Securing range free localization against wormhole attack using distance estimation and maximum likelihood estimation in Wireless Sensor Networks

Author’s Accepted Manuscript Securing Range Free Localization against Wormhole Attack using Distance Estimation and Maximum Likelihood Estimation in Wireless Sensor Networks Gulshan Kumar, Mritunjay Kumar Rai, Rahul Saha www.elsevier.com/locate/jnca

PII: DOI: Reference:

S1084-8045(17)30320-X https://doi.org/10.1016/j.jnca.2017.10.006 YJNCA1986

To appear in: Journal of Network and Computer Applications Received date: 28 January 2017 Revised date: 5 August 2017 Accepted date: 3 October 2017 Cite this article as: Gulshan Kumar, Mritunjay Kumar Rai and Rahul Saha, Securing Range Free Localization against Wormhole Attack using Distance Estimation and Maximum Likelihood Estimation in Wireless Sensor Networks, Journal of Network and Computer Applications, https://doi.org/10.1016/j.jnca.2017.10.006 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Securing Range Free Localization against Wormhole Attack using Distance Estimation and Maximum Likelihood Estimation in Wireless Sensor Networks Gulshan Kumara , Mritunjay Kumar Raib,∗, Rahul Sahac a Assistant

Professor, School of Computer Science and Engineering, Lovely Professional University, India b Associate Professor, School of Electronics and Communication Engineering, Lovely Professional University, India c Assistant Professor, School of Computer Applications, Lovely Professional University, India

Abstract Localization has always been considered to be an important factor in Wireless Sensor Networks (WSNs). Along with accuracy of the location estimation, the security of the location information is a critical issue in localization process. Moreover, as the network environment changes from static to mobile, the probability of the wormhole attack increases. Previous research suggests possible solutions but lag behind to find the applicability in the mobile environment and some of the algorithms are not suited for resource constrained WSNs. Therefore, in this paper we have developed a localization algorithm that prevents wormhole attack in mobile environment. The algorithm uses authentication process to identify any unauthorized nodes using distance estimation method and applies Maximum Likelihood Estimation (MLE) to calculate the required location. The comparison of our algorithm with other contemporary algorithms proves that this algorithm performs efficiently. Keywords: Anchor Nodes, Security, Location, Authentication, Wormhole, Certificate

1. Introduction The proliferation need of monitoring and controlling in the Wireless Sensor Networks (WSNs) has extended the applications [1][2][3] of sensor networks from static to mobile environment. Dynamic networks also create a concern on the changeable locations of the nodes. Location estimation in such mobile environments attracts various attackers to execute their attack procedures and ∗ Corresponding

author Email addresses: [email protected] (Gulshan Kumar), [email protected] (Mritunjay Kumar Rai), [email protected] (Rahul Saha)

Preprint submitted to Journal of LATEX Templates

October 4, 2017

manipulate the location information. Therefore, along with the accuracy of the location estimation, the security [4] of location information is also an important factor in WSNs. Localization errors can be generated due to malicious and nonmalicious reasons; however it affects the performance of the location estimation process. Security attacks in WSNs can be executed by the internal nodes as well as external nodes. Compromised insider nodes misguide about their original location and external nodes execute various forms of intrusion with the location determination process. Therefore, the localization techniques must be secured to avoid or prevent such attacks. The security requirements for secured localization techniques must provide confidentiality of the estimated location along with the authentication for identity of the nodes that participate in the network. Further, information availability to compute proper location is also required for a secured localization process. A number of security attacks [5] have been found in WSNs. Among them, wormhole attack [6][7] is one of the important consideration. This attack creates a high speed link between two radio transceivers so that all the packets can follow that path of routing. Variations of this attack are available such as store-and-forward, bit-by-bit and many more. All such attacks are immune to cryptographic mechanisms and therefore the attackers need not provide extra effort to break in the codes. A number of secured localization algorithms have been introduced so far. For example, Distance bounding protocol [8][9] defines the upper bound on the distance between two communicating nodes. Distance bounding protocols engage the authentication which makes it more efficient for location estimation. Based on this distance bounding approach another secure localization algorithm is shown in [10]. The algorithm called Secure Positioning In sensor NEtworks (SPINE) is a centralized algorithm and uses Verifiable Multilateration [11]. The process is executed in two phases: (i) the sensors measure distance bounds to their neighbors and (ii) the distance bounds are then collected at a central authority and the positions of the nodes are computed. Another algorithm RObust Positioning Estimation(ROPE) uses decentralized approach for robust location computation and verification[12]. Maximum Spoofing Impact (MSI) is introduced in this approach which is defined as the maximum distance between the actual location of the node under attack and any possible spoofed location. A range free secured localization algorithm Secure Range independent Localization (SeRloc) [13] uses sensing nodes and locator nodes for the location estimation. Wormhole attacks are thwarted in SeRloc by its sector uniqueness property and communication range violation property. Similar usage of sensing nodes and locators is also seen in High Resolution Localization (HiRloc) [14]. The sensor nodes rely on beacon information transmitted from the locators. Based on the beacon information, sensors define the sector area as the confined area covered by the transmission of a locator. By collecting beacons from the locators the sensor node computes its location as the Region of Intersection (ROI) of all the sectors the sensor node included in. Since the 2

