A NOVEL TWO-STAGE ALGORITHM PROTECTING INTERNAL ATTACK FROM WSNS

International Journal of Computer Networks & Communications (IJCNC) Vol.5, No.1, January 2013

A NOVEL TWO-STAGE ALGORITHM PROTECTING
INTERNAL ATTACK FROM WSNS

Muhammad Ahmed1, Xu Huang1and Hongyan Cui2
1
Faculty of Information Sciences and Engineering, University of Canberra, Australia
Muhammad.ahmed@Canberra.edu.au, xu.huang@canberra.edu.au
2
SICE, Beijing University of Posts and Telecommunications, China
cuihy@bupt.edu.cn

ABSTRACT
Wireless sensor networks (WSNs) consists of small nodes with constrain capabilities. It enables numerous
applications with distributed network infrastructure. With its nature and application scenario, security of
WSN had drawn a great attention. In malicious environments for a functional WSN, security mechanisms
are essential. Malicious or internal attacker has gained attention as the most challenging attacks to
WSNs. Many works have been done to secure WSN from internal attacks but most of them relay on either
training data set or predefined thresholds. It is a great challenge to find or gain knowledge about the
Malicious. In this paper, we develop the algorithm in two stages. Initially, Abnormal Behaviour
Identification Mechanism (ABIM) which uses cosine similarity. Finally, Dempster-Shafer theory (DST)is
used. Which combine multiple evidences to identify the malicious or internal attacks in a WSN. In this
method we do not need any predefined threshold or tanning data set of the nodes.

KEYWORDS
Wireless Sensor Network, Security, Internal Attack, Cosine similarity, Dempester-Shafer Theory

1. INTRODUCTION
Wireless sensor networks (WSNs) are a new technology for collecting data with autonomous
sensors [1]. This technology is first motivated by military applications. As example battlefield
surveillance, transportation monitoring, and sensing of nuclear, biological and chemical agents
[2-5] is considered. Recently, this technology is widely used in our daily life because they are
low cost, low power, rapid deployment, self-organization capability and cooperative data
processing, such as habitat monitoring [6], intelligent agriculture, home automation [7], etc. A
Typical WSN is shown in Figure1.

Figure1: A typical WSN
DOI : 10.5121/ijcnc.2013.5107 97

Extracting meaningful and actionable information from these applications, however, remains a
challenge. According to the applications of WSN the node deployment strategy is decided.
Normally, in WSN the environment is unknown, hostile, remote harsh fields, disaster are as
toxic environment. Hence there is no standard deployment strategy exist. The deployment
usually done by scatter by a possible way based the application scenario [6].

In all communication networks including WSN, security provisioning is a critical requirement.
WSN has different characteristics compare to the conventional communication networks that
includes open nature of wireless medium, unattended operation, limited energy, memory,
computing power, communication bandwidth, and communication range which makes the WSN
security mechanism tough. Even, it is more susceptible to the security attack in comparing to the
traditional wired network. The security of WSN can be investigated in different viewpoints,
following our paper [8] WSN attacks can be classified as two major categories: external and
internal attacks according to the domain of attacks. External attack is defined as the attack does
not belong to the network and it does not have any internal information about the network such
as cryptographic information. When a legitimate node of the network acts abnormally or illicit
way it is consider as a suspect of an internal attack. To perform the internal attack in the
network it uses the compromised node, which definitely can destroy or disrupt the network. In
this paper we focus on the internal attacks. The major internal attacks in WSN includes Denial
of Service (DoS) attacks, information and selective forwarding, Wormhole attack, node
replication, Sybil attacks or black-grey-sink holes, and HELLO flooding.

Considering the characteristics of WSN many algorithms have developed for the secure
functionality of WSN. Many existing research has focused on the pair wise key establishment of
the node for data exchange, authentication access control and defence against an attack. These
works mainly focused on the traditional cryptographic information, data authentication in order
to build the relationships among the sensors. But, the cryptographic methods sometimes are not
very efficient and effective. The unreliable communications through wireless channel can make
the communication technique vulnerable by allowing the sensor nodes to compromise and
release the security information to the adversary [9]. The compromised entity of the network
appears as a legitimate node. So it is easy for the adversary to launch the internal attacks. When
a node is attacked by a malicious the node will behave aberrantly. It will perform tampering the
massage from other member; data drop or excessive data broadcasting.

Considering the existing work, in order to protect WSN in an efficient way more attention is
necessary. Hence, in our research, we have proposed two-stage mechanism to find the internal
attack in a targeted WSN. In order to do that we first check the transmission range based on
individual regions. If the node is in the transmission range we use the cosine similarity method
to find the abnormally behaved node based on the message frequency with k-means algorithm
[10]. In the final stage to ensure about the internal attack we used Dempester- Shafer Theory
(DST). As DST has the feature of dealing with uncertainty [11]. In the both stages the algorithm
observes neighbour nodes parameters but the judgement is made based on DST. It considers the
observed data as hypothesis. In the observation; it might be uncertain which hypothesis fits best.
Therefore, DST makes it possible to model several single pieces of evidence within multi
hypotheses relations [12]. In our proposed method the system does not need to have any prior
knowledge of the pre-classified training data of the nodes.

The rest of the paper organizes as follows: section 3 presents the related works followed by the
characteristics of WSN. A discussion about internal attack is presented in this section 4. The
generic security requirements and vital security challenges is in section 5 and 6, respectively.
Section 7 describes our network and the system architecture. The description of new algorithm,
concept and implementation process how to detect internal attack is presented in Section 8.

98

Section 9 illustrates the simulation results. A brief discussion is given with our conclusion in
Section 10.

