Secure Data Aggregation Model (SDAM) in Wireless Sensor Networks

Abstract

Nowadays, Wireless Sensor Networks (WSN’s) are becoming more and more promising and applicable to a variety of fields: military, environmental, medical, wild life habitat, and transportation as well wearable devices, target-tracking. WSN are expected to be a main player in the Internet of Things technology. Power management is very important factor in considering WSN’s and it has been demonstrated that communication cost is higher than the computational as nodes consume most of the energy in communication. Added to the fact that sensors could be closely deployed and report the same reading, the data aggregation concept was introduced to resolve those issues and for the sake of better performance at a reduced cost. Nonetheless, sensing devices are prone to failure due to several aspects such as node failure or low batteries as well as being compromised. In this paper, we are introducing a novel method Secure Data Aggregation Model (SDAM) aiming at assuring a secure aggregate communication at a low cost (in terms of resources). Our simulation results showed that implementing SDAM resulted into an increase in the energy efficiency as well as a considerable reduction in cross-layering overhead.

Full text

Secure Data Aggregation Model (SDAM) in Wireless
Sensor Networks
Mohamed Ben Haj Frej
Computer Science and Engineering
University of Bridgeport
Bridgeport, CT 06604
mbenhaj@bridgeport.edu
Khaled M. Elleithy
Computer Science and Engineering
University of Bridgeport
Bridgeport, CT 06604
elleithy@bridgeport.edu
Abstract— Nowadays, Wireless Sensor Networks (WSN’s) are
becoming more and more promising and applicable to a variety
of fields: military, environmental, medical, wild life habitat, and
transportation as well wearable devices, target-tracking. WSN
are expected to be a main player in the Internet of Things
technology. Power management is very important factor in
considering WSN’s and it has been demonstrated that
communication cost is higher than the computational as nodes
consume most of the energy in communication. Added to the fact
that sensors could be closely deployed and report the same
reading, the data aggregation concept was introduced to resolve
those issues and for the sake of better performance at a reduced
cost. Nonetheless, sensing devices are prone to failure due to
several aspects such as node failure or low batteries as well as
being compromised. In this paper, we are introducing a novel
method Secure Data Aggregation Model (SDAM) aiming at
assuring a secure aggregate communication at a low cost (in
terms of resources). Our simulation results showed that
implementing SDAM resulted into an increase in the energy
efficiency as well as a considerable reduction in cross-layering
overhead.
Keywords— availability, base station (BS), cluster head (CH),
secure Data Aggregation model (SDAM), sensor node (SN),
wireless sensor network (WSN).
I. INTRODUCTION
Following to the digital modernization, there has been a
tremendous growth in Wireless Sensor Networks (WSN) field;
with many applications such as in the military, environmental
monitoring, agriculture production and medical field. WSN
consist on a considerable number of motes (sensor nodes)
placed in a known topology or on ad hoc basis in the desired
location in order to collect information [1]. Each mote is a low
end device assuring processing, sensing and communication
abilities. Below are the motes’ main characteristics:
 Limited battery life with no ability to recharge it in
certain situations or because it is not worth it as they
are meant to be disposable
 Low processing capabilities, as they should be
designed based on required resources only, in order
to keep them at a low price.
 Limited data storage capacity
Since the sensors are generally deployed close to each other,
they could be reporting the same values, there have been a
need to summarize the data then forward it to upper nodes for
the sake of energy consumption optimization.
Wireless sensor networks’ data aggregation consists on
reducing the flow of communication and optimizing the
energy consumption.
In data aggregation, designated motes will collect and
process the data packets from their surrounding nodes then
forward them to the base station making a huge impact on the
traffic as well as on the communication time. Thus it does its
job in increasing the efficiency. However it comes with its
inconvenience as it affects other constraints such as security
and delay. The aggregated sensors are vulnerable to intruders
which can inject false data in WSN.
Secure communication requires that the encryption will take
place before transmitting the data. While for aggregation
protocols, it would be preferable to process the data and
forward it to the next hop prior to encryption. So it is a
challenging task to combine both taking in consideration that
we could not apply the computing encryption protocols to
wireless sensor networks due to their limitation in resources.
Although there could be some compromise while adopting
aggregation and encryption, both are deemed essential for a
stable and accurately performing network [2].
Amongst every operation that might be performed by a
sensor node, gathering and transmitting data are the ones that
expend considerably higher vitality when contrasted with
other operations [3]. Data accumulation serves to lessen
correspondence by joining Data parcels. WSNs are as a rule
conveyed in threatening and unattended situations, so the need
for secure Data conglomeration is further emphasized.
II. RELATED WORK
There are numerous Secure Data Aggregation plans
proposed in writing. The base station (BS) takes wise choices