ROI indicates the confined region where the sensor node is located, reducing the size of the ROI leads to an increased accuracy in the localization estimation. To prevent wormhole inclusion, a single hop based approach is also introduced in Distributed Reputation-based Beacon Trust System (DRBTS) Algorithm [15]. Apart from these traditional approaches, some contemporary secured localization processes have been studied by various researchers. Our contribution in this paper emphasizes on the development of a secure localization algorithm that can withstand with this type of attacks and can provide estimation of location in a hostile environment with higher accuracy. The main contributions of this paper are summarized as follows: • We have proposed a wormhole attack resistant scheme for each anchor node to identify their authenticated hop bound unknown neighbours. • We have utilized the resource privileged anchor nodes and base station with their computational ability to increase the localization accuracy of the unknown nodes. • We have conducted the simulation based comparison to validate the efficiency of our proposed secure localization scheme. The rest of this paper is organized as follows. Section 2 reviews the related work on secure localization against wormhole attacks. In Section 3, we have described the network model and the proposed algorithm. Section 4, presents the performance evaluation. Finally, Section 5 concludes this paper. 2. Related Work We have done the literature survey in two categories. Firstly, we have reviewed the existing approaches to protect the localization process against wormhole attack. Secondly, we have reviewed the generalized wormhole attack prevention methods applied in different domains. A robust localization approach against wormhole attack in WSNs has been shown in the work of [16]. The authors have provided a solution called as ConSetLoc for wormhole attack for range-free localization process. They have used the relationship of hop counts and geographical distances of sensor nodes and have designed a partition method of stable anchors sets along with the convex constraints in geometry. This process reduces the effect of false measurement. The authors have shown a filtering strategy for the candidate locations with anchors sets. Though the strategy provides calculation precision of the location, but creates an extensive calculation for filtering process that may not be feasible for a resource constraint WSNs environment. A distance consistency based secure localization process has been identified in [17]. The algorithm consists of three phases: detection of wormholes, valid locator identification and self-localization. The process has been simulated using

3

100

the Received Signal Strength Indicator (RSSI) approach. The work is not validation for time measurement approaches as lower bound and upper bound of time measures has not been prescribed. A label based approach against wormhole attack in DV-Hop localization has been researched in [18]. The algorithm works in three phases. Firstly, the beacon nodes are distinguished and labelled according to their geographic relationships considering the presence of a wormhole. Secondly, the sensor nodes are further distinguished and labelled according to neighbouring beacon nodes. After avoiding the wormhole links, DV-Hop localization is applied on the labelled neighbouring nodes. The labelling process is feasible for static network environment but cannot be applied to a dynamic environment of WSNs because the frequent change of labelling will increase overhead. A proactive countermeasure against wormhole attacks in localization has been shown in [19]. The authors have developed a novel approach Wormhole Free DV-Hop (WFDV) that includes infection prevention process before the DV-Hop localization starts. The prevention process emphasizes on constructing the valid neighbour list and removing the false neighbours. The neighbour list creation process is not feasible and validated for dynamic environment of the network. Apart from the centralized processes, distributed approaches also have been researched. Such an approach called WAPN has been shown in [20]. The authors have used probability distribution (Bernoulli) of the neighbouring-node-number and threshold value to detect a wormhole attack. Another distributed wormhole detection approach for WSNs has been observed in [21]. The proposed algorithm detects wormhole links based on the deviations created in the network. Moreover, it uses a hop counting technique, reconstructs local maps for each node, and then uses a diameter feature to detect errors caused by the wormholes. Both the algorithms do not confirm the complexity level according to the scalability factor of the network and are not verified against distance reduction or enlargement attacks. A wormhole detection and prevention method has been shown in the work of [22]. The proposed method uses route trip time parameter and threshold value to detect the wormhole link. Though these methods detect the wormhole links, the data availability is not ensured. A Finite State Machine (FSM) based approach for wormhole detection is shown in [23]. Channel reciprocity is used in the work by [24]. The proposed algorithm called Secure Channel REciprocitybased WormholE Detection (SCREWED) avoids both false positives and false negatives of wormhole detection. Channel hopping, randomized transmission powers, message integrity codes, as well as a special replay protection mechanism are used in this algorithm. A weighted approach to detect and prevent the wormhole attack is shown in the work by [25]. The concept of Local Most Trustable (LMT) node is used in the algorithm that not only prevents the wormhole attack, but also identifies the wormhole nodes. The election of LMT node is a challenging task to protect the authenticity of this node will disrupt the normal functionalities of the proposed work. A secure neighbour discovery process in presence of wormhole attack has been discussed in the papers [26][27]. A cluster based approach as 4

