Key management techniques in wireless sensor networks

International Journal of Network Security & Its Applications (IJNSA) Vol.6, No.6, November 2014
KEY MANAGEMENT TECHNIQUES IN
WIRELESS SENSOR NETWORKS
Khawla Naji Shnaikat1 and Ayman Ahmed AlQudah 2
1Department of Computer Science, University of Jordan, Amman, Jordan
2Deanship of e-learning, University of Dammam, Ad Dammam, KSA
ABSTRACT
The way that is used to achieve most important security requirements is the cryptography. Cryptography
mainly depends on what is called cryptography keys; the cryptographic keys required to be managed by
using a robustness technique that guarantees the needed security requirements. So first of all, the necessary
keys need to be distributed to the nodes before they are disseminated in the target area, and then let the
sensors that need to communicate establish its secure communication by having a deal on what is called
pair-wise key, and allow the refreshment process for those keys to be occurred successfully when it is
needed, and finally, having the ability to revoke the keys that related to compromised nodes. These phases
are performing the process that is called Key Management. In this paper we will explain different key
management schemes, critique them theoretically, and propose an idea as a way out for the expected
problems in one of these schemes.
KEYWORDS
Wireless Sensor Networks, Key Management, and Cryptography.
1. INTRODUCTION
Wireless Sensor Network (WSN) is the network that consist of many small devices that called
sensors. In literature, sometimes it is considered as a special type of the ad hoc networks [1].
These networks are useful in our life; they are widely used in many applications. WSN is used in
military, commercial, and in ecological. Thus, the communications in these networks must be
secure.Securing the communication in WSN is a very important issue, just because of many
security threats and because of the nature of WSN. We can notice this point by studying the
communication links in these networks, which is the radio links that is subject to many faulty
information and malicious attacks. Sensor devices in general have a limitation in its resources;
these limitations can control the nature for WSNs, also affect the security level for this type of
networks. As an example for such limitations; limited processing power, battery age, transmission
distance, shortage in memory space, random distribution for nodes, and bandwidth [2].
• Limited processing power: Sensor node as any other technical device has its own
hardware and software architecture [3]. The key factor when manufacturing a sensor
node is to keep the hardware and software architecture for the sensor node simple as
much as possible to save node energy [3].
• Battery age: The sensor nodes are fully dependent on its battery, since it is operating
in outdoor environment [4]. The only energy resource for sensor nodes is its battery,
so once it is out the sensor node is considered as dead node [4].
• Transmission distance: The transmission distance or range for wireless devices is
limited. The more distance the sensor node can send, the more energy will be
DOI : 10.5121/ijnsa.2014.6605 49

consumed [2]. Thus, the transmission range for sensor nodes is restricted regarding to
the limitation in sensor resources.
• Shortage in memory space: Sensor nodes just as other related devices have its own
operating system software. This software is loaded to the sensor memory and takes
part of the memory space. Thus, rest of the memory space needs to be managed in an
efficient way [2]. The keys that are needed in the cryptography process in WSN
needs a light management mechanism that do not waste the residual memory space,
since there is another data in WSN that needs this space, while the network is
operating[2].
• Prior deployment knowledge: In general, sensor nodes are deployed in risky areas.
For example, war areas. Thus, the deployment process for sensor nodes is a critical
issue most time and it could be done randomly or dynamically [2]. This issue
contributes in making the location information for the sensor nodes places
insufficient or most likely not available. The key management techniques should not
depend on the sensor location specially when starting keying process [2].
• Bandwidth: The key establishment process in WSN is limited to sensor nodes
constraints. As mentioned before, the hardware architecture for sensor nodes is
simple. The transmitter in such nodes has limited capabilities, so it cannot transmit a
large set of data at the same time. This means that the bandwidth capacity for sensor
nodes is small. Therefore, the key management techniques should take in its account
this issue [2].
Since the WSN has such constraints, and is vulnerable to attackers, securing it is considered a
great challenge. Many algorithms have been proposed in the literature to achieve this task. WSN
needs a cryptography algorithm that must be selected carefully, and the most important factor for
those algorithms is solving the Key agreement or management problem, that is needed to provide
an encrypted and authenticated data transmission between the sensors nodes to have a secure
channel. The cryptography algorithm has to meet many securities requirements such as,
confidentiality, integrity, authenticity, and availability (CIAA) [5].
• Confidentiality: This is one of the most important security requirements. It means that
the secrecy of the exchanged data between nodes must be guaranteed, and could not
be detected by any illicit sides. Therefore, the key establishment mechanism is
considered best if it has the ability to guarantee the protection for the keys, so the
exchanged data will be secured [2].
• Integrity: This security requirement means that any type of modification for the data
must be prevented. The way to guarantee this issue in WSN is to use key
management techniques that forbid any illicit node from making any changes on the
keys [2].
• Availability: In general, availability means to have the service all the time. In WSN
there are many factors that cause a denial of service. The constraints that are found in
sensor resources have a great effect on the availability requirement. For example, as
mentioned before, the battery life for sensor nodes is short; this is cause an
interruption in the service. So the being used key establishment technique should take
into its account how to maintain a lasting service in the WSN.
• Authenticity: In security, the authentication means that the node needs to proof its
identity. The key management mechanism should not allow for any strange node to
participate in any communication process that takes place in the WSN.
With the existence of all previous difficulties that related to the sensor devices nature, the key
establishment techniques need to be efficient in a way that can handle such issues. To judge the
efficiency and goodness for the key establishment mechanism, many metrics that was mentioned
50