2. RELATED WORK
Multi-hop communication is used in WSNs in order to increase the network capacity as in
multi-hop routing; messages may traverse many hops before reaching their destinations.
However, the deployed sensor nodes are normally physically unprotected to make the network
cost effective. They are always deployed in open or hostile environments in a possible way.
Hence, they can be easily captured and compromised by the adversary, and then it can easily
extract sensitive information, control of the compromised nodes. So, those nodes end up by
giving the service for the adversary. Therefore, when a node is compromised, an adversary
gains access to the network and can produce malicious activities. The attacks are involved in
different fashion such as corrupting network data or even disconnecting major part of the
network.

The network layer attacks are discussed extensively by Karlof and Wagner in [13].They
mentioned altered or replayed routing information and selective forwarding, node
replication, Sybil attacks or black-grey-sink holes, and HELLO flooding. Other than
that some papers even discussed several attacks in term of network’s resiliency. In [14],
the researchers discussed how to keep WSN routing protocols as stateless as possible to
avoid the proliferation of specific attacks and provide for a degree of random behaviour
to prevent the adversary from determining which the best nodes to compromise are.
They defined three items, namely (i) average delivery ratio, (ii) average degree of
nodes, and (iii) average path length to describe the networks resiliency. Obviously, the
more efficient and effective ways are needed.

In [15] the authors addressed pollution attacks against network coding systems in
wireless mesh networks. They proposed a lightweight scheme, DART that uses time-
based authentication in combination with random liner transformations to defend
against pollution attacks.

A few papers also address pollution attacks in internal flow coding systems use special
crafted digital signatures [16-17] or hash functions [18-19]. Recently some papers
discuss the preventing the internal attacks by related protocols [20, 21].

It is noted that resiliency of WSNs are related to the security of WSNs [22], where a
definition of network resiliency was discussed based on the comparisons with other
similar terminologies such as robustness etc. In this paper we follow the definition of
[23], i.e. resiliency is the ability of a network to continue to operate in presence of k
compromised nodes to present our definition of “resiliency degree” and an algorithm to
control the compromised nodes with SDT in the targeted WSN. Here we automatically
take the assumption that the attacks are from internal when we highlighted the nodes
become “compromised nodes.”

In cryptographic approaches, the source uses cryptographic techniques to create and
send additional verification information that allows nodes to verify the validity of coded
packets.
In terms of the attack model, synoptically speaking there are two types of international
attacks, namely (i) exceptional message attack, by which the attacks will tamper the
99

message content or generate fake messages and (ii) abnormal behaviour attacks, by
which the transmission will be abnormally changed such as dropping the messages,
forwarding the message to a particular receivers, broadcasting redundant or meaningless
messages to increasing the traffic load in the network, etc. As we are focusing on the
controllable resiliency based on the internal attackers we shall focus on the case
abnormal attributes and some of cases (i) can be extended to what we discussed in this
paper.

3. CHARACTERISTICS OF WSN
WSN is currently used for real-world unattended physical environment to measure numerous
parameters. So, the characteristics of WSN must be considered for efficient deployment of the
network. The significant characteristics of WSN are described as follows [24]:

Low cost: Normally hundreds or thousands of sensor nodes are deployed to measure any
physical environment in WSN. In order to reduce the overall cost of the whole network the cost
of the sensor node must be kept as low as possible.

Energy efficient: Energy in WSN is used for different purposes such as computation,
communication and storage. Sensor node consumes more energy compare to any other for
communication. If they run out of the power they often become invalid as we do not have any
option to recharge. So, the protocols and algorithm development should consider the power
consumption in the design phase.

Computational power: Normally the node has limited computational capabilities as the cost and
energy need to be considered.

Communication Capabilities: WSN typically communicate using radio waves over a wireless
channel. It has the property of communicating in short range, with narrow and dynamic
bandwidth. The communication channel can be either bidirectional or unidirectional. With the
unattended and hostile operational environment it is difficult to run WSN smoothly. So, the
hardware and software for communication must have to consider the robustness, security and
resiliency.

Security and Privacy: Individual sensor node should have adequate security mechanisms in
order to avoid illegal access, attacks, and accidental damage of the information inside of the
sensor node. Furthermore, additional privacy mechanisms must also be included.

Distributed sensing and processing: The large number of sensor node is distributed uniformly or
randomly. In WSNs, each node is capable of collecting, sorting, processing, aggregating and
sending the data to the sink. Therefore the distributed sensing provides the robustness of the
system.

Dynamic network topology: In general WSNs are dynamic network. The sensor node can fail
for battery exhaustion or other circumstances, communication channel can be disrupted as well
as the additional sensor node may be added to the network that result the frequent changes in the
network topology. Thus, the WSN nodes have to be embedded with the function of
reconfiguration, self-adjustment.

Self-organization: The sensor nodes in the network must have the capability of organizing
themselves as the sensor nodes are deployed in an unknown fashion in an unattended and hostile

100

environment. The sensor nodes have work in collaboration to adjust themselves to the
distributed algorithm and form the network automatically.

Multi-hop communication: A large number of sensor nodes are deployed in WSN. So, the
feasible way to communicate with the sinker or base station is to take the help of an
intermediate node through routing path. If one node needs to communicate with the other node
or base station which is beyond its radio distance it must be through the multi-hop route by
intermediate nodes.

Application oriented: WSN is different from the conventional network due to its nature. It is
highly dependent on the application ranges from military, environmental as well as health
sector. The nodes are deployed randomly and spanned depending on the type of use.

Robust Operations: Since the sensors are going to be deployed over a large enough to cover the
required area and sometimes hostile environment. Hence, the sensor nodes have to be fault and
error tolerant. Therefore, sensor nodes need the ability to self-test, self-calibrate, and self-repair

Small physical size: Sensor nodes are generally small in size with the restricted range. Due to
its size its energy is limited which makes the communication capability low.