countermeasure against wormhole attack has been shown by [28]. The proposed algorithm Control Traffic tunneling Attack (CTAC) is applicable for both static and mobile network environment. The selection of a cluster-head is a critical issue in the algorithm which is not addressed in a deterministic way. Local connectivity based approaches are discussed in [7][29]. The proposed algorithms are distributed and well suited for static networks because the analysis of local connectivity in a heterogenic network with large number of nodes is troublesome and will be more complex. 3. Proposed Network Model The objective of the proposed algorithm is to provide a preventive mechanism against wormhole attack. The foremost process in this regard is to attain a strong authentication process that can support one hop as well as multihop routing mechanism. The algorithm starts with an initialization phase that deals with distribution of certificates by the Base Station (BS). After the distribution of the certificates, mutual authentication is executed. Once the nodes are proved to be legitimate, the distance estimation phase starts among the anchor nodes and the unknown nodes. After the estimation of distances, the Base Station (BS) is able to localize the unknown nodes applying Maximum Likelihood Estimation (MLE) method. The algorithm is summarized in Algorithm 1. The proposed algorithm also supports mobility of the nodes (anchors as well as unknown nodes). 3.1. Network Model In the proposed algorithm, we have categorized the nodes into three categories. Firstly, the base station (BS) controls the overall network environment. Secondly, the anchor nodes aj ∈ A have the capability of complex computation with privileged energy resources. Finally, the resource constrained unknown nodes ui ∈ U whose locations are to be estimated. The anchor nodes and the unknown nodes are considered to be mobile and randomly deployed in the network. The anchors are having a variable range of transmission with an average transmission range Ravg given as:  min e∈E ψ(|e|) Ravg = (1) m Where m is the number of anchor nodes in the network, ψ(|e|) is the weighing function of a connection between an anchor node and an unknown node and interpreted as: ψ(|e|) ∼ |e|α where 2 ≤ α ≤ 4

(2)

The network model is having no central control of deploying the sensor nodes in the network. For the ease of representation, the wireless sensor network model is considered as two dimensional and represented by a graph G(V, E) where,V

5

Table 1: List of Symbols and Notations.

Description

Notation/Symbol

Average transmission range Weighing function Base Station Anchor node Unknown node Certificate of node (Node can be anchor node aj and/or unknown node ui ) IP address of node (Node can be anchor node aj and/or unknown node ui ) Public key of node (Node can be anchor node aj and/or unknown node ui ) Timestamp when the certificate was created Expiry time of the certificate Private key of the base station Random nonce Private key of anchor node aj Private key of unknown node ui System error Error generated by the malicious node Public key of base station Distance vector of unknown node Relative mobility

Ravg ψ(|e|) BS aj ui CertN

Equation no. 1 12 4 5 6 14 5 7 8 14 68 4 5 6 7 8 14

IPN

456

KN +

456

t et KBS− N Kaj− Kui− δ  KBS+ b ui RMti,j

456 456 456 78 7 8 11 13 11 12 14 15 16

is a set of vertices and E is a set of edges. The size of the network can be given as: |V | = |A| + |U | (3) Where, |A| is the size of anchor node set A, |U | is the size of the unknown node set U and A, U ⊆ V . The assumptions for our proposed approach have been listed below. • The unknown nodes and anchor nodes are mobile. • Base Station is assumed to be trusted and is considered to be key distributor and certificate authority. • Anchor nodes and unknown nodes are deployed with their pairs of private and public key predetermined. • Base Station shares the public key only to the legitimate unknown nodes and anchor nodes predefined.