in the literature, could be used to assess the technique. Here we will brief some of them to see if
the key management technique could overcome the obstacles that mentioned before. Theses
metrics are resistance, resilience, revocation, the scalability [2].
• Resistance: As we said previously, the sensor nodes are distributed in dangerous and
far target areas which make it hard to monitor it all the time. This makes the sensor
nodes subject to be physically seized by the attackers. The attacker or adversary once
catch the sensor node will try to get any information that could help in revealing the
secret keys and data in the WSN. In addition, the attacker may make many changes to
compromise some nodes in the network, and place these nodes in the WSN, by this
the attacker roles the WSN and the confidentiality for WSN is penetrated [2][4].
• Resilience: If the adversary success in capturing some sensor nodes, the key
establishment mechanism should prevent him from making any use of that. This
means that the technique should guarantee the robustness for the sensor nodes against
the physical attack. The attacker must not success in his trials to get the secret keys
and penetrating the confidentiality for the WSN [2].
• Revocation: This comes to get the nodes that were attacked and pervaded by the
adversary back. The key establishment mechanism should guarantee this issue in
WSN, and should maintain it with respect to the limited resources for sensor devices
[2].
The sequence goes as follow; the adversary captures the sensor node and tries to reveal the secret
keys to reveal the data. The key management technique could not prevent the nodes from being
physically captured, but it should guarantee the resilience. If the key establishment mechanism
did not guarantee the resilience, then it should guarantee the revocation for the seized nodes. As a
result, through providing and maintaining the resistance, resilience, and revocation. The key
establishment technique is considered and evaluated to be efficient.
• Scalability: In general, scalability means that the solution for small size problem
should work properly if the problem size gets larger. WSN is subject to be larger
while it operates, i.e. new nodes could be inserted to the network. The key
management technique should be scalable and allow for the new nodes to work in a
secure manner fairly [2].
• Adjustability: In the context of key management, being adjustable means that the
mechanism should work properly if any network conditions change. For example,
having dead nodes in the networking or losing connectivity.
The previous introduction leads us to the importance of having an efficient key management
technique. This technique approach sets a key between whole sensor nodes that aim to exchange
data in secure manner, handle the join and disjoin of nodes, the ability to work in undefined
deployment environment, and could prevent the unauthorized nodes from establish
communication with the network nodes. And it needs to success in fulfilling those goals under
WSN constraints, which means considering simplicity in working to not consuming the sensors
resources.
51

52
2. PHASES FOR THE KEY MANAGEMENT PROCESS
Various schemes have been proposed in the literature for key management techniques, these
schemes have focused on many phases that are needed for this process to secure the WSN and to
overcome the aforementioned obstacles in WSN. We illustrate here three key management
schemes and explain the most important phases for each one of them. The three management
techniques that we analysing are:
2.1. The First Technique that mentioned in [6]
As the name implies, this technique is designed for the Heterogeneous Sensor Networks (HSN)
that is formed of many clusters. Each cluster is composed of one highly equipped sensor node that
is called the cluster head or sink [6], and a number of less equipped sensor nodes, which are the
typical sensor nodes. This key management scheme is having the following phases:
2.1.1. Pre-distribution Phase
This step is happened before the deployment of the sensors, there are many mechanisms to do this
step, such as, the Pair-wise Key Pre-distribution, the Master Key Based Pre-distribution, the Base
Station Participation, and the Probabilistic Key Pre-distribution. The Base Station (BS)
mechanism is used in the Key-chain approach. So that, some calculations need to take place prior
to the nodes deployment process. These calculations start with the generation of two key chains.
These chains generation is done by the Base Station (BS) using the two one-way functions F1 and
F2:
{n10k10, n11k11, n12k12, n13k13..., n1nk1n}
{n20k20, n21k21, n22k22, n23k23..., n2nk2n}
Where k1 (n+1) = F1 (k1n), k2 (n+1) = F2 (k2n); n1n is the ID of key k1n on the first key-chain, and n2n is
the ID of key k2n on the second key-chain. Each sensor, i.e., typical node, is pre-loaded with n1, n2,
and an initial key Kinit.
n1 and n2 are the IDs of each key on the two key chains, that have been generated by BS for each
sensor, and Kinit = K1n1 ⊕ K 2n2 .
The Cluster Head, i.e. the highly equipped node will be pre-loaded with F1, F2, K10, and K20,
where F1 and F2 are the two one-way functions, K10 and K20 are the first keys on each key-chain.
2.1.2. Pair-wise Key Establishment
This step may happen in different forms depending on who are these pairs. We have pair-wise
key establishment and authentication that happens between nodes of the same type, cluster key
establishment and authentication that happens between two different types of sensor nodes, and
the global key establishment and authentication that happens in what called the distributed WSNs
that has only a manager node without the existence of the cluster head.
The Key-chain technique uses two types of key establishment, as a first step, it makes a pair-wise
key between the cluster head and sensor, but before doing it, the node that has the ability to play
the cluster head role will generate two random numbers N1, N2 and calculate the cluster key Kbrod,
as Kbrod = K1N1 ⊕ K2N2. The cluster head could do this calculation because it is pre-loaded with F1
and F2, the two one-way functions.