4. INTERNAL ATTACKS OF WSN
In the manufacturing process the cost-effectiveness of the sensor nodes is considered. In most
application of WSN the sensor nodes are usually not even physically well protected.
Considering the characteristics of WSNs it can easily be summarised that WSN is very easy to
be attacked internally. An adversary can easily corrupt the network by gaining the internal
information of the node. The attacks are involved in corrupting network data or even
disconnecting major part of the network. Following our previous paper [8] we have described
the major internal attacks as follows.

Denial of Service (DoS) attacks: DoS attack is an explicit attempt to prevent the authentic user
of a service or data. The common method of attack involves overloading the target system with
requests, such that it cannot respond to genuine traffic. As a result, the user can not get the
access to the desired service or system. The basic types of attack involved: Jamming, Tapering,
collision, Homing, flooding, etc. If the sensor network encounters DoS attacks, the attack
gradually reduces the functionality as well as the overall performance of the wireless sensor
network. Projected use of sensor networks in sensitive and critical applications makes the
prospect of DoS attacks even more alarming.

Wormhole attack: Just like the theoretical wormholes in space, this attacker can send packets,
routing information, ACK etc., through a link outside the network to another node somewhere
else in the same network. This way an attacker can fool nodes into thinking they are neighbours,
when they are actually in different parts of the network. This can also confuse routing
mechanisms that rely on knowing distances between nodes. A wormhole attack can be used as a
base for eaves dropping, not forwarding packets in a DOS like manner, alter information in
packets before forwarding them etc.

Sinkhole attack: This is a DOS attack, where a malicious node advertises a zero cost route
through itself. If the routing protocol in the network is a “low cost route first “protocol, like
distance vector, other nodes will chose this node as an intermediate node in routing paths. The
neighbours of this node will also chose this node in routes, and compete for the bandwidth. This
way the malicious node creates a black hole inside the network.

101

Sybil attack: The Sybil attack targets fault tolerant schemes such as distributed storage,
disparity, multipath routing and topology maintenance. This is done by having a malicious node
present multiple identities to the network. This attack is especially confusing to geographic
routing protocols as the adversary appears to be in multiple locations at once.

Selective forwarding attack: In this attack, malicious nodes can decide not to forward packets of
certain types or to from certain nodes. Even though the protocol is completely resistant to the
sinkholes, wormholes, and the Sybil attack, a compromised node has a significant probability of
including itself on a data flow to launch this type of attack if it is strategically located near the
source or a base station.

Spoofing attack: In this attack, a malicious node may be able to create routing loops,
wormholes, black holes, partition the network and etc., by spoofing, altering or replaying
routing information.

Hello flood attack: Many protocols require nodes to broadcast HELLO packets to announce
themselves to their neighbours. A node receiving such a packet may assume that it is within the
radio range of the sender but this assumption may be false.

Flooding attack: In this attack, a malicious node may send continuous connection requests to a
victim node effectively flooding the victim’s network link.

In network coding intermediate nodes actively mix or code input packets and forward the
resulting coded packets. The very nature of packet mixing also subjects network coding
systems to a severe security threat, knows as a pollution attack, where attackers inject corrupted
packets into the network. Since intermediate nodes forward packets coded from their received
packets, as long as least one of the input packets is corrupted, all output packets forwarded by
the node will be corrupted. This will further affect other nodes and result in the epidemic
propagation of the attack in the network. In [14], it addressed pollution attacks against network
coding systems in wireless mesh networks.

5. GENERIC SECURITY REQUIREMENTS OF WSN
The nature of a WSN leads a challenge to provide full security to the network. The ultimate
security requirement is to provide confidentiality, integrity, authenticity, and availability of all
messages in the presence of resourceful adversaries. In order to provide the complete security in
a WSN all message have to be encrypted and authenticated. An adversary can use natural
impairments to modify the original message or information as well as can make the information
unavailable because of WSN nature and uncontrolled environments. Security requirements in a
WSN are similar to the wireless ad hoc network [8, 25].

WSNs have the general security requirements of data confidentiality, authentication, integrity,
freshness and secure management.

Confidentiality: An adversary can choose any node to eavesdrop as long as it is within the radio
range as the signals are transmitted over the wireless channel. So, it is a threat for the data
confidentiality as the attacker can gain the cryptographic information.

Authentication: To determine the legitimate node and whether the received data has come from
the authorized node or not authentication is important.

Integrity: Information moving through the network could be altered. So integrity is important to
trust the received information from the network.
102

Freshness: To save the network from the replay packets it is important to ensure that the
received data is fresh and unused.

Secure management: It is important to manage the distribution of cryptographic keying material
in the network.

6. VITAL CHALLENGES OF WSN SECURITY
A WSN has three major properties that made the security mechanism challenging.

a. Resource Constraints

b. Operational Environment, and

c. Wireless Multihop Communication.

It is commonly assumed that sensor nodes are highly resource constrained; e.g., the resources
are comparable to the Berkeley MICA2 motes and TMote mini is presented in the Table 1 Thus,
security protocols for WSNs must be executable on the available hardware and especially must
be very efficient in terms of energy consumption and execution time.

Table1: Existing Sensor Platform [26], [27]

Characteristics Mica2 TMote mini

RAM (Kbytes) 4 10

Program Flesh Memory 128 48
(Kbytes)

Maximum data rate (Kbps) 76.8 250

Power Draw: Receive (mW) 36.81 57

Power Draw: Transmit (mW) 87.90 57

Power Draw: sleep (mW) 0.048 0.003

The operational environment of most WSNs is assumed to be unattended or even hostile. Since
sensor nodes are usually not assumed to be physically protected by some tamper-resistant
hardware, an adversary is able to compromise sensor nodes. Thus, even if security mechanisms,
such as node-based authentication, are deployed, an adversary is able to participate in the
network since the adversary has access to all data [14], e.g., cryptographic keys stored on the
node. Thus, security protocols must be able to operate even if sensor nodes are compromised.