6

Initialization Phase Keys are basically pre-generated among the BS, anchor nodes and unknown nodes in the topology. Each of such nodes receives exactly one certificate CertN , after getting authenticated itself, as follows: BS → CertN = [IPN , KN +, t, et ]KBS−

(4)

Where, IPN is the IP address of a particular node, KN + is the public key of that node, t is the timestamp when the certificate was created and et is the expiry time of the certificate. This total certificate is digitally signed by KBS− , the private key of the BS working as certificate server. All nodes must update themselves with a fresh certificate. For an anchor node aj and an unknown node ui we can write the above format in the following way. BS → Certaj = [IPaj , Kaj+ , t, et ]KBS−

(5)

BS → Certui = [IPui , Kui+ , t, et ]KBS−

(6)

Distance estimation phase After this certificate generation, the anchor node initiates a neighbour discovery process and provides end-to-end authentication. The anchor node aj begins this neighbour discovery by broadcasting a neighbour discovery packet (NDP) to its one-hop neighbors along with its certificate and puts on the timer. The broadcast message is sent in the following format: aj broadcasts N DP : [N ]Kaj− , Certaj

(7)

where N is the random nonce created by the anchor node and encrypted by the private key Kaj− of the anchor node aj . The nonce is used to uniquely identify that the packet is coming from a legitimate anchor node. The receiving unknown node ui uses sender’s public key extracted from the certificate to validate the signature and also verifies if the certificate is expired or not. The receiving unknown node signs the nonce sent by the anchor node, appends its own certificate, and replies back to the anchor node. The signature prevents spoofing attacks that may alter the route or form loops. This process of mutual authentication will help to prevent any kind of outsider attacks as the keys are pre-distributed by the BS. ui → aj : [N ]Kui− , Certui

(8)

where, Kui− is the private key of the unknown node ui and Certui is the certificate of the unknown node ui . When the anchor node receives the message from the unknown node, it stops the timer and calculates the Round Trip Time (RTT). Once the RTT is calculated, the approximated distance dij between the anchor node aj and unknown node ui can be estimated as: RT T = timerstop − timerstart 7

(9)



dij = c × RT T , where c is the speed of light

(10)

This calculated distance is further adjusted according to the system error and or any malicious error as:  d ij = dij × (1 ± δ) × (1 ± ), ∀ i = 1, 2, ..., n and j = 1, 2, ..., m

(11)

Where, δ is the system error generated by relative mobility RMta,u among nodes and  is the error created by maliciousness of the compromised nodes. The threshold values of δ and  are predetermined as given below.  ∈ [−max , max ], where 0 ≤  ≤ 1

(12)

δ ∈ [−δmax , δmax ], where 0 ≤ δ ≤ 1

(13)

Once the anchor node calculates this estimated distance, it is then forwarded to the BS encrypted with the public key of BS and along with the anchor node’s certificate. aj → BS : [d (14) ij ]BS K+ , Certaj All the anchor nodes send the distances among the unknown nodes and themselves to the BS. After receiving the message from the anchor nodes, BS decrypts the message with its private key and generates distance vector for each of the unknown node ui as:   bu = [d (15) i1 , di2 , ......, dim ], ∀ i = 1, 2, ...., n i

Finally, it uses MLE to estimate the location of an unknown node . One important consideration in this approach of MLE is that, it requires atleast three non-collinear anchor nodes to be applied. Another important feature of our proposed algorithm emphasizes the mobility of the nodes. We consider that all the nodes (both anchor and unknown) are having relative mobility (RM). The relative mobility between an unknown node ui and anchor node aj at a given time t is given by,   RMti,j = dij t − dij t−1 (16) RMti,j is positive if node ui is moving away from aj and negative if ui is coming closer to aj . Though the mobility is incorporated in the algorithm, nodes (both the anchor nodes and the unknown nodes) are assumed to be pseudo static i.e. they are static for a very short time interval for the localization process and this does not incorporate any significant error in the estimation. The benefits of our proposed algorithm in regard of preventing wormhole attack are as follows. Firstly, one hop neighbourhood concept will minimize the complexity and the probability of security breach through intermediate nodes of the multi hop communication. Secondly, mutual authentication will help to prevent any intrusion from the outside attackers. Finally, the speed of light “c used in the distance estimation will prevent the generation of high speed link required to execute wormhole attack because there cannot be any higher speed link in which the transmission speed will be more than that of the light. 8

Location estimation phase The location estimation problem for the location xu of an unknown node ui with noisy distance vector bui to m anchors at known locations xa1 , xa2 , ....,xam is formulated using maximum likelihood given as: ˜ T C −1 [d( ˜ xu ) − d] xu ) − d] x u = arg min[d(

(17)