The scenario of forming the cluster head starts with broadcasting a Hello message by the cluster
head to the nearby sensors, this message includes the ID of the cluster head and will be called
message 1. Each sensor receive a Hello message will join the cluster of the cluster head that sends
a Hello message with the best signal noise ratio (SNR). After choosing the cluster head, sensor
sends message to the cluster head, this message contains the sensor ID, and the two random
numbers n1 and n2, and this message is message 2 in this scenario. When the cluster head receives
message 2, it calculates the Kinit for each sensor joined its cluster and find a new key called the
pair-wise key Kpair, where Kpair = K|N1-n1 |⊕ K|N2-n2|. This step comes to authenticate the sensor's
legitimacy. Then the cluster head generates a random number R1 for the next time
communication. After this the cluster head will send message 3 for the sensor, which contains an
encryption for the R1 value, Kpair, and Kbrod, this encryption will be done using the Kinit. So only
the legal sensor has the right Kinit, and could decrypt message 3 to obtain the information with it.
If the sensor is a legal one, it will get the values of R1, Kpair, and the cluster key Kbrod, and stores
those values. Then it will reply the cluster head with message 4, which has the encryption of R1
with Kpair. The cluster head must decrypt message 4, and checks R1, in case it matches the original
value, the cluster head would store the sensor ID and Kpair. Till now the cluster head and each
sensor belong to its cluster establish pair-wise key for future communication.
As a second step in the Pair-wise key establishment phase, the Key-chain approach uses another
type of pair-wise key establishment, the type that happens between two nodes of the same
category, i.e. between the clusters heads. Communications between the BS and the cluster heads
could be achieved by using the relaying strategy. All cluster heads send data to the BS via multi-hops
of other cluster heads. At beginning, the distant or the far away cluster head tries to join into
the close cluster, and that means it will be a child node for the cluster head of this cluster. After
the successfully joining, both cluster heads broadcast the random number N2, and this is message
1 for this scenario. Then both cluster heads can calculate the pair-wise keys use the following
equation, where N2
53
' is the other node random number.
'
Kpair = K2N2 ; N2 = N2
K2N2 ⊕ K2N2; N2 N2
'
At last, cluster head distribute the cluster key to child cluster head in message 2, after encryption
this information using the calculated Kpair in the previous step. By now, all the networks keys
establishment processes have been finished. In the future communication, the produced keys can
be used to encrypt sensitive data, so protecting the WSN from attacks.
2.1.3. Key Renovation
This step means having the ability to re-keying the sensors with new keys as a way to have an
intrusion detection mechanism to detect compromised nodes. So the key renovation and
revocation phase is an essential component in key management techniques.
Using the Key-chain technique that mentioned in [6], each sensor in the cluster has a unique pair-wise
key, if we don't consider the probability of more than two sensors pre-loaded with the same
n1, n2. If a node is compromised, we only need to delete the related pair-wise key in its cluster
head. This will not affect the other nodes and links. As soon as a node is compromised or the key
period expired, the cluster head will renew the cluster key. Then cluster head generate another
two random numbers N1’, N2’, calculate the new cluster key Kbrod’, and distribute it to the cluster
numbers encrypting with each pair-wise key. To reduce the communication costs, a piece of
message can include several nodes key renovation information. The format of the message is:

International Journal of Network Security Its Applications (IJNSA) Vol.6, No.6, November 2014
54
id1 || id2 || id3…|| EKpair_1 (Kbrod
') || EKpair_2 (Kbrod
') || EKpair_3 (Kbrod
') …
Where, id1, id2...are sensors IDs; Kpair_1, Kpair_2…are sensors pair-wise keys. Renovation
message is broadcasted to all sensors in the cluster. The corresponding sensors receive the
message and get the new cluster key. To keep the security of pair-wise key, cluster head must
renew the pair-wise key periodically. The process is similar to cluster key renovation.
2.2. The Second Technique that Mentioned in [7]
In this approach, they depended on the idea of a group key management, where the group key is
used to secure the communication for each group in the WSN, and it is usually created and
updated by a central key server (sink node), and then securely distributed to all members in the
same group.
Working under this scheme requires some assumptions that need to be taken, such as that the
sensor nodes in the network are uniquely identified, i.e. has a unique ID number i. Each node is
innocent before deployment, and cannot be compromised during the first few minutes following
deployment. Compromised nodes shall be revoked by the sink. Each node can use the watchdog
mechanism [8] to identify the compromised nodes. Sensor nodes are grouping into clusters. Each
cluster has a leader, referred to as a cluster head (CH). CHs are elected by specific algorithms [9]
and form a second tier network. Each CH stores the identities of other CHs after the CHs
elections. In this work, they used the threshold cryptography [10], which is used for distribution
of trust in key management and a (n, k) threshold scheme allows n parties to perform
cryptographic operations, so that any k parties can jointly perform key discovery whereas (k-1)
parties cannot devise any information, even after collusion.
The sink node is the node who decides how many groups (clusters) can be formed in the network,
each group will take a unique 2t degree bivariate polynomial, and since the count starts from zero,
the maximum degree will be equal to 2t–1, as in the following equation, g(x, y) =
is constructed over a prime finite field, where t is related to the number of
nodes in each group, and y is a variable that represent the current group key version which started
with zero value when the node is deployed and increased by one each time the key is updated.
Accordingly, each sensor node u gets its own secret g(u, y), as illustrated in Fig.1, node V1 … V6
and CH get their own secrets g(V1, y) … g(V6, y) and g(CH, y).
Figure 1. Group Key Pre-distribution [7]

55
2.2.2. Pair-wise Key Establishment
As mentioned before, the pair-wise key establishment could happened in many forms, depending
on the type of the pair nodes. In this approach, the communication is happened between the CH
and the member nodes in the cluster, so to secure this communication we need this key, which is
called the hierarchical key. To generate this key the CH uses a one-way hash function and the ID
of the member node that will communicate with the CH to get H(ID1) …. H(IDn), as an output of
the hash function, where n is the number of member nodes in the cluster. Then the CH will send
this key to the corresponding node through unicast traffic, and it saves all the generated keys in
the hierarchical key table in its memory. As a result each member node will have its own secret
key, and the CH knows all the hierarchical keys that related to all the member nodes in its cluster.
2.2.2.1. Group Key Recovery
This step represents the reforming of the group key that was given in the pre-distribution phase.
This process is done by the CHs. So to get the current group key, the CH shall look for help from
t member nodes, by broadcasting the help request to all member nodes in its cluster. In its turn,
each member node that receives such request and is willing to trust its cluster leader will return its
own secret share, g(u, y). As illustrated in fig.2.
As a result, the CH gets the shares from each member node in its cluster, but there is a possibility
to receive incorrect shares from compromised nodes, so if the CH succeeds in getting to correct
personal secrets, it can obtain the current group key by using the threshold secret sharing
approach, which divide data (D), i.e. the group key here, into n pieces, D1, …, Dn, in such a way
that:
• Knowledge of any k or more Di pieces makes D easily computable.
• Knowledge of any k-1 or less Di pieces leaves D completely undetermined (in the
sense that all its possible values are equally likely).
The CH can get the group key by estimating (t+1) points using the same polynomial that used in
the pre-distribution phase, the CH can evaluate g(0, y) and therefore obtains the current group key
Kg = g(0, y).
After calculating the new Kg, the CH needs to send this key to all the member nodes in a secure
manner, so it uses the hierarchy key for each member node to encrypt the Kg and send it to each
member node.
Figure 2. Group Key Reconstruction [7]

