chro-
sequence is 100 ms and the size of each packet which are
mosome.
transferred by each sensor node is 512 Bytes. Each sensor
Step 6: Repeat the above steps till the final vector elements. node is operated by battery which supplies energy to the sen-
sor node for uninterrupted data transmission and reception.
2.3 Classification Each battery in sensor node is equipped 1000 J as initial
energy.
In this work, ANFIS classifier is implemented for the classifi- Table 1 shows the analysis of performance of the pro-
cation of trust node from un- trust node in WBAN networks. posed method with conventional methodologies on 10% of
The architecture of ANFIS classifier which is used in WBAN un-trusted nodes in WBAN network. In this paper, 5 sen-
network constitutes input layer, output layer and hidden layer, sor nodes are assumed as un-trusted sensor nodes over 50
as depicted in Fig. 3. This classifier architecture is explained sensor nodes in an individual WBAN network. These un-
in the following equations. trusted sensor nodes affect the performance of the entire
This ANFIS architecture contains five independent layers WBAN network in terms of Packet Delivery Rate (PDR),
as Layer1, Layer2, Layer3, Layer4 and Layer5. The Layer1 classification rate and latency. The ANFIS classifier in this

123
Cluster Computing

Table 2 Analysis of performance of the proposed method with conven-

tional methodologies on 20% of untrusted nodes in WBAN network
Methodologies Classification PDR (%) Latency (ms)
rate (%)

ANFIS classifier 96.76 97.4 12.98

(Proposed)
SVM-linear 94.64 95.7 19.75
SVM-quadratic 93.85 94.3 22.86
SVM-polynomial 94.75 95.1 21.56

120

100

80

Values
60
Fig. 3 Design of ANFIS architecture
40
Table 1 Analysis of performance of the proposed method with conven- 20
tional methodologies on 10% of untrusted nodes in WBAN network
0
Methodologies Classification PDR (%) Latency (ms) ANFIS classifier SVM-Linear SVM-Quadratic SVM-Polynomial
rate (%) (Proposed)
Methodologies
ANFIS classifier 97.86 98.1 10.76
(Proposed) Classification rate (%) PDR (%) Latency (ms)

SVM-linear 95.29 96.2 18.61

Fig. 5 Graphical comparisons on 20% of untrusted sensor nodes
SVM-quadratic 94.75 95.5 21.94
SVM-polynomial 95.45 96.7 20.56 untrusted nodes in WBAN network. In this paper, 10 sen-
sor nodes are assumed as un-trusted sensor nodes over 50
120 sensor nodes in an individual WBAN network. These un-
100 trusted sensor nodes affect the performance of the entire
WBAN network in terms of Packet Delivery Rate (PDR),
80
classification rate and latency. The ANFIS classifier in this
Values

60 paper achieved 97.76% of classification rate, 97.4% PDR

40 and 12.98 ms of latency. In conventional methodologies as
20 SVM-Linear achieved 94.64% of classification rate, 95.7%
PDR and 19.75 ms of latency. SVM-Quadratic achieved
0
ANFIS classifier SVM-Linear SVM-Quadratic SVM-Polynomial 93.85% of classification rate, 94.3% PDR and 22.86 ms of
(Proposed) latency. SVM-Polynomial achieved 94.75% of classification
Methodologies rate, 95.1% PDR and 21.56 ms of latency. Figure 5 shows the
Classification rate (%) PDR (%) Latency (ms) graphical comparisons on 10% of un-trusted sensor nodes.

Fig. 4 Graphical comparisons on 10% of untrusted sensor nodes

4 Conclusions
paper achieved 97.86 % of classification rate, 98.1% PDR
and 10.76 ms of latency. In conventional methodologies as In this paper, 5 and 10 sensor nodes are assumed as un-
SVM-Linear achieved 95.29% of classification rate, 96.2% trusted sensor nodes over 50 sensor nodes in an individual
PDR and 18.61 ms of latency. SVM-Quadratic achieved WBAN network. These un-trusted sensor nodes affect the
94.75% of classification rate, 95.5% PDR and 21.94 ms of performance of the entire WBAN network in terms of Packet
latency. SVM-Polynomial achieved 95.45% of classification Delivery Rate (PDR), classification rate and latency. The
rate, 96.7% PDR and 20.56 ms of latency. Figure 4 shows the classification rate of the proposed trust sensor node classifica-
graphical comparisons on 10% of un-trusted sensor nodes. tion system is improved using optimized genetic algorithm.
Table 2 shows the analysis of performance of the pro- The ANFIS classifier in this paper achieved 97.86 % of clas-
posed method with conventional methodologies on 20% of sification rate, 98.1% PDR and 10.76 ms of latency.