Where d˜ is the m × 1 distance measurement vector, d( xu ) is another m × 1 vector [|| xu − xa1 ||, || xu − xa2 ||, ....., || xu − xam ||] and C is the covariance matrix of the distance measurement errors. We have to find the maximum likelihood estimator that minimizes the above quadratic equation. In order to get a simple estimator, d(xu ) can be linearized using the approach of Taylor series around a reference point x0 given as: d(xu ) ≈ d(x0 ) + dx (x0 ) + dx (x0 )(xu − x0 )

(18)

Where dx (x0 ) is in R2 and represents a (m − 1) × 2 matrix of partial derivatives of d with respect to x evaluated at x0 . A recursive solution to the maximum likelihood estimator can then be obtained as:    T −1   −1  T  

˜ + d S d dx xu,k S −1 d−d x x xu,k (19) xu,k+1  = x u,k x u,k x u,k

200

Further, to reduce the impact of distance estimation errors due to system error or malicious environment, we have used Cayley-Menger determinant. The concept is illustrated in the figure- for an unknown node ui having three distance estimations to three anchor nodes a1 ,a2 and a3 . a3

di3

ui di1

di2

d23

d13 a1

d12

a2

Figure 1: Propagation time estimation process

The Cayley-Menger determinant for the above example is given by:    0 d212 d213 d2 1 i1   2 d12 0 d223 d2i2 1   2 D(a1 , a2 , a3 , ui ) = d13 d223 0 d2i3 1  d2i1 d2i2 d2i3 0 1    1 1 1 1 0

(20)

A generic output of the Cayley-Menger determinant is given by a theorem shown in [30]. The theorem is as follows. 9

Theorem 1. Consider an n-tuple of points x1 , x2 , ... xn in m-dimensional space with n ≥ m + 1. The rank of the Cayley-Menger matrix M(x1 ,.....,xn ) is at most m + 1. According to this theorem, the value of D(a1 , a2 , a3 , ui ) is in R2 and given as : D(a1 , a2 , a3 , ui ) = 0

(21)

The inter anchor distances d1 2, d1 3 and d2 3 can be calculated from the known anchor positions. The true distances among the unknown node and the anchor nodes for the above example (shown in the figure-) is a function of the measured distances given as:  dij = f (d ij ), where di j = dij + ζij f or i = 1 and j = 1, 2, 3 and ζij = δij + ij (22) To generalize the above equation for n unknown nodes and m anchor nodes, we can rewrite the above equation as:  dij = f (d ij ), where di j = dij + ζij f or i = 1, 2, ............n and j = 1, 2, .......m (23) Putting this value of Equation 21 into Equation 20 we get the following according to the theory shown in [31] ζ T Aζ + ζ T b + c = 0

(24)

Where, ζ = [ζ1 , ζ2 , ....., ζj ]T , A is the matrix for the distances among the m anchor nodes and b is the vector for the measured distances from an unknown node to the anchor nodes given in Equation 14. An unknown node forming k quadrilaterals with one-hop anchor nodes will generate k independent equations like Equation 23. These equality constraints can be used with Equation 16 by Lagrange multipliers [32]. Finally, we have applied gradient descent algorithm [33] is used to search for the solution that gives the best estimation of unknown node’s location satisfying the descending condition as: F (∂k+1 ) < F (∂k ) (25) where ∂ = ( xu − xu ) and xu is the original location of ui Considering the variation of F-value along with the half line starting at x and with descent direction h, the Taylor expansion for the above function can be written as: F (∂ + αh) = F (∂) + αhT F  (∂) + O(α)2 (26) given that,



hT F (∂) < 0

(27)

αe = arg minα>0 F (∂ + αh)

(28)

10

Algorithm 1 Distance estimation by Anchor nodes 1: 2: 3: 4: 5: 6: 7: 8: 9: 10: 11: 12: 13: 14: 15: 16: 17: 18:

procedure Distance estimation by Anchor nodes Input: anchor node set A, unknown node set U Output: Maximum likelihood estimator for x u BS generates identities IDaj for all anchor nodes and identities IDui for all unknown nodes for each node aj ∈ A do Broadcast NDP message aj starts the timer for each node ui under Ravg of any aj ∈ A do ui sends aj : [N ]Kui− , Certui end for aj receives the message and stops the timer  Calculate dij = c × RT2 T  Calculate d ij = d × (1 ± δ) × (1 ± ) ij

aj sends d ij to BS end for   BS: Calculate bui = [d i1 , di2 , ....., dim ] ∀ i = 1, 2, ...., n Apply MLE end procedure

From the above equation, performing with a positive step value αh, the relative gain in the function satisfies the following. lim

α→0

 F (∂) − F (∂ + αh) 1 T  = − h F (∂) = −||F (∂)||Cosθ α||h|| ||h||

(29)



where, θ is the angle between the vector h and F (∂). This shows that we get the greatest gain rate if θ = π, i.e. if we use the steepest descent direction hsd given by  hsd = −F (∂) (30) This steepest descent method gives the best solution locally. To get a global solution we can use Newton’s method as shown in [34]. 4. Results In this section, we have evaluated the proposed algorithm based on the following parameters as shown in Table 2 We have compared the simulated results with the two recent algorithms: 1) Partitioning method with convex constraints for secure localization against wormhole [16], 2) Label based DVHop Secure Localization [18]. The performances of the algorithms are measured on two metrics: localization error rate and wormhole induced localization error.