The wireless communication enables an adversary to eavesdrop, inject, drop, or alter messages
or to perform denial of service (DoS) attacks by jamming the wireless channel. In contrast to
most other wireless networks, the communication is performed in a multihop way. This
introduces additional challenges. Compromised nodes may be part of a route, enabling them to
modify forwarded messages, or a compromised node injects a large amount of false messages to
drain the energy resources of all forwarding nodes.
103


7. NETWORK AND SYSTEM ARCHITECTURE
A network with N uniformly distributed sensor nodes over the area of Nx× Ny with self-
organized m clusters into squared field in a 2D scenario. Sensing nodes are responsible to
collect and forward the monitored data around them. Then, the collected data sent to the sinker.
In order to detect the abnormal behaviour of the sensor nodes we use ABIM and DST
mechanism if it is within the transmission range. We will consider the system is synchronized.

In addition, as a case study, for a temperature measurement we will consider that the sensor
deployed area with temperature varying from 8 degree to 14 degree in °C. Based on the
Gaussian distribution, within 2 sigma (standard deviation) we will accept the temperatures.
According to Holder et al [28], the choice of sigma value depends on the data set. 95.46% of the
sensor data will fall within 2 standard deviation of the mean with 2 sigma consideration. The
sampling rate is set to 0.1 Hz means 6 massages in a minute.

We will consider the temperature reading is normal or the node is behaving normally, if we find
that the reading match withour sigma value and with the one hop neighbours. If the outcome
from ABIM is abnormal node we will go ahead for the second stage implementation with DST.
In DST we will check both physical (temperature) and transmission packet drop rate (PDR)
parameter of the node.

Our temperature measurement in WSN system is based on a single sinker with randomly
distributed static node. We assume the neighbour node with one hop will observe the data of
the suspected internal attacker. Observed physical parameter (temperature) and transmission
behaviour (packet drop rate) is considered as independent events. The observation of the events
becomes the pieces of evidences. In the decision making process with Dempster-Shafer Theory
we will combine the independent pieces of evidences.

Figure 2: Three neighbour observing the attacker with one hop

Let’s take the above scenario described in Figure 2, the neighbour nodes X , Y and Z will
observe the suspected internal attacker node A for its temperature (T) and packet drop rate
(PDR). Before we go further discussion for Figure 2, we need to brief the ABIM and DST to

104

present our novel algorithm, which will described in section five. Then we shall apply our
designed algorithm to Figure 2 as study case for our initiative.

8. NETWORK MODEL AND DESIGNED METHOD
The system under consideration consists of an area of interest where region wise detection
requirements are provided by the end user. We model the area of interest as a grid of Nx× Ny
points. The ratio of the detection to miss requirements at every point on the grid are ordered in
two NxNy × 1 vector of the ratio of the probability, pd/ pm. There are two common sensing
models found in literature, binary detection model and the exponential detection model. Both
models share the assumption that the detection capability of a sensor depends on the distance
between the sensor and the phenomena, or target to be detected. Following [4] notations we
have the case that for the binary detection model, the probability of detection pd (t,s) is given as:

 1 if d ( t , s ) ≤ rd
p d (t, s ) = 
 0 if d ( t , s ) > rd (1)

whererd is the detection radius and d(t,s) is the distance between the target’s position “t” and the
sensor location “s” on a plane. The exponential model is a more realistic model, where the
probability of detection corresponds to

 e − α d ( t , s ) if d ( t , s ) ≤ rd
p d (t, s ) = 
 0 if d ( t , s ) > rd (2)

whereα is a decay parameter that is related to the quality of a sensor or the surrounding
environment. In the exponential model of equation (2), even if a target is within the detection
radius, there is a probability that it will not be detected, which means it will be missed. As this
model is closer to the realistic case, we shall use this model.

The process of linking individual sensors’ detection characteristic to the overall probability of
detection requirements on the grid is mathematically quantified using miss probabilities, pmiss =
1 −pd, where pd is the probability of detection. The overall miss probability M(x, y) corresponds
to the probability that a target at point (x, y) will be missed by all sensors, which is

M ( x, y ) = ∏ p miss (( x , y ), ( i , j )) u ( i , j )
( i , j )∈ Ω (3)

where u(i,j) represents the presence or absence of a sensor at the location (i, j) on the grid, and
corresponds to

 1, if there is a sensor at ( i , j )
u (i , j ) = 
 0, if there is no sensor at ( i , j ) (4)

Taking the natural logarithm of the both sides in equation (3), we have

m ( x, y ) = ∑ u (i , j ) ln p
( i , j )∈Ω
miss (( x , y ), ( i , j ))
(5)

where m(x, y) is so-called the overall logarithmic miss probability at the point (x, y). Thus we
have the function b(x, y) as
105


 ln p miss (( x , y ), ( 0 , 0 ), d (( x , y ), ( 0 , 0 ) ≤ rd
b( x, y ) = 
 0 , d (( x , y ), ( 0 , 0 )) > rd
(6)

The overall logarithmic miss probabilities for all points on the grid can be arranged in a vector
m of dimension NxNy× 1 that corresponds to equation (7) as shown below:

m = [ m ( x , y ), ∀ ( x , y ) ∈ Ω ] T

u = [ u ( i , j ), ∀ ( i , j ) ∈ Ω ] T

and m = Bu (7)

The ((i-1)Ny + j)-th element of u indicates the number of sensors deployed at point (i, j) on the
grid. The matrix B is of dimension NxNy × NxNy, and it contains

{ b ( x − i , y − j ), ∀ ( x , y ) ∈ Ω , ( i , j ) ∈ Ω }

b(x−i, y−j) corresponds to the (r, c)-th entry of B, where r = (x−1)Ny + y and c = (i−1)Ny + j.