123
 Cluster Computing

References 17. Peng, H., Tian, Y., Kurths, J., Li, L., Yang, Y., Wang, D.: Secure and
energy-efficient data transmission system based on chaotic com-
1. Jing, L., Ming, L., Bin, Y., Wenlong, L.: A novel energy efficient pressive sensing in body-to-body networks. IEEE Trans. Biomed.
MAC protocol for wireless body area network. China Commun. Circ. Syst. 11(3), 558–573 (2017)
12, 11–20 (2015) 18. Sun, S., Gong, J., He, J., Peng, S.: A spreading activation algorithm
2. Misra, S., Moulik, S., Chao, H.-C.: A cooperative bargaining solu- of spatial big data retrieval based on the spatial ontology model.
tion for priority-based data-rate tuning in a wireless body area Cluster Comput. 18(2), 563–575 (2015)
network. IEEE Trans. Wirel. Commun. 14(5), 2769–2777 (2015) 19. Yu, R., Mak, T.W.C., Zhang, R., Wong, S.H., Zheng, Y., Lau,
3. Walsh, M.J., Hayes, M.J.: Throughput rate control for an 802.15.4 J.Y.W., Poon, C.C.Y.: Smart healthcare: cloud-enabled body sensor
wireless body area network using static and low order anti-windup networks. In: 2017 IEEE 14th International Conference on Wear-
techniques. In: Proc. Mediterranean Conf. Control Autom, pp. 1–6 able and Implantable Body Sensor Networks (BSN) (2017)
(2007) 20. Sun, Y., Wong, C., Yang, G.-Z., Lo, B.: Secure key generation using
4. Ren, Z., Qi, X., Zhou, G., Wang, H.: Exploiting the data sensitiv- gait features for body sensor networks. In: 2017 IEEE 14th Inter-
ity of neurometric fidelity for optimizing EEG sensing. IEEE J. national Conference on Wearable and Implantable Body Sensor
Internet Things 1(3), 243–254 (2014) Networks (BSN) (2017)
5. Yi, C., Wang, L., Li, Y.: Energy efficient transmission approach 21. Stollenwerk, A., Sehl, F., Marx, G., Kowalewski, S., Janisch, T.:
for WBAN based on threshold distance. IEEE Sensors J. 15(9), Enrichment of a diving computer with body sensor network data.
5133–5141 (2015) In: 2017 IEEE 14th International Conference on Wearable and
6. Kelly, F., Maulloo, A., Tan, D.: Rate control in communication net- Implantable Body Sensor Networks (BSN) (2017)
works: Shadow prices, proportional fairness and stability. J. Oper. 22. Samanta, A., Misra, S.: Energy-efficient and distributed network
Res. Soc. 49(3), 237–252 (1998) management cost minimization in opportunistic wireless body area
7. Bae, J.N., Choi, Y.H., Kim, J.Y., et al.: Efficient interference cancel- networks. IEEE Trans. Mobile Comput. 17, 376–389 (2017)
lation scheme for wireless body area network. J. Commun. Netw. 23. Boudargham, N., Bou Abdoy, J., Demerjianz, J., Guyeuxx, C.,
13(2), 167–174 (2011) Makhoul, A.: Investigating low level protocols for wireless body
8. Wang, L., Goursaud, C., Nikaein, N., et al.: Cooperative scheduling sensor networks. In: 2016 IEEE/ACS 13th International Confer-
for coexisting body area networks. IEEE Trans. Wirel. Commun. ence of Computer Systems and Applications (AICCSA) (2016)
12(1), 123–133 (2013) 24. Zhao, B., Gu, Y., Ruan, Y., Chen, Q.: Two game-based solution con-
9. Cheng, S.H., Huang, C.Y.: Coloring-based inter-WBAN schedul- cepts for a two-agent scheduling problem. Cluster Comput. 19(2),
ing for mobile wireless body area networks. IEEE Trans. Parallel 769–781 (2016)
Distrib. Syst. 24(2), 250–259 (2013) 25. Sandhya, R., Sengottaiyan, N.: S-SEECH secured-scalable energy
10. Alam, M.M., Berder, O., Menard, D., Sentieys, O.: TAD-MAC: efficient clustering hierarchy protocol for wireless sensor network.
traffic-aware dynamic MAC protocol for wireless body area sensor In: International Conference on Data Mining and Advanced Com-
networks. IEEE. J. Emerg. Sel. Topics Circ. Syst. 2(1), 109–119 puting (SAPIENCE) (2016)
(2012)
11. Kumari, P., Anjali, T.: Securing a body sensor network. In: 2017
9th International Conference on Communication Systems and Net-
works (COMSNETS), vol. 12(8) (2017) K. Kalaiselvi received her Bache-
12. Kim, B.-S., Kang, S.Y., Lim, J.H., Kim, K.H., Kim, K.-I.: A lor of Engineering degree in Com-
mobility-based temperature-aware routing protocol for wireless puter Science from Annamalai
body sensor networks. In: 2017 International Conference on Infor- University Chidambaram in 2003,
mation Networking (ICOIN), vol. 7(22) (2017) Master of Engineering degree in
13. Jijesh, J.J. Shivashankar,: ” A survey on wireless body sensor net- Computer Science from Annama-
work routing protocol classification. In: 2017 11th International lai University Chidambaram in
Conference on Intelligent Systems and Control (ISCO), vol. 5(11) 2005. She is currently working
(2017) as an assistant professor in the
14. Wang, Z., Yang, N., Guo, M., Zhao, H.: Human-human Interac- Department of Computer Science
tional synchrony analysis based on body sensor networks. In: IEEE and Engineering at Saveetha Engi-
Transactions on Affective Computing, vol. 12(23) (2017) neering College, Chennai. Curren-
15. Sulimov, A.I., Karpov, A.V., Lapshina, I.R., Khuzyashev, R.G.: tly, she is pursuing Ph.D. in Infor-
Analysis and simulation of channel non-reciprocity in meteor-burst mation and Communication Engi-
communications. IEEE Trans. Antennas Propag. 65(4), 2009–2019 neering from Anna University,
(2017) Chennai. She has published more than 10 research papers in journals
16. Liu, Y., Chen, Q., Tang, X., Cai, L.X.: On the buffer energy aware and conferences. Her main research interests include Wireless body
adaptive relaying in multiple relay network. IEEE Trans. Wirel. area networks, Wireless Sensor networks, IOT and Artificial Intelli-
Commun. 16, 6248–6263 (2017) gence. She is a member of CSI and ISTE

