Performance and Energy-Tuning Methodology for

Wireless Sensor Networks Using TunableMAC

Ekereuke Udoh Vladimir Getov
Distributed and Intelligent Systems Research Group Distributed and Intelligent Systems Research Group
University of Westminster University of Westminster
London, United Kingdom London, United Kingdom
[email protected] [email protected]

Abstract— Energy-efficiency and performance are at it is imperative to look at the existing approaches to curbing
the forefront with regards to wireless sensor networks due DOS attacks. This is discussed in Section II. Section III
to the resource-constrained nature of the sensors on the discusses the research methodology, followed by a discussion
network. Most of the energy in a sensor is consumed by of the results in Section IV. Section V concludes the paper and
the radio and this therefore creates the need for a more outlines directions for future work.
efficient use of the Media Access Control (MAC) layer
II. RELATED WORK
which controls access to the radio. The Castalia
framework which runs on the OMNET++ simulation A generic framework that optimizes the performance of
platform provides a MAC layer protocol – TunableMAC existing clustering protocols (e.g. UHEED) by using
– which is used in this paper for tuning of performance Simulated Annealing and K-Beam algorithms is proposed in
and consumed power. Our goal is to improve as much as [21]. However, this is mainly aimed at clustering and routing
protocols. In [22], the relationship between node density and
possible the performance/energy balance in terms of
certain network parameters such as received signal strength
resources used up by security features, while attempting
indicator (RSSI) and Link quality indicator (LQI) are
to preserve the overall lifespan of the wireless sensors. analysed, with reference to DOS attacks.
This paper investigates performance parameters for
TunableMAC such as energy consumed, latency, Gateway MAC (GMAC) was designed to protect against
throughput and network lifetime based on simulated broadcast attacks [13] which are usually targeted at the MAC
temperature sensors. A 5-step methodology is proposed layer. GMAC focuses on the network architecture as a way of
that can be helpful for minimizing the impact of denial- conserving energy and reducing the risk of DoS attacks by
of-sleep (DOS) attacks. Hence, the benefit of this research using cluster-based approach, making it better than SMAC,
is that it feeds into the development of a novel MAC TMAC and BMAC in terms of network lifetime. However, it
protocol that is energy-aware and can autonomously is relatively low in terms of autonomy as it focuses on the
MAC layer of the network.
guard against energy drain attacks such as DOS attacks.
In [8], the hash-based scheme, which also uses a cluster-
Keywords—TunableMAC protocol, performance tuning, based approach, is proposed to be better than the random vote
OMNET++, wireless sensor networks, energy aware scheme and round robin scheme. This evaluation is based on
the time and energy consumed in order to select a cluster head.
I. INTRODUCTION
Wireless sensor networks (WSNs) can be used in resource- In [4], Clustered Adaptive Rate Limiting (CARL)
constrained environments where there is limited or no access approach is designed to protect against an unauthenticated
to external energy sources e.g. underwater exploration and broadcast attack. It does this by using a host-based intrusion
battlefield surveillance. In addition, the sensors are also system which classifies incoming packets based on
constrained in terms of resources such as battery life as well authentication tests and anti-replay checks. Furthermore, it
as low memory due to their miniature size and need for limits the rate at which the radio remains active as a way of
portability [2]. To conserve energy, these sensor nodes go into minimizing the broadcast attack, however this method can
sleep mode to save energy. This makes them prone to energy- affect performance by limiting valid packets from being
drain attacks such as DOS attacks which prevent the sensors received.
from sleeping, thereby draining their energy, and reducing The Fake Schedule Switch Scheme involves a node
their lifespan significantly from 3.5 years to 3 days [4]. initiating a fake schedule switch if it does not receive an
In [5], DOS attacks are classified into six methods: sleep acknowledgment after sending a message [12]. This tricks the
deprivation, barrage, synchronization, replay, collision, and attacker into launching their attack at the wrong time and
broadcast attacks. These attacks are possible due to could minimize collision, exhaustion, and broadcast attacks.
vulnerabilities like frame collisions, message overhearing and However, this is applicable with MAC protocols that support
idle listening [3]. To mitigate these attacks, certain approaches Request-to-Send (RTS) and Clear-to-Send packets (CTS).
have been proposed and compared, however the evaluation of In [9], a secure wakeup scheme is proposed which
these approaches are qualitative in nature with a focus on their assumes that a radio can perform some check on the validity
strengths and weaknesses [5]. of a packet while the sensor node is still in sleep state. The
Although this paper focuses on TunableMAC as a form of radio is assumed to be able to store a list of tokens with which
test bed for analysing and tuning the changes in performance, it compares the tokens sent to it. While this seems energy-