11

Table 2: Simulation parameters and rules

Parameters Simulation Area No. of Unknown Nodes No. of Anchor Nodes Mobility Propogation Antenna Simulation time

Values 100 × 100 m2 100 to 200 10 to 40 Random TwoRayGround Omni antenna 10 minutes

Comparison of Localization Error 0.75

Proposed Algorithm Chen et al. [18] Niu et al. [16]

0.7 Localization Error Rate

0.65 0.6 0.55 0.5 0.45 0.4 0.35 0.3 0.25 0.1

0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28 Anchor Nodes Ratio

0.3

Figure 2: Anchor Nodes Ratio vs Localization Error Rate

Localization accuracy is an important parameter for evaluating the efficiency of localization algorithms. In the proposed work, the localization accuracy is defined by the localization error rate between the actual location and the calculated node position. In our simulation, we have varied the ratio of anchor nodes from 0.1 to 0.3 with increments of 0.02. Simulation result, shown in Figure 2, shows that the localization error rate decreases with the increasing number of anchor nodes. Similarly, location accuracy is also tested by varying the average transmission range of anchor nodes’ in presence of wormhole. Further, we have also shown the average localization error in Table 3. Theoretically, the range of the neighbour area for an anchor node is approximately its transmission range Ravg , since one node can only hear from its neighbours within the transmission range Ravg . But because of the shortcut path (the worst case scenario in our model as the use of light speed and certificate verification can easily remove the probability of wormhole) the computed map for the neighbour area of that node will be distorted, and so the diameter 12

Table 3: Performance under different parameters

Parameters No. of Anchors 10 10 10 10 10 10 20 20 20 20 20 20 30 30 30 30 30 30 40 40 40 40 40 40

No. of Unknown Nodes 100 120 140 160 180 200 100 120 140 160 180 200 100 120 140 160 180 200 100 120 140 160 180 200

Performance Average Localization Error 0.0540 0.0570 0.0592 0.0631 0.0667 0.0701 0.0531 0.0554 0.0569 0.0619 0.0643 0.0689 0.0519 0.0539 0.0547 0.0587 0.0617 0.0657 0.0498 0.0523 0.0534 0.0573 0.0610 0.0653

Computational time (seconds) 18.6453 21.4523 23.4736 24.4532 24.0372 26.1012 20.2232 23.0052 23.9613 25.0123 25.8196 27.0713 20.9813 22.6352 23.7879 25.2107 26.1213 28.3022 24.8372 25.0345 25.9432 27.2643 29.7412 31.1123

of that computed local map will be larger than the physical one. One of the observations regarding the effect of multiple wormholes has been shown in Figure 3. The locations of the ends of the two wormholes are highlighted by black circles in the same figure. From this figure, we can see that even two wormholes lie very close to each other, the peaks of diameters have still appeared in the nodes which are close to the ends of the wormholes. Continuing the above effect of wormholes on the average transmission ranges, we have analysed another parameter called Wormhole Induced Localization Error (WILE) that evaluates how much localization is affected by the wormhole attack. It is calculated based on localization error using the formula: W ILE =

errors with wormhole(s) − normal error normal error

13

(31)

Diameter

100 90 80 70 60 50 40 30 20 10 0

100 90 80 70 60 50 40 30 20 10 0

0

10

20

30

40 X

50

60

70

80

90 1000

10

20

30

40

50

60

70

80

100 90

Y

Figure 3: Diameter measurement with two wormholes

The experiments have also been done with varying the percentage of wormhole. Wormhole Induced Localization Error with 2% wormholes Wormhole Induced Localization Error (%)

Wormhole Induced Localization Error (%)

Wormhole Induced Localization Error with 1% wormholes 160

Proposed Algorithm Chen et al. [18] Niu et al. [16]

140 120 100 80 60 40 10

15

20 25 30 35 40 45 Transmission Range of Anchor Nodes (Meter)

200

Proposed Algorithm Chen et al. [18] Niu et al. [16]