Essentially, b(x−i, y−j) quantifies the effect of placing a sensor at the point (i, j) on the
logarithmic miss probability at the point (x, y) on the grid. If there are some compromised nodes
distributed in a WSN, how those compromised nodes could be detected by their so-called
abnormal attributes among the network, such as irregular change of hop count that implicates
sinkhole attacks; the signal power is impractically increasing which may indicate wormhole
attacks; abnormally dropping rate traffic behaviours related the related nodes most likely to be
compromised, etc.

In the initial stage we have designed ABIM that is sensitive to the abnormal event. In the
conventional cryptographic way it is not possible to detect the internal attacker because of the
unpredictable wireless channel. The unreliable channel makes it easy to compromise the node
and establish untrustworthy relationship [29]. The attacker always behaves abnormally, so it is
mandatory to identify the misbehaved node to secure the network.

WSN is densely deployed and continuously observe the phenomenon, this characteristics drive
WSN node normally encounter the spatial-temporal correlation. In our research we considered
the message generated from the nodes is similar for a defined period with the sampling rate if
0.1Hz (1 message per 10 second). Considering the limited storage of the sensor we store
minimum information of the Message in S. The message mi is consists of the content of the
representative message ( ∂ ) and frequency of the message ( α ). m i = ∂ i , α i , The set of the
message is shown in the equation (8)

S = {m1 , m2 , m3 ,..........mn } (8)

It is the set that will store the latest message that is sent to the network recently. When a new
message mnew is sent it arrives at the cluster head which can be authenticated by the similarity
function with S. The difference between the detected and average temperature is divergence. If
we denote Di (mnew ) as the divergence between the new and the normal message we have the set
as equation (9) [12].

106


{ }
D i ( mnew ) = D1 ( mnew ) D 2 ( mnew )..... D m ( mnew ) (9)

where,

D i (mnew ) = m M − mnew

Based on equation (8) if the data is different from the content considering the Gaussian
distribution of temperature and the threshold than it is new message. The threshold is defined as
i
the mean of the data set. If D (mnew ) is not within the threshold it is considered as new message.
For further authentication we will use the cosine similarity with frequency consideration. If we
consider new message frequency ω , the cosine similarity is in equation (10).

α .ω
COSIM =
α ω
(10)

If the two frequencies are similar it is considered as normal message otherwise it is considered
as false message and the node will be considered as abnormal node.

Algorithm 1

I. Get
mnew

M
For i = 1 to
i
If MinTh ≤ D (mnew ) ≥ MaxTh

printf “Good Node”

else go to II

II. for i = 1 to T

Execute the equation (10)

If COSIM i ≤ 0.6

printf “the node is an internal attacker”

else

Go to step I

end

107

In the simulation we set the sampling rate 0.1Hz from the 6 minutes observed imperial data and
our case study we have the calculation for the consign similarity for a one hop neighbour with
the abnormally behaved node.

α = {6 5 6 5 6 4}
ω = {1 0 3 2 1 5}

Cosine Similarity (COSIM) = 36 / (13.19) *(6.32)

= 36 / 83.42

= 0.43

With the decision of ABIM decision we further implement Dempester-shafer theory (DST). In
DST, probability is replaced by an uncertainty interval bounded by belief and plausibility.
Belief is the lower bound of the interval and represents supporting evidence. Plausibility is the
upper bound of the interval and represents the non-refuting evidence [30]. In this reasoning
system, all possible mutually exclusive hypothesis (or events) of the same kind are enumerated
in the frame of discernment also known as universal discloser θ . A basic belief assignment
θ
(BBA) or mass function is a function m: 2 → [0, 1], and it satisfies two following condition

m(φ ) = 0 (11)

∑ m( A ) = 1
j
A⊆θ (12)

In which φ is the empty set and a BBA that satisfy the condition m (φ ) = 0 . The basic
probability number can be translated as m(A) because the portion of total belief assigned to
hypothesis A , which reflects the evidences strength of support. The assignment of belief
function maps each hypothesis B to a valuebel (B) between 0 and 1. This defined as

bel ( B ) = ∑ m( A )
j: A j ⊆ A
j
(13)

The upper bound of the confidence interval is the plausibility function, which accounts for all
the observations that do not rule out the given proposition. It maps each hypothesis B to a value
pls (B) between 0 and 1, can be defined as follows.

pls( B ) = ∑ m( A )
j: A j ∩ B ≠φ
j
(14)

The plausibility function is a weight of evidence which is non-refuting to B. equation (15)
shows the relation between belief and plausibility.

pls( B) = 1 − bel(~ B) (15)

108

The hypothesis not B is representing by ~B. The functions basic probability numbers, belief and
plausibility are in one-to-one correspondence and by knowing one of them, the other two
functions could be derived [31]. Figure 3 shows the graphical representation of the above
defined measures belief and plausibility.

Figure 3: Measure of belief and plausibility

Assuming m1(A) and m2(A) are two basic probability number by two independent items of
evidence means two independent neighbour node which act as observers in the same frame of
discernment. The observations (the pieces of evidence) can be combined using Dempster’s rule
of combination (known as orthogonal sum) as in equation (16).