in view of the Data that is assembled amid Data collection
stage. Identifying anomaly values in the data frames is a vital
errand in choice making for different applications like
observing, issue determination and interruption location in
WSNs. Exceptions are characterized as "estimations that
altogether go a miss from the ordinary example of detected
data". The plausible wellsprings of anomalies in WSN
incorporate clamor and blunders, occasions, and vindictive
assaults [3]. Off base data will antagonistically influence the
estimation of the collected data which is utilized by the BS for
deciding. For instance, if the sensor nodes are conveyed in
backwoods to sense temperature esteem, then an exception in
temperature quality can show a flame and the BS will need to
call the flame station [4]. Henceforth, anomaly identification
is essential for distinguishing crisis in the system. Along these
lines, the exception recognition instrument ought to be
consolidated with data conglomeration plans [5]. In this paper,
they have proposed a secure data conglomeration conspire that
recognizes exception values. Their plan distinguishes
exception values and in the meantime gives data realness and
data uprightness. They have utilized multivariate information
investigation method to handle anomaly in associated
variables. It permits the BS to identify the false data infusion
amid the data conglomeration and to recognize the
compromised node.
A broad study of anomaly identification plans has been
conducted. The authors have considered a bunch based
topology with hubs sorted as: base station (BS), cluster head
(CH) and basic sensor nodes (SN). They have utilized
multivariate information investigation method to handle
anomaly in associated variables. They have utilized the PCA
model for flaw discovery [3]. The CH recognizes the arriving
information as ordinary or unusual. They have utilized factual
methods to discover sensor information to enhance the
exception discovery precision. They have utilized a
conglomeration tree as a system model. They have planned
anomaly recognition plan taking into account commit
disseminate system. The authors have presented a model that
executes inquiries, distinguishes and sends clients an
arrangement of readings that are expected to be exceptions [4].
They have utilized to distinguish a hub as anomaly hub. The
authors have proposed a SDA plan which utilizes a bunch
based system model. The group head runs the peculiarity
recognition calculation and sends the IDs and check of the
anomaly values to the guardian group head. Their
methodology is taking into account hearty vital segment
investigation [4]. To uncover the abnormalities the authors
have utilized the relationship among the detected information.
III. CLUSTERING – DATA AGGREGATION
A. Wireless Sensor networks Limitations
Since their elaboration, there have been continuous efforts to
design efficient and simple WSN’s. Hence, there are some
constraints to be considered while designing a wireless sensor
network with a simple topology [6].
 Energy considerations: nodes can do limited
computations as they have less memory and are limited in
power supply. Nodes completely depend on their batteries
for data transmissions as well as all the other tasks.
 Reliability and Fault Tolerance: the ability to keep
transmitting the information even when some of the nodes
become unavailable by taking another path.
 Scalability: when the number of nodes increases, it
becomes complex to manage and organize the nodes
which could affect the overall communication.
 Data Delivery Models: There are different models such as
event driven, query driven, and hybrid. These modes
control the flow of data from sensors to sink nodes. The
type of model used is based upon the applications.
 Quality of Service: there is a set of protocols to measure
the energy utilized, time taken for routing the information
and to prioritize the data transfer from sensor to sink
node.
 Network Dynamics: it is not always the case that all the
components in the WSN such as nodes, sink nodes,
sensors are static. There is a need to also support
mobility.
B. Clustering process:
Clustering process steps consists on the following:
 Get ready information: institutionalizing property and
characterizing measurement.
 Choosing property: picking viable property from essential
properties and putting them away into a vector.
 Collecting property: changing the picked properties to
new properties.
 Clustering: picking some sort of separation capacity that
has the fitting properties as the estimation of closeness
and after that grouping.
 Evaluating on the after effect of grouping: performing
assessment of three sorts: outside legitimacy assessment,
inside legitimacy assessment and relativity test
assessment [7].
C. Data aggregation process:
As previously mentioned, wireless sensor networks consist
on tiny sensors deployed in large number in the field. Each
node is a low end device that integrates data processing,
wireless communication and sensing abilities. Setting a base
node and data aggregation has to be done towards other nodes
in the network hierarchy and leads into saving energy. In case
of Multi hops network the data aggregation will consist on
getting the min, max and sum and sending it to the upper node
[8]. The data aggregation process totally depends on
mathematical models. It will decide the following:

 How to locate a bunching inclination in the
information.
 How to locate a superior approach to seek the bunch
and in addition the gathering hubs.
 How to discover a test system in order to demonstrate
that the segments are right and secure [9].
IV. SECURE WIRELESS SENSOR NETWORKS
A wireless sensor network is considered to be secured if it
offers all the following services:
 Authentication and Data Confidentiality: nodes can
carry confidential information of an organization. In
order to prevent any hacking attacks, nodes should
communicate within secured channels after the
authentication. Before giving access, WSN should
confirm the identity of all its nodes. The best way to
protect the data is to send it in an encrypted form by
using encoding and decoding techniques.
 Data Integrity: providing exact data sent by the
sensors without any alterations.
 Data Freshness: Data transferred should be fresh and
new. Nodes get busy unnecessarily if same message
is sent repeatedly, this could reduce the battery life of
the node and reduce the efficiency of the system.
 Availability: A node is completely dependent on it
battery life. A node should be able to send a message
if its battery life is going down and let the other
nodes take another path to communicate.
 Self-Organization: all the dynamic nodes should
organize themselves to form a multi path
communication system making the fault tolerance of
the wireless network more efficient.
 Secure Localization: Location of all the sensors in a
wireless sensor network should be traced accurately.
All these features make an ideal wireless sensor network.
But in reality, it is a bit difficult to design a WSN which could
strictly follows all the protocols. Formally a WSN should
authenticate all the nodes, detect and prevent any malicious
attacks, and have a data recovery solution in case one or more
nodes have been compromised
V. PROPOSED SOLUTION – SECURE DATA
AGGREGATION MODEL
In this paper we are proposing a new secure data
aggregation consisting of a sub layer between the MAC and
the network layers. SDAM’s primary objective is to ensure
secure communication by eliminating the dilemna of having
the data encrypted before aggregation while the aggregation
protocols perform better on unencrypted data.
A. SDAM Components:
The SDAM consists of three modules: Operations sub-layer,
Data aggregation module and Communications sub-layer
(please refer to figure 1).
 Operations Sub-layer responsible on checking based
on the MAC address of the device to determine if it is
coming from a legitimate or illegitimate node. If the
node is deemed illegitimate, the packets will be
dropped.
 Data Aggregation Module: responsible for removing
the repetition by applying aggregation algorithms as
well as deciding on the size of the packets.
 The Communications sub-layer is responsible to
decide on the number of packets that are needed to
aggregate and then forward them to the MAC layer.
Both incoming and outgoing traffic are sent by the
MAC layer and then forwarded to the SDAM
module.
Below is the diagram of the proposed solution:
Network Layer
MAC Layer
Operations
Sub-layer
Communications
Sub-layer
Data Aggregation
Figure 1: Secure Data Aggregation Model
VI. SIMULATION RESULTS
This section shows the experimental performance of our
proposed SDAM. The performance of the proposed model
was measured by using Network Simulator-2 (NS2) on

Ubuntu 14.2 operating system. The measurements are done on
a network size of 700 m x 700 m. We have assumed that there
are two types of nodes in the network:
 Homogenous
 Heterogeneous.
The homogenous nodes are distributed in the network that
sense and relay the data. The heterogeneous nodes are head
nodes in each cluster with more power and bandwidth. Each
homogeneous node initially uses 4 joules energy, whereas
each heterogeneous node uses 16 joules energy.
We have implemented our proposed model on each cluster
head node and assumed that the topology is static. Our goal in
this simulation is to apply the secure aggregation. The base
station is located at point (0, 850). The size of the packets is
128 bytes. The remaining parameters are given in Table 1.
PARAMETERS
VALUES
Size of network
1600 × 1600 square
meters
Number of nodes
180
Number of aggregation generated
1050
Energy of homogenous node
4 joules
Energy of heterogeneous node
16 Joules
Packet size
128 bytes
Data Rate
280
kilobytes/second
Sensing Range of node
35 meters
Simulation time
450 Seconds
Average Simulation Run
05
Base station location
(0,850)
Transmitter Power
13.2 mW
Receiver Power
11.4 mW
Table 1: Simulation Parameters
Based on the initial simulation results, we will be focusing
on evaluating the following metrics:
 Energy Efficiency
 Cross-Layering Overhead
A. Energy Consumption
Our generated scenario consists of 180 nodes that are
randomly distributed. 10% of the nodes are malicious
constitute a hurdle on the way of aggregation. Our proposed
SDAM is implemented on each cluster head node. Once, the
nodes collect the data from the sensing event that are
forwarded to the cluster head node, which is responsible to
apply aggregation and balance the communication between
MAC sub-layer and network layer.
Based on the results on Figure 2. Based on the results, we
observed that 1.34 joules energy is consumed when using
proposed model whereas, during the simulation period, while
1.53 joules energy is consumed without using SDAM.
Applying SDAM made the network more efficient in terms
of energy consumption and more secure as it detects the
malicious nodes and avoids data loss. Without SDAM, the
packets are captured by the malicious nodes because the
sensor nodes do not have the capability to identify the possible
malicious activities so that the captured packets are lost and
retransmitted.
As a result, the additional energy is consumed for
retransmission the packets. We have proved that our proposed
model saved 0.19 joules energy that is approximately 9.8%
saving in the entire network and that our proposed model
could be used to extend the network lifetime.
Figure2: Energy consumption using SDAM and without SDAM approaches
during the entire simulation time
B. Cross-Layering Overhead
Using the same simulation setup we have measured the
overhead cross-layering for 1800 aggregations (Figure 3-a)
and for 3600 aggregations (Figure 3-b).