56
In this scheme, this step means having the ability to re-keying the group key if an adversary
captured a node, cause the adversary will be able to steal its key material, therefore the group key
needs the update process that is done by following the next steps:
• If a node (A) becomes a compromised node, this approach supposes that the neighbor
node (B) is the node who detects the new state for (A), and (B) after detecting this
case needs to do a check to see whether (A) is a CH. If yes, (B) sends the identity of
(A) to the sink node directly. In its turn, the sink node will inform the nodes who are
members at the compromised cluster to do a new CH election. Otherwise, the sink
node discards the message. On the other hand, if (B) detects that (A) is a normal
node, the identity is send by (B) to the CH, and if the message convinced the CH that
(A) became a compromised node and the CH authenticated the message, it erases the
hierarchy key that belongs to the node (A) from its memory, then complete the
procedure of the re-keying process. Otherwise, it ignores the message from (B).
• The involved CH sends a group re-keying message with timer T to other CHs in the
group, where T is for synchronizing group re-keying. In the first place, the other CHs
need to check the legitimacy of the message, if it is valid, the system will go back to
do the group key recovery operation again, and then follow the re-keying process
steps. Otherwise, ignoring the message.
• As a result to the previous steps, the CHs have generated a new version of the group
key, the published this key to all the member nodes after encrypted it with the
hierarchy key related to each member node, but this distribution for the new Kg does
not include the compromised nodes, thus the CH already erase their corresponding
hierarchy keys.
2.2.3.1. Node Join and Leave
This step is considering as a form or a type of the key renovation phase, the node joins means the
adding of new node to a specific group, and while this node will need the key materials used in
this network, so this scheme supported its needs by letting the sink node assigns the new node (U)
into a specified group and loaded it with the current group key Kg, and the personal secret g(u, y),
as well. So when the new is deployed into the network, it broadcasts its ID with a join requesting
message to the neighbors, with the help of the sink node, the CHs within the communication
range check the validity of the message, if legal, all the involved CHs replied the new node with a
message that has its cluster ID. The new node studies the offers, and picks the offer that related to
the nearest CH to save its energy resources, so the winner CH establishes a hierarchical key with
the new node. After that, the new node is ready to function in the network just as the already
existed once.
The nodes leaving means that the network lost some of its nodes, and that is happened for many
reasons, such as, the enervation of the sensor power, or losing the controlling on some node
because they become compromised nodes by the adversary. If the leaving node was a normal
node, i.e. not CH, its CH should erase its hierarchical key from its memory, and perform the
group keying process again as mentioned before. But if the leaving node was a CH, nodes in its
cluster must do an election process to select a new honest CH, and that happened with the help of
the sink node, then the new CH will repeat the phases of calculating new hierarchical keys for its
members, and updating the Kg.

57
2.2.3.2. Node Isolation
This is needed, because this approach depends on its nodes to function well, so if a large number
of nodes become compromised, the CH may not be able to obtain the new group key due to the
lack of assistant from the member nodes. As a result, all the compromised nodes in this cluster
must be isolated. In this scheme, the CH who fails in reforming the group key sends a cluster
dissolving message to the sink node and to the innocent nodes that remains in his cluster, the
purpose of this message is to let them know that he is willing to quit, and his cluster is won't be
exist anymore, it erases the hierarchy key table, and this CH will also help the innocent nodes in
his cluster to join the a new cluster, it did this by sending a joining message to the other CHs in
the group, this message has the identities for all the innocent nodes in the old cluster. The CHs
check the validity of this message, if so, the replied the CH with a joining accept message. The
CH studies all the received offers and picks the one from the nearest CH, then informs all the
innocent nodes to join. After the successful join for the nodes, the new CH for those nodes will
establishes hierarchical keys with them, and then doing the group key renewing again.
2.3. The Third Technique that Mentioned in [11]
This scheme tries to find a key management approach with less energy consumption, it based on
dividing the WSN into three-level models as illustrated in Fig.3. The first level, the network is
organized with Normal Nodes (NN), the second level with Cluster Heads (CH) and on the top
Figure 3. Network Model [11]
level is the Base Station (BS).
As the figure shows, the communications in this network happens between the normal nodes and
the corresponding cluster head, between the neighbors CHs, and between the CHs and the BS.
The BS has unlimited resources capabilities and strong computational power, the CHs also
powerful devices but not as the capabilities for the BS, and the normal node is the one with
limited resources and low memory space.