∑ m (A ) m (A )
i , j: Ai ∩ A j = B
1 i 2 j

(m1 ⊕ m2 )( B ) =
1− ∑ m (A ) m (A )
i , j: Ai ∩ A j =φ
1 i 2 j
(16)

where ⊕ represents the Dempster’s combination operator that combines two basic probability
assignments or basic belief assignments (BBA) into the third [17]. To normalize the equation
we consider L is a normalization constant defined by the equation (17), More than two belief
function can be combined with pairwise in any order.

1
L=
K (17)

where ,

K = 1− ∑ m (A ) m (A )
i , j: Ai ∩ A j =φ
1 i 2 j

The combination rule assigns the belief according to the degree of conflict between the
evidences and assigns the remaining belief to the environment and not to common hypothesis. It
makes possible to combine with most of their belief assigned to the disjoint hypothesis without
the side effect of a counterintuitive behaviour. Belief resembles the certainty factors or
evidences [18]. The conflict between two belief functions bel1andbel2, denoted by the Con(bel1,
bel2) is given by the logarithm of normalization constant [19] shown in equation (18)

Con (bel1 , bel 2 ) = log( L ) (18)
109

If there is no conflict between the bel1and bel2 than Con (bel1 , bel 2 ) = 0 and if there is nothing in
common between two evidences Con(bel1 , bel 2 ) = ∞ [20]. The DST automatically incorporates
the uncertainty coming from the conflicting evidences. Following the reference [20] we can
come up with a Dempester-shafer combination, which can be given as in equation (19)

L ∑ m (A ) m (A )
i , j: Ai ∩ A j = B
1 i 2 j

m( B ) = (m1 ⊕ m2 )( B ) =
1 + log( L) (19)

In order to find the internal attacks in our case we can execute the above framework with
equation (19). The algorithm used to do the simulation is shown below. The temperature
threshold ∂ T and ∂ PDR is the threshold for the packet drop rate which is set based on the training
data.

Algorithm 2

I. Get the view of the neighbor node view

Input: mT , mPDR , ∂ T , ∂ PDR

mT [ ] BPA assignment

mPDR [ ] BPA assignment

II. Execute the equation (9)

mT , PDR [ ]

If m( B ) < 0.6

Output result accepted

printf “the node is an internal attacker”

else

Go to step I

end

DST application in our system works by considering the independent event as temperature T
and PDR as described in section 4. Our case the universal discloser or the set of local element
can be observed by the one hop neighbouris θ = {T , PDR} . Hence the power set becomes

2θ = {φ , {T }, {PDR}, {unknown }}

110

Where,

{unknown} = {T }∪ {PDR}
In our specific case study and in simulation we have the imperial data as below. The observation
of nodeA by nodesX, Y and Z

mT ( X ) = 0.7 ; mT (Y ) = 0.75 ; mT ( Z ) = 0.65 ; mT (U ) = 0.1

m PDR ( X ) = 0.75 ; mPDR (Y ) = 0.7 ; mPDR ( Z ) = 0.75

Using the equation (19) the observation by X , Y and Z the combination becomes,

mT , PDR ( X ) = mT ( X ) ⊕ mPDR ( X )

mT , PDR (Y ) = mT (Y ) ⊕ mPDR (Y )

mT , PDR ( Z ) = mT ( Z ) ⊕ mPDR ( Z )

9. RESULTS AND SIMULATIONS
Temperature measurement is considered in our experimental work with randomly deployed
WSN. For the temperature range we have considered Gaussian distribution mean with 2 sigma
similar to the approach taken by holder et al. [10], even though in holder experiment he used 1
sigma for the constrain of data set but we assume we have sufficient data set to choose 2 sigma.
In the simulation environment the parameter we have set is shown as follows,

Table 2: The Parameters

Parameters Values

Packet Size 500 bytes

Initial Energy 2J

Packet Size 500 bytes

Regional Area (0,0) to (500,500)

The experiment was done in the MATLAB environment. In the deployed sensor field area we
have set the temperature range 8 to 14 degree centigrade. Gaussian distribution is used for the
mean and data threshold for ABIM with the training data. In the DST implementation we have
simulated 200 different observations by the neighbour nodes.

The abnormal behaviour of the node was identified according to the methodology of ABIM
which is described in section 5. Figure 4 shows the randomly distributed sensor filed and the

111

detected abnormal node in red (node 16). The value was set 29 degree to make node 16 as a
suspected node for the simulation purpose.

With the abnormal node detection of ABIM we farther implement Dempester-shafer Theory
(DST). Figure 5 describe the observation about the node A. it shows the observation by node X
(node 2), Y (node 7) and Z (node 12). from the figure it is clearly seen that three nodes
observation gives the common result between 75 % to 85% that the node A (Node 16) is
compromised or an internal attacker.

Figure 4. Sensor field

1

0.95

0.9
% of individual node observation

0.85

0.8

0.75

0.7

0.65

0.6

0.55
0 20 40 60 80 100 120 140 160 180 200
Nodes View in different time

Figure 5.Observation of node A by X,Y,and Zin Figure 2.
112


10. CONCLUSION
We have carefully investigated internal attacks for WSN and create a novel algorithm for
protecting WSNs from the internal attacks based on evaluation using ABIM and DST with our
case study, temperature measurements in this paper. We first illustrate the method to identify
the compromised nodes by the abnormal attributes. Internal attacker always mismatch with the
normal behaviour of the node in different both in physical and transmission parameters. With
the neighbour normal node evaluation using DST we have identified the internal attack in
Wireless Sensor Networks. The simulation results, shows different observation and DST
combination result for identified internal attacker.

In future, we would like to implement the algorithm in the hardware level to test in real time
environment.

ACKNOWLEDGEMENT
One of authors would like to thank the works has been partly supported by National Natural
Science Foundation of China (Grant No. 61201153), the National 973 Program of China under
Grant (Grant No. 2012CB315805), and the National Key Science and Technology Project
(Grant No. 2010ZX03004-002-02).