978-1-7281-7315-3/20/$31.00 ©2020 IEEE

Authorized licensed use limited to: California State University Fresno. Downloaded on June 18,2021 at 23:35:18 UTC from IEEE Xplore. Restrictions apply.
efficient, it is not clear how the radio achieves this and how Consumed energy. This is the total amount of energy
much energy is spent. used to transmit data from a source to its destination. The unit
Absorbing Markov Chain, proposed in [14], is a for consumed energy is joules/bit.
mathematical model which can be useful for detecting a DOS Network Lifetime. This is how long the nodes on a
attack by making some probabilistic calculations on the network can stay alive from the point when they start working
expected death time of a sensor network while monitoring the
to the point where the nodes fail due to energy-drain attacks.
network flow. However, this method only is limited to
detecting the attack, without any controls or measures to B. TunableMAC Parameters
mitigate it. This section explains the various parameters of
A hierarchical collaborative model proposed in [2] utilises TunableMAC parameters that can be tuned [20].
distributed anomaly detection whereby the load of detecting
the anomaly is spread across multiple nodes to minimise the Justification for choosing TunableMAC. One of the
burden on a single node. To achieve this, nodes are categorised main reasons for using TunableMAC was the flexibility it
in various roles and clusters. The weakness of this model is gives, allowing a user to alter a number of parameters (8 to
that packet overhead may be high in some cases. 10) in order to suit the network needs.
Cross-layer mechanism as the name implies focuses on Duty cycle. The duty cycle is expressed as a percentage
data gathered from multiple layers of the network – MAC, or fraction of time for which the node listens or for which the
network and physical layers. They use a combination of node is active in duty.
routing table, RSSI and data fragment rejection to detect sleep
deprivation attack, replay attack and barrage attack Listen interval. While duty cycle is a fraction of time the
respectively. However, this mechanism has only been tested node listens, listen interval is the actual time for which the
on one protocol (SMAC) and only works in scenarios where node listens.
RTS and CTS are supported.
Beacon Interval Fraction. The presence of a duty cycle
In [11], the two-tier secure scheme is proposed with a means that a node that wants to transmit to sleeping nodes
strong focus on reducing the complexity of the security needs to wake up those nodes. This can be done using beacons
process and conserving energy while trying to tackle power as a form of preamble before sending the actual message.
exhausting attacks, specifically forge and replay attacks. It
uses a hash-chain generated dynamic session key which is Probability of transmissions. This is used alongside the
useful for authentication and encryption. number of transmissions or retransmissions to calculate the
Zero-Knowledge Protocol is proposed in [10] and is used expected number of successful transmissions per node.
alongside the interlock protocol for authentication via key
Number of transmissions. This has to do with the
transfer. The protocol claims to protect against man-in-the-
middle and replay attacks. However, there is no research on number of times data is transmitted.
the energy costs of this approach, especially as it involves Random Transmission Offset. This is the random time
RSA key generation and hash generation and distribution. for which a node delays before information is transmitted.
III. PERFORMANCE TUNING METHODOLOGY Retransmission Interval. This is the interval between
The performance tuning methodology was used to transmissions and is also expressed as an integer data type.
evaluate WSN MAC protocols based on their performance
via simulation experiments using OMNET++ and the Castalia Backoff Type. Based on CSMA technique, a node backs
framework. The methodology is divided into 5 steps: off each time the channel is busy. How long the node backs
off is determined by the back-off type.
1. Decide on what performance metrics to measure
2. Decide on TunableMAC parameters to use C. TunableMAC Limitations
3. Choose a set of values for each parameter There is no security to prevent DOS attacks as nodes can
4. Choose scenario and run simulations be kept awake through a stream of beacons. TunableMAC
5. Build a predictive model based on results does not support unicast and this leads to a waste of energy as
information is always sent to all neighboring nodes. Finally,
A. Performance Metrics it does not support RTS/CTS: Therefore, there is no form of
Certain performance metrics had to be used and these collision avoidance.
metrics include:
IV. WIRELESS SENSOR NETWORK SIMULATION
Latency. This has to do with any form of delay that RESULTS
happens during communication in the wireless sensor A. Introduction
network. Latency is measured in units of time – e.g. seconds. This simulation involves 16 temperature sensors
Throughput. As mentioned earlier, this is defined as the arranged in the form of a grid. These sensors sample their
amount of data successfully transferred from the source to temperature readings when it gets above 15 degrees. Any
destination within a given period. Throughput is usually node that senses a value above the threshold then broadcasts
measured in bits/second. However, in the simulator, this is this value. The value propagation which records how many of
referred to as the transmitted packets per time it took to the nodes received the broadcasted value is then recorded for
transmit those packets. each node. Energy consumed by the node is also recorded as