58
The BS is the node that administrates this network, and to ease its task, each sensor in the network
is given a unique ID which is equals to the column number in the generated matrix and points to
the value of that column, which will be saved in the sensor. All the nodes in this network model,
including the BS, are pre-loaded with the following uniform rule, and if needed, more than one
uniform rule will be used: ai = ai-1 ⊕ (ai ai+1).
Where ai is the ith bit of a binary string, ai+1 is the right bit of ai and ai-1 is the left bit of ai. The BS
also set with an initial state for a binary string, and a variable called KM that represents the
number of iteration that will be applied using the uniform rule to generate a key.
2.3.2. Pair-Wise Key Establishment
This is to obtain the keys that are needed for the three types of communications in this network
that mentioned before. First the initialization process for the network must take place. The BS
will generate a matrix Ckp at random in a WSN with N nodes and it is equal to the number of
columns in the generated matrix, as well, where p is the number of generated keys, and k is the
length of the needed keys, while P-N represents the maximum number of new sensors that could
be added to this network. As we said, this phase will include the following steps:
2.3.2.1. Key Establishment between the BS and the CH
After finishing the computing for the BS and the CH, this process begins, with the following
sequence:
• Clusters need to be formed, and then the BS broadcasts a message to all CHs in the
network, informing them to start the key establishment process.
• Cluster heads send their IDs to the BS.
• The BS will use the IDs values it receives to refer to the matrix Ckp and obtains the
column value, since the IDs represented the columns numbers in this matrix, then the
BS will use the value of KM and the uniform rule to compute the pair-wise key.
• As a reply from the BS, it sends the value of KM and the list of the generated IDs to
the corresponding CHs.
• Each CH checks out if its ID included in the list or not.
• CHs calculate their keys upon the column value they have and the KM they got from
the BS, and save the keys.
2.3.2.2. Key Establishment between CH and another CH
The cluster heads already have a binary string that represented an initial state that used to generate
the keys that are needed to be calculated before; the steps of this process are as follow:
• Thus each CH has a column value that pre-loaded on it; neighbor cluster heads sends
its column value for each other’s, so every CH will has its column value Ci its
neighbor column value Cj.
• Both CHs that did the previous exchange calculate Ci Cj accordingly, and the output
of this calculation is used as an initial state values for the uniform rule.
• CHs calculate respectively depending on the uniform rule and the KM value as the
number of iterations, and then save the new keys.

59
2.3.2.3. Key Establishment between Cluster Head and Normal Sensor Nodes
This step is little bit similar to the one happened between CHs and the BS, but here it needs to
work in two stages, one between the normal node the CH, and the second one between the CH
and the BS, as following:
• Normal sensor nodes within a cluster send their ID to the corresponding CH.
• In its turn, the CH at this point just passes the list of received IDs to the BS.
• The BS queries the matrix Ckp, and then sends the column value according to the IDs
list and attaching the value of KM with it to the CH.
• Upon receiving the previous information from the BS, the CH does its calculation,
and ate the same time, broadcasts KM and the list of IDs to the sensor nodes within
the cluster.
• Normal sensor nodes check out to see if their IDs are in the received list, and also it
computes using KM and the uniform rule its key, and save the result.
As mentioned before, this step is to rekeying and updating the whole keys in the network, and in
this approach this is done when a CH is captured by an adversary. This scheme assumes that the
BS station is the one who senses that the CH was captured. So when the BS detects the capture of
a cluster head, it must notify the other CHs in the network to refresh the keys. And the BS will
generate a new random number as a value for KM that represented the number of iterations
applied to the uniform rule. So the whole key establishment's stages will be repeated, just by
using the new value of KM and the same binary string that was assigned to each sensor.
2.3.3.1. Sensors Addition and Deletion
As usual, this step is considered as a form or part of the key renovation phase, and just as when
the network nodes are deployed for the first time, the new node that will to join the network will
be pre-loaded with the same needed information, such as the node ID that represented the column
number in the matrix Ckp, the column value, and a uniform rule. When the node is deployed into
the network, it firstly broadcasts its ID; the nearest CH that hears this ID will pass it to the BS,
which checks it. After this, the process of key establishment is begun and it is just as the one
happened between the CH and the normal nodes.
Nodes deletion in this network model happens for the normal nodes only, because it has less
energy level comparing to the CHs nodes. So when the power of the normal node is
approximately depleted, it informs its corresponding cluster head which in its turns will erase the
ID and the key related to this node.
3. ANALYSIS FOR THE AFOREMENTIONED TECHNIQUES
The Analysing for the previous approaches will be discussed by summarizing the advantages and
the disadvantages for each scheme, and by trying to find out some recovery ideas for the
weakness point in those techniques from our point of view.
• First Technique
This approach uses two one-way hash functions in generating two key chains, this will prevent
the adversary from getting a successful trial to discover the key materials, but the expected