REFERENCES
[1] Huang X, Ahmed M, Sharma D, “Timing Control for Protecting from Internal Attacks in Wireless
Sensor Networks”, IEEE, ICOIN 2012, Bali, Indonesia, February 2012.
[2] D. Li, K. D. Wong, Y. H. Hu, and A. M. Sayeed, “Detection, classification, and tracking of targets”,
IEEE Signal Processing Mag., vol. 19, pp. 17-29, Mar 2002.
[3] C. Meesookho, S. Narayanan and C. Raghavendra, “Collaborative classification applications in
sensor networks”, Proc. of Second IEEE Multichannel and Sensor array signal processing workshop,
Arlington, VA, 2002.
[4] T. He, S. Krishnamurthy, J. A. Stankovic, T. Abdelzaher, L. Luo, R. Stoleru, T. Yan, L. Gu, J. Hui,
B. Krogh, “An energy-efficient surveillance system using wireless sensor networks”, MobiSys'04,
Boston, MA, 2004.
[5] B. Sinopoli, C. Sharp, L. Schenato, S. Shaffert, Sh. S. Sastry, “Distributed control applications within
sensor networks”, Proc. Of the IEEE, August 2003.
[6] P. Sikka, P. Corke, P. Valencia, C. Crossman, D. Swain, and G. Bishop-Hurley, "Wireless ad hoc
sensor and actuator networks on the farm," 2006, pp. 492-499.
[7] Z. Feng, "Wireless sensor networks: a new computing platform for tomorrow's Internet," 2004, pp. I-
27 Vol.1.
[8] Ahmed M., Hunag X., Sharma D., “A Taxonomy of Internal Attacks in Wireless Sensor Network”,
ICIS 2012, Kuala Lumpur, Malaysia 2012.
[9] Ahmed M., Hunag X., Sharma D., “A Novel Framework for Abnormal Behaviour Identification and
Detection for Wireless Sensor Networks”, ICIS 2012, Kuala Lumpur, Malaysia 2012.
[10] Shakhnarovish G., Darrell T., Indyk P., “ Nearest-neighbor meathods in learning na d vision”, (MIT
press 2005)

113

[11] Khalaja F., Khalajb M., Khalaj A.H., “Bounded Error for Robust Fault Detection under Uncertainty,
Part 1: Proposed Model Using Dempster-Shafer Theory”, Journal of Basic and Applied Scientific
Research 2012, 2(2)1233-1240, ISSN 2090-4304 (2012)
[12] Auden J, “ A logic for uncertain Probabilities”, international journal of uncertainity, Fuzziness and
Knowledge-Based Systems, Vol. 9, No. 3, June 2001.
[13] C. Karlof and D. Wagner, “Secure routing inn wireless sensor networks: attacks and
countermeasures,” Ad Hoc Networks, Vol. 1 no. 2-3, pp. 293-315, August 2003.
[14] Ochirkhand Erdene Ochir, Marine Minier, Fabrice Valois, and Apostolos Kountouris, “Resiliency of
Wireless Sensor Networks: Definitions and Analyses”, 2010 17th International Conference on
Telecommunications, pp828-835.
[15] Jing Dong, Reza Curtmola, and Cristina Nita Rotaru, “Parctical Defenses Against Pollution Attacks
in Intra-Flow Network Coding for Wireless Mesh Networks,” WiSec’09, March 16-18, 2009, zurich,
Switzerland, pp111-122.
[16] D. Charles, K. Jain, and K. Lauter, “Signatures for network coding,” 40th Annual Conference on
Information Sciences and Systems, 2006.
[17] Z. Yu, Y. Wei, B. Ramkumar, and Y. Guan, “An efficient signature-based scheme for securing
network coding against pollutions attacks,” in Proc. Of INFOCOM OS , Phoenix, AZ, April, 2008.
[18] M. Krohn, M. Freedman, and D. Mazierres, “On-the-fly verification of rateless erasure codes for
efficient content distribution,” Security and Privacy, 2004. Proc. 2004 IEEE Symposium on, pp.226-
240, 9-12 May 2004.
[19] C. Gkantsdis and P. Rodriguez, “ Cooperative security for network coding file distribution,” Proc. Of
INFOCOM 2006.
[20] Ahmad Ababnah, Balasubramaniam Naatarajan, “Optimal control based strategy for sensor
deployment.” IEEE Tran. On Systems, Man, and cybernetics, Part A: Systems and Humans, vol. 41,
no. 1 Jan. 2011.
[21] Ahmed Sobeih, Jennifer C Hou, Lu-Chuan Kung, Ning Li, Honghai Zhang, Wei Peng Chen, Hung-
Ying Tyan, Sun Yat-Sen and Hyuk Lim, “J-Sim: A simulation and emulation environment for
wireless sensor networks,” IEEE Wireless Communications, August 2006, pp104-119.
[22] Xu Huang, Muhammad Ahmed, and Dharmendra Sharma, “A Novel Algorithm for Protecting from
Internal Attacks of Wireless Sensor Networks” 2011 Ninth IEEE/IFIP International Conference on
Embedded and Ubiquitous Computing , Melbourne, Oct 24-26, 2011.
[23] Ochirkhand Erdene-Ochir, Marine Mibier, Fabrice Valois and Apostolos Kountouris, “Resiliency of
wireless sensor networks: definitions and analyses,” 2010 17th International Conference on
Telecommunications. Pp-828-835.
[24] C. Buratti , A. Conti , D. Dardari and R, Verdone, “An Overview on Wireless Sensor Networks
Technology and Evolution” , Sensors 2009, ISSN 1424-8220, pp 6869-6896, 2009.
[25] K. Lu et al., "A Framework for a Distributed Key Management Scheme in Heterogeneous Wireless
Sensor Networks", IEEE Transactions on Wireless Communications, vol. 7, no. 2, Feb. 2008, pp.
639-647.
[26] https://www.eol.ucar.edu/rtf/facilities/isa/internal/CrossBow/DataSheets /mica2.pdf, accessed on
November 24, 2011.
[27] http://sentilla.com/files/pdf/eol/Tmote_Mini_Datasheet.pdf, accessed on November 24, 2011.
[28] C. Holder, R. Boyles, P. Robinson, S. Raman, and G. Fishel, “Calculating a daily Normal
temperature range that reflects daily temperature variability”, American Meteorological Society, June
2006.
[29] W. T. Zhu, Y. Xiang, J. Zhou, R. H. Deng, and F. Bao, “Secure localization with attack detection in
wireless sensor networks,” International Journal of Information Security, vol. 10, no. 3, pp. 155-171,
2011.
114