Authorized licensed use limited to: California State University Fresno. Downloaded on June 18,2021 at 23:35:18 UTC from IEEE Xplore. Restrictions apply.
well as the number of packets transmitted by the nodes. In this
scenario, only one node senses temperature beyond 15
degrees. The results are dependent on a number of parameters
associated with the TunableMAC protocol used in the
scenario. The parameters include duty cycle, beacon interval
fraction, and TX Power.

TABLE I. TUNABLE PARAMETERS USED

Duty Cycle Beacon TX Power

Interval
Fraction
1 0.02, 0.05, 0.1, 1.0 0dBm
0.5, 0.8
Fig. 2. Energy consumption based on duty cycles.
2 0.1 0.02, 0.05, 0.1, 0dBm
0.5, 0.8 The graph in Fig. 2 shows the effect of duty cycling on
3 0.1 1.0 -15, -10, -5, - energy consumption. It is clear that energy consumption
1, 0 increases as the duty cycle increases meaning that the
relatively highest consumed energy happens when duty cycle
is 0.8. The more a node listens, the more energy it consumes.
B. Varying Duty Cycle In [19] the authors conclude in their results and discussions
TABLE II. THROUGHPUT RESULTS
that energy consumption increases as duty cycle increases.
Among the three parameters (duty cycle, beacon interval
Duty Duty Duty Duty Duty faction and transmit power), the duty cycle relatively has the
Cycle Cycle Cycle Cycle Cycle greatest impact on energy consumption with its highest being
= 0.02 = 0.05 = 0.1 = 0.5 = 0.8 0.557.
0.992 0.975 0.992 0.933 0.867
TABLE IV. TRANSMITTED PACKETS

Duty Duty Duty Duty Duty

Cycle Cycle Cycle Cycle Cycle
= 0.02 = 0.05 = 0.1 = 0.5 = 0.8
117.017 45.825 22.808 3.733 1.733

Fig. 1. Throughput values based on duty cycles.

The graph in Fig.1. shows the value propagation which

indicates how many of the nodes received the propagated
value as the duty cycle is varied. One point to note is that the
change in value propagation is not linear and this is due to the Fig. 3. Transmitted packets based on duty cycles.
variations and randomness in the start times of the nodes (lack
of synchronization of sleep cycles). In [18] one of the ways The graph in Fig. 3 shows the impact of the duty cycle on
to ensure synchronization of schedules is for each node to the transmitted packets. Apparently, the lower the duty cycle,
send a SYNC message to other nodes to make them aware of the higher the transmitted packets. This is due to the fact that
its schedule. The main irregularity lies between the second a node spends less time listening and therefore can send more
and third bar from the left, where the duty cycle is 0.05 and data.
0.1 respectively. However, the graph still shows at large that
the value propagation reduces as duty cycle increases. The C. Varying Beacon
lowest value is 0.867 which happens when the duty cycle is
at the highest relatively (0.8). TABLE V. THROUGHPUT VALUES

Beacon Beacon Beacon Beacon Beacon

TABLE III. CONSUMED ENERGY RESULTS
Fraction Fraction Fraction Fraction Fraction
Duty Duty Duty Duty Duty = 0.02 = 0.05 = 0.1 = 0.5 = 0.8
Cycle Cycle Cycle Cycle Cycle 0.183 0.242 0.304 0.813 0.912
= 0.02 = 0.05 = 0.1 = 0.5 = 0.8
0.146 0.124 0.143 0.376 0.557

Authorized licensed use limited to: California State University Fresno. Downloaded on June 18,2021 at 23:35:18 UTC from IEEE Xplore. Restrictions apply.
 The graph in Fig. 6 shows the effect of varying the beacon
interval fraction on the transmitted packets. The higher the
beacon interval fraction, the higher the number of transmitted
packets.
D. Varying TX Power

TABLE VIII. THROUGHPUT RESULTS FOR TX POWER

TX TX TX TX TX
Power = Power = Power = Power = Power =
-15dBm -10dBm -5dBm -1dBm 0dBm
0.063 0.113 0.558 0.954 0.979
Fig. 4. Througput based on beacon interval fraction.