Figure 3a: Cross-layering Overhead With and Without SDAM
Figure 3b: Cross-layering Overhead With and Without SDAM- Doubling the
number of Aggregations
Our simulation results shows the following: the overhead
with SDAM is: 0.7 milliseconds for 1800 aggregations and 0.8
for 3600 aggregations Without the SDAM, it is 1.64
milliseconds for 1.92 milliseconds for 3600 aggregations.
The simulation results shows 43 % improvement for 1600
aggregations and around the same improvement (42 %) with
3600 aggregations.
These results indicate that our proposed SDAM solution
shows considerable improvement in the cross-layering
overhead along with its energy efficiency optimization.
VII. CONLUSION
Considering the limited resources of the wireless sensor
networks and the impact of the use of encryption on
aggregated data; In this paper we have proposed a simplistic
and efficient method. The approach consisted 0n adding
modules able to authenticate non-legitimate nodes in the
network as well as facilitate the communications, in an energy
efficient way as shown by our simulation results. Our
simulation results also showed an around 40% improvement in
the cross-layering overhead.
ACKNOWLEDGMENT
Special thanks to Dr. Abdul Razaque, from the Cleveland
State University, who has been with a help especially in the
simulation section.
REFERENCES
[1] Razaque, Abdul, and Khaled Elleithy. "Modular Energy-Efficient and
Robust Paradigms for a Disaster-Recovery Process over Wireless Sensor
Networks." Sensors 15, no. 7 (2015): 16162-16195.
[2] Karthikeyan, B., Velumani, M., Kumar, R., and Inabathini, S.R.:
‘Analysis of data aggregation in wireless sensor network’, in Editor
(Ed.)^(Eds.): ‘Book Analysis of data aggregation in wireless sensor
network’ (IEEE, 2015, edn.), pp. 1435-1439.
[3] Rezvani, M., Ignjatovic, A., Bertino, E., and Jha, S.: ‘Secure data
aggregation technique for wireless sensor networks in the presence of
collusion attacks’, Dependable and Secure Computing, IEEE
Transactions on, 2015, 12, (1), pp. 98-110.
[4] Bharuka, K., and Jinwala, D.C.: ‘A secure data aggregation protocol for
outlier detection in wireless sensor networks using aggregate Message
Authentication Code’, in Editor (Ed.)^(Eds.): ‘Book A secure data
aggregation protocol for outlier detection in wireless sensor networks
using aggregate Message Authentication Code’ (2014, edn.), pp. 1-6.
[5] Sahana, S., and Amutha, R.: ‘Data aggregation in wireless sensor
networks’, in Editor (Ed.)^(Eds.): ‘Book Data aggregation in wireless
sensor networks’ (2014, edn.), pp. 1-6.
[6] Jose, J., Manoj Kumar, S., and Jose, J.: ‘Energy efficient recoverable
concealed data aggregation in wireless sensor networks’, in Editor
(Ed.)^(Eds.): ‘Book Energy efficient recoverable concealed data
aggregation in wireless sensor networks’ (2013, edn.), pp. 322-329.
[7] Durrani, N.M., Kafi, N., Shamsi, J., Haider, W., and Abbsi, A.M.:
‘Secure multi-hop routing protocols in Wireless Sensor Networks:
Requirements, challenges and solutions’, in Editor (Ed.)^(Eds.): ‘Book
Secure multi-hop routing protocols in Wireless Sensor Networks:
Requirements, challenges and solutions’ (IEEE, 2013, edn.), pp. 41-48.
[8] Mingxin, Y., Jingsha, H., and Xuguang, S.: ‘Research on Secure Data
Aggregation in Wireless Sensor Networks Based on Clustering Method’,
in Editor (Ed.)^(Eds.): ‘Book Research on Secure Data Aggregation in
Wireless Sensor Networks Based on Clustering Method’ (2011, edn.),
pp. 1-4.
[9] Ben Othman, S., Trad, A., Youssef, H., and Alzaid, H.: ‘Secure data
aggregation in wireless sensor networks’, in Editor (Ed.)^(Eds.): ‘Book
Secure data aggregation in wireless sensor networks’ (IEEE, 2013,
edn.), pp. 55-58.
[10] CAO, X.-m., LI, W.-l., and YANG, G.: ‘Research on Secure Data
Aggregation in Wireless Sensor Network’, Computer Technology and
Development, 2012, 11, pp. 060.