[30] K. Sentz, “Combination of Evidence in Dempester-Shafer Theory”, System Science and Engineering
Department, Binghamton University, SAND 2002-0835, April 2002.
[31] U. Rakowsky, “Fundamentals of the Dempster-Shafer theory and its applications to system safety
and reliability modelling” RTA # 3-4, 2007, December – Special Issue.
[32] D. Koks, S. Challa, “An Introduction to Bayesian and Dempster-Shafer Data Fusion” , Published by
DSTO Systems Sciences Laboratory, Australia, November 2005.
[33] M. Tabassian, R. Ghaderi, R. Ebrahimpour, “Combination of multiple diverse classifiers using belief
functions for handling data with imperfect labels” Expert Systems with Applications 39, Elsevier
2011.
[34] F. Campos, S. Cavalcante, “An Extended Approach for Dempster-Shafer Theory” Information Reuse
and Integration, 2003. IRI 2003. IEEE 2003.

Authors
Muhammad Raisuddin Ahmed currently serves as Lecturer (Teaching Fellow) at
the Faculty of Information Sciences and Engineering, University of Canberra
(UC), Australia. He was a distinguished member of the Board of directors of
ITE&E Canberra Division, Engineers Australia in 2011. Besides, from March
2009 until July 2011, he was working as a Research officer and Project
coordinator of BushLAN project at the Plasma research Laboratory, Research
School of Physics and Engineering, at the Australian National University
(ANU), Australia. During this time he was also an academic in the College of
engineering and computer science at ANU from February 2010 till November
2011. He is pursuing his PhD at the UC, Australia. He has received Master of Engineering studies in
Telecommunication and a Masters of Engineering Management degree from the University of
Technology, Sydney (UTS), Australia. He obtained his Bachelor of Engineering (Hons) Electronics
Majoring in Telecommunications degree from Multimedia University (MMU), Malaysia. His Research
interest includes: Wireless Sensor Networks, Distributed Wireless Communication, Blind Source
Separation, RF technologies, RFID implementation.

Professor (Dr) Xu Huang has received the B.E. and M.E. degrees and Ph.D. in
Electrical Engineering and Optical Engineering prior to 1989 and the second
Ph.D. in Experimental Physics in the University of New South Wales, Australia
in 1992. He has earned the Graduate Certificate in Higher Education in 2004 at
the University of Canberra, Australia. He has been working on the areas of the
telecommunications, cognitive radio, networking engineering, wireless
communications, optical communications, and digital signal processing more
than 30 years. Currently he is the Head of the Engineering at the Faculty of
Information Sciences and Engineering, University of Canberra, Australia. He is
the Course Conveners “Doctor of Philosophy,” “Masters of Information Sciences (by research),” and
“Master of Engineering.” He has been a senior member of IEEE in Electronics and in Computer Society
since 1989 and a Fellow of Institution of Engineering Australian (FIEAust), Chartered Professional
Engineering (CPEng), a Member of Australian Institute of Physics. He is a member of the Executive
Committee of the Australian and New Zealand Association for Engineering Education, a member of
Committee of the Institution of Engineering Australia at Canberra Branch. Professor Huang is Committee
Panel Member for various IEEE International Conferences such as IEEE IC3PP, IEEE NSS, etc. and he
has published about two hundred papers in high level of the IEEE and other Journals and international
conference; he has been awarded 9 patents in Australia.

115

A/Prof.Hongyan Cui has received the Ph.D. in School of Telecommunications
Engineering in Beijing University of Posts and Telecommunications in 2006. She
is engaged in communication network research and development work since
2000. She has been participated in two National 973 Projects, four 863 Projects,
two National Nature Funds, a ministerial project, and a corporate-funded
research project. She has published over 30 papers in the important journals /
conferences, two books since 2003. She applied eight patents. She is the reviewer
of the " Chinese Journal of Electronics"," Journal of Communications " ,"Journal
of Beijing University of Posts and Telecommunications", "IEEE Networks Magazine SI", "Chaos" etc.
She has trained 34 undergraduate students for graduation design,, and guided 31 Masters, in which 16
have graduated , and now she also assisting guided 3 doctoral students. Her research interest is future
networks architecture, ESN, and Clustering.

116

A NOVEL TWO-STAGE ALGORITHM PROTECTING INTERNAL ATTACK FROM WSNS

Recommended

More Related Content

What's hot (20)

Similar to A NOVEL TWO-STAGE ALGORITHM PROTECTING INTERNAL ATTACK FROM WSNS (20)

More from IJCNC (11)

Recently uploaded (20)

A NOVEL TWO-STAGE ALGORITHM PROTECTING INTERNAL ATTACK FROM WSNS