The graph in Fig. 4 shows the effect of the beacon interval

fraction on the value propagation. Value propagation
increases as the beacon fraction increases. This means that
more nodes are likely to receive the propagated value when
more beacons are sent.

TABLE VI. CONSUMED ENERGY

Beacon Beacon Beacon Beacon Beacon

Fraction Fraction Fraction Fraction Fraction
= 0.02 = 0.05 = 0.1 = 0.5 = 0.8
0.134 0.135 0.135 0.137 0.142 Fig. 7. Throughput based on transmission power.

Fig. 7 shows the effect of varied transmitted packet on the

value propagation and it clearly indicates that the value
propagation increases as the radio transmit power increases.

TABLE IX. CONSUMED ENERGY FOR TX POWER

TX TX TX TX TX
Power = Power = Power = Power = Power =
-15dBm -10dBm -5dBm -1dBm 0dBm
0.135 0.135 0.138 0.143 0.143

Fig. 5. Consumed energy based on beacon interval fraction.

The graph in Fig. 5 shows the impact of the beacon

interval fraction on the consumed energy. The higher the
beacon interval fraction, the higher the energy consumed.

TABLE VII. TRANSMITTED PACKETS

Beacon Beacon Beacon Beacon Beacon

Fraction Fraction Fraction Fraction Fraction
= 0.02 = 0.05 = 0.1 = 0.5 = 0.8
0.367 0.725 1.217 9.75 17.337
Fig. 8. Consumed energy based on varied TX power.

Fig. 8 shows the effect of the radio transmit power on the

consumed energy. The energy increases as the transmit power
increases. Hence, the reason why most radios for wireless
sensor networks do not exceed 0dBm of transmit power.

TABLE X. TRANSMITTED PACKETS FOR TX POWER

TX TX TX TX TX
Power = Power = Power = Power = Power =
-15dBm -10dBm -5dBm -1dBm 0dBm
1.438 2.587 12.842 21.946 22.521
Fig. 6. Transmitted packets based on beacon fraction.