problem is with pre-loaded the CHs in this network with the two hash functions and the beginning
key for each chain, and since this approach didn't present any solutions for facing the problem
when the CH become a compromised node, and if such thing happened, it will be easy for the
attacker to control the network that depends on this approach, although the approach assumes that
the CH node is equipped with a tamper resistance hardware to protect it. This is not good enough
because the tamper resistance is not absolute; an opponent with access to semiconductor test
equipment can retrieve key material from a smart card chip by direct observation and
manipulation of the chip's components, [10] add to the previous, the cost that it takes in equipped
all the CH nodes with this hardware.
Another expected problem with this technique, is the way it used to form the clusters, it uses the
SNR mechanism, and this one works by let the CHs send a Hello message for all of the nearby
nodes, and the nodes will attend the cluster that belongs to the CH that sends a Hello message
with the highest SNR value, this gives the adversary a great chance to attack such approach cause
he could put a spoofed CH that propagate a Hello message that has a greater SNR value than the
one related to the honest CH, thus ruin the network and let the attacker control the mechanism.
One more thing to the previous comments, the approach, in our opinion, may need another step to
be done in the pre-loaded phase, which is using a DB for the random numbers that are generated
for the nodes, this is to avoid the probability of having one or more sensors pre-loaded with the
same random numbers, it also will help in making a flag for these numbers to let the admin
distinguish between the ones for normal nodes (denoted by small n in the original paper) and the
ones related for CHs (denoted by capital N in the original paper), and this may be needed,
because when the CH try to join another cluster, it broadcasts one of its random numbers in the
network, so the normal node will receive this extra message and process it. But if we separate the
ranges for both kind of random numbers and used a multicast traffic instead of the broadcast, this
may decrease the overhead on the network and contributes in saving the sensor power, thus
improve the efficiency of this technique. Another idea to do that is to apply the approach of
gateway CHs to achieve the purpose here.
60
• Second Technique
This scheme depends on the idea of hashing the sensor IDs to generate what called the hierarchy
key that used between the normal node and the CH. The expected problem is that when sending
this key for the normal node, the CH sends H(ID) to each sensor, and this is subject to the
attackers attack, because they may steal this key and makes use of it, especially it is the one that
used for encrypting another sensitive info which is the group key. As a trial to solve this problem,
the scheme could apply the MAC technique for this communication to protect such info.
Another issue with this approach, that it sends a lot of unicast traffic that make an extra overhead
on the network, besides to the big assumptions it supposed for the re-keying phase which
consumes many of the sensors resources.
• Third Technique
This scheme assumes that the BS has unlimited resources capabilities and strong computational
power, thus most of the materials that are needed for generating the keys are located in the BS
and tied to it.
This approach doesn't allow the nodes, regardless to its category type, to exchange the secret
keys; it just exchanged the IDs. These IDs points to binary strings which are used to produce the
needed keys after applying the uniform rule KM times on it, but it doesn't handle the situation
when the ID that related to a sensor doesn't existed in the received list, this will leave such sensors

without a key, so it won't be able to send its data in a secure manner or it may use the pre-loaded
binary string as a key, and this is subject to be stolen by the attacker.
As a proposed solution for such issues, the next section has modifications for the previous
scheme, as a trial for a proposed scheme.
61
4. PROPOSED IDEA AND FUTURE WORK
Here we are trying to find a compromised technique that solves some of the expected problems in
the previous techniques. Our technique will use same matrix that mentioned in [10], shown in
table 1, but with making some modifications to improve the technique. We will use the idea of
pre-loaded the sensor with an initial binary string that is used in forming the clusters in the WSN.
Table 1. Matrix that located in the BS
ID1=1 ID2=2 ID3=3 ID4=4 ID5=5 … IDn=P
1 0 1 1 … … 1
0 1 1 1 … … 0
1 1 1 0 … … 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
k k k k k k k
4.1. Cluster Forming
• Divide the nodes into two halves, even half and odd one.
• Divide the target area (field where the sensors will be deployed), into two halves,
area O (odd), and area E (even). So when disseminate the sensors, we need to deploy
the even half in E, and the odd half in O.
• The CH with an odd ID (3, 5, 7...), will allow for the nodes that has a binary string
with an odd parity to join its cluster.
• The CH with an even ID (2, 4, 6...), will allow for the nodes that has a binary string
with an even parity to join its cluster.
4.2. Keying the Network
• CHs send their IDs to the BS, the BS applying the uniform rule on the column values
related to these IDs with number of times equal to the number of ones in the binary
string, so no need for generating KM.
• When the CHs receive back the list after the BS computation, if its ID doesn't exist, it
generate its own key by applying the rule with a number of rounds equal to the
number of ZEROS in its binary string, then sends a message to the BS in a form of:
(ID,0), to let the BS do the same calculation to get the same result for the key value,
and if the binary string doesn't have 0s, the CH depends on the number of 1s as the
number of rounds, and also sends the message (ID,1) to the BS to do the same.
• Same approach between normal nodes and CH.
Our proposed idea similar to the scheme in [11], that both of us use the binary matrix idea. On the
other hand, we differ in preloading the normal sensors with an initial binary string that is used in