123
Cluster Computing

G. R. Suresh received his Bache- V. Ravi received his Bachelor’s

lor of Engineering degree in Elec- Degree in Electrical and Electron-
tronics and Communication from ics Engineering from K.S. Ran-
Manonmaniam Sundaranar Uni- gasamy College of Technology,
versity in 1997, Master of Engi- Tiruchengode in 2002, Master’s
neering degree in communication Degree in Power Systems from
systems from Madurai Kamaraj Annamalai University, Chidamba-
University, Madurai in 2000 and ram in 2004, Doctor of Philos-
Ph.D. degree in the Faculty of ophy in Electrical Engineering
Information and Communication from Anna University, Chennai in
Engineering from Anna Univer- 2013. He is presently working as
sity, Chennai in 2010. He is cur- a Professor in the Department of
rently working as a Professor & Electrical and Electronics Engi-
Head, Department of ECE, neering at K.S.R. College of Engi-
Rajalakshmi Institute of Technol- neering, Tiruchengode, Namakkal
ogy, Chennai. He has more than 18 years of experience in teaching in the year 2007 to Till date. He has more than 13 years of experience
at under graduate and graduate level. He has published more than in teaching at under graduate and graduate level. His areas of interest
60 research papers in journals and conferences. He has received the are Power Systems, Soft Computing Techniques and Smart Machines.
computer engineering division prize from the Institution of Engineers He has guided 15 under graduate and nine post graduate projects in
(India) for the year 2009. He is an active reviewer in the journals, different areas of Electrical and Electronics Engineering. He is a life
IET image Processing and International journal of Electronics. His member in Indian Society for Technical Education. He has presented
research area includes Medical Signal & Image processing, Speech seven international conferences and published 13 papers in interna-
processing and WSN in Telemedicine. He is a member of IEEE, IET tional journals.
and life member of ISTE.

123