Authorized licensed use limited to: California State University Fresno. Downloaded on June 18,2021 at 23:35:18 UTC from IEEE Xplore. Restrictions apply.
 [6] J. Rezaei, “Best-worst multi-criteria decision-making
method,” Omega, vol. 53, pp. 49–57, 2015.
[7] L. Xu and J.B. Yang, “Introduction to multi-criteria
decision making and the evidential reasoning approach,”
Manchester School of Management, 2001.
[8] M. Pirretti, S. Zhu, N. Vijaykrishnan, P. Mcdaniel, M.
Kandemir, and R. Brooks, “The sleep deprivation attack
in sensor networks: Analysis and methods of defense,” Int.
J. Distrib. Sens. Networks, vol. 2, no. 03, pp. 267–287,
2006.
[9] R. Falk and H.J. Hof, “Fighting insomnia: A secure wake-
up scheme for wireless sensor networks,” Proc. 3rd Int.
Conf. on Emerging Security Information, Systems and
Technologies (SECURWARE), pp. 191–196, 2009.
Fig. 9. Transmitted packets based on TX power.
[10] S. Naik and N. Shekokar, “Conservation of energy in
wireless sensor network by preventing denial-of-sleep
attack,” Procedia Computer Science, vol. 45, pp. 370–379,
Fig. 9 shows the effect of varying the transmit power on
2015.
the number of transmitted packets. It is evident that the higher
the transmit power, the higher the packets transmitted. There [11] C.T. Hsueh, C.Y. Wen, and Y.C. Ouyang, “A secure
scheme against power exhausting attacks in hierarchical
is a huge difference between the change in TX packets from wireless sensor networks,” IEEE Sens. J., vol. 15, no. 6,
15dBm to -10dBm and then from -10dBm to -5dBm. pp. 3590–3602, 2015.
Transmit power can be said to be directly proportional to the
[12] C. Chen, L. Hui, Q. Pei, L. Ning, and P. Qingquan, “An
value propagation, consumed energy and transmitted packets. effective scheme for defending denial-of-sleep attack in
wireless sensor networks,” Proc. of 5th Int. Conference on
V. CONCLUSION AND FUTURE WORK
Information Assurance and Security, IAS’09, pp. 446–
Part of the relevance of this research is that it becomes 449, IEEE Xplore, 2009.
easier to understand the impact of certain parameters on [13] M.I. Brownfield, “Energy-efficient wireless sensor
various aspects of a wireless sensor network thereby giving network MAC protocol,” PhD Thesis, Virginia Tech,
more insight as to where to focus on in terms of increasing 2006.
energy-efficiency. This then makes it easier to know which of [14] T. Bhattasali and R. Chaki, “AMC model for denial-of-
the parameters to tune in order to increase energy-efficiency sleep attack detection,” Journal of Recent Research
with minimal effect on throughput which can be measured Trends, pp. 1–4, 2012.
from the value propagation. [15] B. Tourancheau, F. Rousseau, A. Duda, L. Damon, and R.
Guizzetti, “Energy cost of security in an energy-harvested
Ongoing research makes use of this methodology as part IEEE 802.15.4 wireless sensor network,” Proc. 3rd
of developing an energy-efficient and autonomous protocol Mediterranean Conference on Embedded Computing
to tackle energy-drain attacks like DOS attacks. This research (MECO), pp. 198–201, IEEE Xplore, 2014.
opens up opportunities to look into how an energy-efficient [16] S. Isaiadis and V. Getov, “Integrating mobile devices into
and self-adaptive protocol which protects sensors from the Grid: Design considerations and evaluation," Proc. of
energy-drain attacks (denial-of-sleep attacks) can be Euro-Par 2005 Conference, LNCS, vol. 3648, pp. 1080–
developed. 1088, Springer, 2005.
[17] S. Panichpapiboon, G. Ferrari, and O.K. Tonguz,
REFERENCES “Optimal transmit power in wireless sensor networks,”
[1] V. Kumar, S.B. Dhok, R. Tripathi, and S. Tiwari, “Cluster IEEE Transactions on Mobile Computing, vol. 5, no. 10,
size optimization with Tunable Elfes sensing model for pp. 1432–1447, 2006.
single and multi-hop wireless sensor networks,” Int. [18] T. van Dam and K. Langendoen, “An adaptive energy-
Journal of Electronics, vol. 104, no. 2, pp. 312–327, 2017. efficient MAC protocol for wireless sensor networks,”
[2] T. Bhattasali, R. Chaki, and S. Sanyal, “Sleep deprivation Proc. 1st ACM International Conference Embedded
attack detection in wireless sensor network,” Int. J. Network Sensor System, pp. 171–180, ACM, 2003.
Comput. Appl., vol. 40, no. 15, pp. 19–25, 2012. [19] R. Tahar, A. Dhraief, A. Belghith, H. Mathkour, and R.
[3] M. Brownfield, Y. Gupta, and N. Davis, “Wireless sensor Braham, “Autonomous and adaptive beaconing strategy
network denial-of-sleep attack,” Proc. 6th Annual IEEE for multi-interfaced wireless mobile nodes,” Wireless
SMC Information Assurance Workshop, pp. 356–364, Communications & Mobile Computing, vol. 16, no. 12,
IEEE Xplore, 2005. pp. 1625–1641, Wiley, 2016.
[4] D.R. Raymond and S.F. Midkiff, “Clustered adaptive rate [20] B. A. Networks, “User's Manual,” March 2011.
limiting: Defeating denial-of-sleep attacks in wireless [21] A. Alchihabi, A. Dervis, E. Ever, and F. Al-Turjman, “A
sensor networks,” Proc. IEEE Military Communications generic framework for optimizing performance metrics by
Conference (MILCOM), pp. 1–7, IEEE Xplore, 2007. tuning parameters of clustering protocols in WSNs,”
[5] D.E. Boubiche and A. Bilami, “A defense strategy against Wireless Networks 25, pp. 1031–1046, Springer, 2019.
energy exhausting attacks in wireless sensor networks,” J. [22] E. Udoh and V. Getov, "Performance analysis of denial-
of Emerging Technologies in Web Intelligence, vol. 5, no. of-sleep attack-prone MAC protocols in wireless sensor
1, pp. 18–27, 2013. networks," Proc. 20th Int. Conference on Computer
Modelling and Simulation (UKSim-AMSS), pp. 151–156,
IEEE Xplore, 2018.

Authorized licensed use limited to: California State University Fresno. Downloaded on June 18,2021 at 23:35:18 UTC from IEEE Xplore. Restrictions apply.