forming the clusters in the WSN, and this will help in overcoming the expected problem in [11].
Add on that, we use different technique in dealing with the target terrain that the sensors will be
deployed in, and different approach in forming the network clusters and keying the network as
explained before.
In the near future we are planning to implement our proposed idea, and conduct some
experiments as a trial for improving our point of view, and do some sense comparisons between
our proposed techniques and the aforementioned one.
62
5. ANALYSIS FOR THE PROPOSED IDEA
This approach won't need much cost cause of the efforts that needed to classify the sensors into
two groups, and in dividing the network area comparing to predefined location techniques such as
the one mentioned in [11], thus no need to the pre-knowledge for each sensor location, all we
need to know which part of the network field will has the odd sensors and which one will hold the
even ones, thus the group location not individual coordinates for all sensors.
Another point, related to the security part of this approach that if the attacker tried to access the
network, by adding a new sensor, he won't know in which part he could add his node, and if he
successes by chance, he won't know the rules this technique depends on in making the keying
process, which is sending the number of one's or number of zeros, even though he steal the
exchanged management messages, he won't be able to get the complete picture of the approach.
6. CONCLUSION
The previous paper focused on a very significant security issue, which is the Key management in
WSN. This problem has attracted the researcher's attention and makes them find out many
techniques to solve it out. We explained three techniques for key management problem in WSN,
and the work went in classifying those techniques upon three phases. Phase one, is the pre-loading
stage, then the pair-wise key process, and finally the key renovation phase. Then we made
some analyzing for those mechanisms by highlighting their advantages and disadvantages. After
that, we presented some modification as a proposed idea to map new key management technique
with less overhead by making use of pre-loaded information for security purposes in forming
secure clusters and decrease some work efforts such as generating random number to get the
number of needed iterations that would applied into uniform rule, and good robustness in facing
WSN attacks. Hence, this paper is important to highlight the main steps for the key management
process in WSN, and helps in simplifying the idea behind it.
REFERENCES
[1] Mohammed A. Abuhelaleh, and Khaled M. Elleithy, (2010), “SECURITY IN WIRELESS
SENSOR NETWORKS: KEY MANAGEMENT MODULE IN SOOAWSN”, International
Journal of Network Security Its Applications (IJNSA), Vol.2, No.4, pp. 67-78.
[2] Yang Xiao, Venkata Krishna Rayi, Bo Sun, Xiaojiang Du, Fei Hu, and Michael Galloway, (2007),
A survey of key management schemes in wireless sensor networks. Science direct, Computer
Communications, pp. 2314–2341.
[3] http://nesl.ee.ucla.edu/tutorials/mobicom02/slides/Mobicom-Tutorial-2-MS.pdf (2014), Viewed
online on 16th Nov 2014.
[4] Chi-Yuan Chen, and Han-Chieh Chao, (2011), A survey of key distribution in wireless sensor
networks, Published online in Wiley Online Library (wileyonlinelibrary.com). DOI:
10.1002/sec.354.

[5] Ritu Sharma, Yogesh Chaba, and Yudhvir Singh, (2010), An IPC Key Management Scheme for
Wireless Sensor Network, 2010 1st International Conference on Parallel, Distributed and Grid
Computing (PDGC - 2010), pp. 251-255.
[6] Guohua Ou1, Jie Huang, and Juan Li, (2010), A Key-Chain Based Key Management Scheme for
63
Heterogeneous Sensor Network, pp. 358-361.
[7] Yuan Zhang, Yongluo Shen, and SangKeun Lee, (2010), A Cluster-Based Group Key
Management Scheme forWireless Sensor Networks, 2010 12th International Asia-Pacific Web
Conference, pp. 386-388.
[8] S.Marti, T.Giuli, K.Lai, and M.Baker, (2000), Mitigating routing misbehavior in mobile ad hoc
networks”, Proc. ACM International Conference on Mobile Computing and Networking, pp. 255–
265.
[9] K.Kim, Y.Zhang, W.Yang, and M.Park, (2008), An authentication protocol for hierarchy-based
wireless sensor networks, in Proc. IEEE International Symposium on Computer and Information
Sciences, pp. 1–6.
[10] A. Shamir, (1979 ), How to share a secret,Communications of the ACM, pp. 612–613.
[11] Lanying Li, and Xin Wang, (2010), A high security dynamic secret key management scheme for
Wireless Sensor Networks, Third International Symposium on Intelligent Information
Technology and Security Informatics, pp. 507-510.
Authors
Khawla Naji Shnaikat: has a B.Sc. degree in Computer information Systems/ specialist in computer
networks with 3.25 GPA, and the M.Sc. degree in Computer Science with 3.94 GPA from University of
Jordan. Find more on: Linked in.
Ayman Ahmed AlQudah: has a B.Sc. in software engineering, July 2008 from Hashemite University,
Jordan and has an M.Sc. in Information Technology Management (ITM), July 2012, from the University
of Sunderland, UK. Find more on: Linked in.

Key management techniques in wireless sensor networks

More Related Content

What's hot (18)

Viewers also liked (15)

Similar to Key management techniques in wireless sensor networks (20)

Recently uploaded (20)

Key management techniques in wireless sensor networks