180 160 140 120 100 80 60

50

10

15

20 25 30 35 40 45 Transmission Range of Anchor Nodes (Meter)

(a)

(b) Wormhole Induced Localization Error with 10% wormholes Wormhole Induced Localization Error (%)

Wormhole Induced Localization Error (%)

Wormhole Induced Localization Error with 5% wormholes 300

Proposed Algorithm Chen et al. [18] Niu et al. [16]

250

200

150

100

50 10

15

20 25 30 35 40 45 Transmission Range of Anchor Nodes (Meter)

50

50

400

Proposed Algorithm Chen et al. [18] Niu et al. [16]

350 300 250 200 150 100 50 10

(c)

15

20 25 30 35 40 45 Transmission Range of Anchor Nodes (Meter)

50

(d)

Figure 4: Wormhole Induced Localozation Error(%) vs Transmission Range of Anchor Nodes(Meter)

Result shown in Figure 4, signifies to the fact that the proposed algorithm is comparatively better in minimizing the wormhole induced localization error 14

against the increasing number of wormholes. Further, the different scenarios of wormholes confirms that wormhole attacks hardly can affect the localization in our proposed algorithm and therefore provide the efficiency to the proposed algorithm. 5. Conclusion Security is always an issue of concern in localization algorithms. A number of algorithms are introduced with security aspects so far, but the complexity of the algorithms in the resource constrained WSNs has not been emphasized. In this paper, we have addressed this problem considering the wormhole attack as most impactful and severe in WSNs and provided a solution with our proposed algorithm. The algorithm prevents wormhole attack in a very easy process utilizing the resources of the anchor nodes and BS, so that the resource constrained unknown nodes do not have overhead of calculation. The simulation results also prove the efficiency of the proposed algorithm in terms of localization accuracy. We have also introduced a new parameter called as wormhole induced localization error to analyze the effect of wormhole in the scenarios. The comparison results show that our proposed algorithm outperforms the algorithms in comparison. The most important feature of our algorithm is that it supports mobility of the nodes and therefore it is robust and adaptable for dynamic network environments. References [1] D.D. Perkins, R. Tumati, H. Wu, I. Ajbar, Localization in Wireless Ad Hoc Networks, in: Resour. Manag. Wirel. Netw., Kluwer Academic Publishers, Boston, 2005: pp. 507-542. doi:10.1007/0-387-23808-5 18. [2] A. Boukerche, H.A.B.F. Oliveira, E.F. Nakamura, A.A.F. Loureiro, Vehicular Ad Hoc Networks: A New Challenge for Localization-Based Systems, Comput. Commun. 31 (2008) 28382849. doi:10.1016/j.comcom.2007.12.004. [3] K.K. Chintalapudi, On the Feasibility of Ad-Hoc Localization Systems, Tech. Rep., Comput. Sci. Dep. Univ. South. California,. (2003) 117. [4] R.F. Cao Xiao-mei , Yu Bo , Chen Gui-hai, Security Analysis on Node Localization Systems of Wireless Sensor Networks, J. Softw. 19 (2008) 879887. [5] Fatema, N., Brad, R.: Attacks and Counter attacks on Wireless Sensor NetworksInt. J. Ad hoc, Sens. Ubiquitous Comput., 2013, 4, (6), pp. 115. [6] Hu, Y.-C.H.Y.-C., Perrig, a., Johnson, D.B.: Wormhole attacks in wireless networksIEEE J. Sel. Areas Commun., 2006, 24, (2), pp. 370380. [7] Ban, X., Sarkar, R., Gao, J.: Local connectivity tests to identify wormholes in wireless networksProc. Twelfth ACM Int. Symp. Mob. Ad Hoc Netw. Comput., 2011, pp. 13:113:11. 15

[8] Rasmussen, K.B., apkun, S.: Location privacy of distance bounding protocolsProc. 15th ACM Conf. Comput. Commun. Secur. - CCS 08, 2008, (i), pp. 181191. [9] Brands, S., Chaum, D.: Distance-Bounding ProtocolsAdv. Cryptol. EUROCRYPT 93, 1994, 765, pp. 344359. [10] Srdjan Capkun, J.-P.H.: Secure positioning in wireless networksIEEE J. Sel. Areas Commun., 2006, 24, (2), pp. 221232. [11] Srdjan Capkun: Securing Localization in Wireless Networks (using verifiable multilateration and covert base stations Secure localization and time synchronization for wireless sensor and ad hoc networks, advances in information security, 2007, 30, pp. 249278. [12] Lazos, L., Poovendran, R., Capkun, S.: Rope: robust position estimation in wireless sensor networks, in IPSN 2005. Fourth International Symposium on Information Processing in Sensor Networks, 2005. (2005), pp. 324331 [13] Lazos, L., Poovendran, R.: SeRLoc: Secure Range-Independent Localization for Wireless Sensor NetworksNetw. Secur., 2004, pp. 2130. [14] Lazos, L., Poovendran, R.: HiRLoc: High-resolution robust localization for wireless sensor networksIEEE J. Sel. Areas Commun., 2006, 24, (2), pp. 233246. [15] Srinivasan, A., Teitelbaum, J., Jie, W.: DRBTS: Distributed reputationbased beacon trust system, in Proceedings - 2nd IEEE International Symposium on Dependable, Autonomic and Secure Computing, DASC 2006 (2006), pp. 277283. [16] Niu, Y., Gao, D., Gao, S., Chen, P.: A Robust Localization in Wireless Sensor Networks against Wormhole AttackJ. Networks, 2012, 7, (1), pp. 187194. [17] Chen, H., Lou, W., Sun, X., Wang, Z.: A Secure Localization Approach against Wormhole Attacks Using Distance ConsistencyEURASIP J. Wirel. Commun. Netw., 2010, 2010. [18] Chen, H., Lou, W., Wang, Z., Wu, J., Wang, Z.: Securing DV-Hop localization against wormhole attacks in wireless sensor networksPervasive Mob. Comput., 2014, pp. 114. [19] Labraoui, N., Gueroui, M., Aliouat, M.: Secure DV-Hop localization scheme against wormholeTrans. Emerg. Telecommun. Technol., 2012, (December 2011), pp. 303316. [20] Kong, F., Li, C., Ding, Q., Cui, G., Cui, B.: WAPN: a distributed wormhole attack detection approach for wireless sensor networksJ. Zhejiang Univ. Sci. A, 2009, 10, (2), pp. 279289.

16

300

[21] Xu, Y., Chen, G., Ford, J., Makedon, F.: DETECTING WORMHOLE ATTACKS IN WIRELESS SENSOR NETWORKS, in CRITICAL INFRASTRUCTURE PROTECTION (2007), pp. 207-219. [22] Amish, P., Vaghela, V.B.: Detection and Prevention of Wormhole Attack in Wireless Sensor Network using AOMDV protocolProcedia Comput. Sci., 2016, 79, pp. 700-707. [23] Shalinie, G.A.S.M.: Wormhole-Free Routing and DoS [24] Krentz, K., Wunder, G.: 6LoWPAN Security: Avoiding Hidden Wormholes using Channel Reciprocity, in 4th International Workshop on Trustworthy Embedded Devices (2014), pp. 13-22. [25] Shi, F., Liu, W., Jin, D., Song, J.: A countermeasure against wormhole attacks in MANETs using analytical hierarchy process methodologyElectron. Commer. Res., 2013, 13, (3), pp. 329-345. [26] Stoleru, R., Wu, H., Chenji, H.: Secure neighbor discovery and wormhole localization in mobile ad hoc networksAd Hoc Networks, 2012, 10, (7), pp. 1179-1190. [27] Hayajneh, T., Krishnamurthy, P., Tipper, D., Le, A.: Secure neighborhood creation in wireless ad hoc networks using hop count discrepanciesMob. Networks Appl., 2012, 17, (3), pp. 415-430. [28] Khalil, I., Awad, M., Khreishah, A.: CTAC: Control traffic tunneling attacks countermeasures in mobile wireless networksComput. Networks, 2012, 56, (14), pp. 3300-3317. [29] Dimitriou, T., Giannetsos, A.: Wormholes no more? Localized wormhole detection and prevention in wireless networksLect. Notes Comput. Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics), 2010, 6131 LNCS, pp. 334-347. [30] Crippen, G.M., Havel, T.F.: Distance Geometry and Molecular Conformation1988, (January 1988). [31] Blumenthal, L.M.: Theory and Applications Distance Geometry (Oxford University Press, 1953). [32] Cao, M., Anderson, B.D.O., Morse, A.S.: Localization with imprecise distance information in sensor networksProc. 44th IEEE Conf. Decis. Control. Eur. Control Conf. CDC-ECC 05, 2005, 2005, pp. 2829-2834. [33] Cao, M., Anderson, B.D.O., Morse, A.S.: Sensor network localization with imprecise distancesSyst. Control Lett., 2006, 55, (11), pp. 887-893. [34] K. Madsen, H.B. Nielsen, O.T.: METHODS FOR NON-LINEAR LEAST SQUARES PROBLEMS (Informatics and Mathematical Modelling Technical University of Denmark, 2004). 17