Performance of Adaptive Beam Nulling in Multihop Ad Hoc networks under Jamming

Transcription

1 Performance of Adaptive Beam Nuing in Mutihop Ad Hoc networks under Jamming Suman Bhunia, Vahid Behzadan, Pauo Aexandre Regis, Shamik Sengupta Dept. of Computer Science and Engineering, University of Nevada, Reno, USA, Abstract In a mutihop ad hoc network, end-to-end data transmissions traverse through mutipe inter-node wireess inks. A jammer can disrupt the entire data transfer of a network by intentionay interfering with inks between a subset of nodes. The impact of such attacks is escaated when the jammer is moving. Whie the majority of current ad hoc protocos consider omnidirectiona transmission and reception, adaptive antennas can be utiized for spatia fitering of the jamming signa. This paper investigates the performance of empoying adaptive beam nuing as a mitigation technique against jamming attacks in mutihop ad hoc networks. Considering a moving jammer, the survivabiity of inks and connectivity in such networks are studied by simuating various node distributions and different mobiity patterns of the attacker. In addition, the impact of errors in estimation of direction of arriva and beamforming on the overa network performance are aso examined. The resuts of this study indicate a significant improvement in retaining connectivity under jamming when adaptive beam nuing is appied. Keywords Beamforming, Beam Nuing, mutihop, Ad hoc, Anti-Jamming, Mobiity. I. INTRODUCTION The ecosystem of wireess communications is evoving towards independent, sef-configuring architectures. Dependence of teecommunications on the infrastructure is envisioned to be argey diminished in a gradua move towards ad hoc networking. Especiay, mutihop ad hoc networks are predicted to pay a key roe in future mission critica communications, such as emergency radio networks in disaster zones, tactica mobie networks, and UAV communications. But the security of such networks heaviy depend on the reiabiity of the wireess inks. The open nature of wireess medium eaves the inks inherenty vunerabe to interference and jamming. In hostie environments such as battefieds, disrupting such inks by means of jamming is an essentia aim of an adversary s eectronic warfare operations. Hence, mitigating jamming attacks has been a crucia research issue for the wireess community [1]. Some we-known categories of anti-jamming techniques proposed in the iterature are those that utiize speciay designed signa coding and moduation, such as Frequency Hopping Spread Spectrum (FHSS) [2] and Direct Sequence Spread Spectrum (DSSS) [3]. The downside associated with this cass of techniques is their arger bandwidth requirement, which considering the state of the overcrowded eectromagnetic spectrum, can prove to be costy. To preserve the scarce bandwidth, an aternative is to appy Spatia Fitering with beamforming antenna arrays [4]. This approach expoits the beamformers abiity to detect the Direction of Arriva (DoA) of signas. This direction is then used to modify the array s response so the interference sources are paced in the nus of the antenna. Beamforming antenna systems that impement This research was supported by NSF CAREER grant CNS #13466 A F G B Jamming range Origina path E H C D Omnidirectiona path Adaptive beam path Fig. 1: A comparison of routing in omnidirectiona vs beam nuing schemes under jamming this mechanism are referred to as Adaptive Nuing Antennas (ANA). The fow of information in mutihop ad hoc networks can be disrupted by jamming a subset of nodes in a region. Traditionay, ad hoc configurations assume an omnidirectiona antenna for communications. In mutihop networks, data is routed over mutipe hops to reach a destination that is not within communication range of the source. By adopting beam nuing techniques, a node can adapt its radiation pattern so a nu is created in the direction of a jammer. This aows maintaining the inks which are not affected by the jammer. Figure 1 provides an exampe of end-to-end data deivery in an ad hoc network. In the absence of a jammer, a packet foows through the path A B C D when a nodes empoy omnidirectiona antennas. In this configuration, the jammer can effectivey jam nodes B, C and E. The routing protoco discovers the ink faiures and reroutes packets through A F G H I D. This way packets are deivered but with an increased end-toend deay and congestion on ink G H. However, if nodes expoit beam nuing, nodes B, C and E can successfuy avoid the jammer. Now packets can be deivered through A B E C D. Hence it can be seen that, in the presence of a jammer, adaptive beam nuing not ony maintains the connectivity of the nodes inside the affected region, but aso ensures ess congestion on the remaining inks. The majority of the iterature on ANAs rey on the assumption that the jammers are stationary with respect to beamformers (e.g. [], [6], [7], [8], [9]), but with the recent expansion and growth of mobie wireess technoogies, this assumption does not necessariy hod true. Aso, there is a ack of pubicy avaiabe anaysis on the network performance of ad hoc networks utiizing adaptive nuing antennas under I

2 jamming. This paper aims to fi this gap by providing a network-oriented anaysis via investigating the effects of a moving jammer on the connectivity and ink survivabiity of an ad hoc network of nodes equipped with ANAs. For this purpose, mutipe simuations have been performed to study the impact of jamming based on connectivity, number of isands, and number of surviving inks for different node densities and jammer s mobiity modes. The simuation resuts show that the proposed mechanism can achieve up to 7.27% of improvement in connectivity over the omnidirectiona antenna case. The remainder of this paper is organized as foows: Section II provides a background on beam nuing techniques. The proposed methodoogy of adaptive beam nuing is presented in Section III. Section IV describes the simuation setup and resuts. Finay Section V concudes the paper. S1 S2 Sn Nuing Channes { { Weights W1 W2 Wn Process Combine (for Feedback Systems) To Communications Receiver II. BACKGROUND This section presents a discussion on terminoogy and concepts of adaptive nuing antennas. This is not intended to be a thorough overview of the beamforming and nuing techniques, but aims to provide the very basics to equip the reader with enough background to understand the rest of this paper. Interested readers are referred to [1] as a comprehensive source on beamforming and nuing antennas. A. Antenna Terminoogy Antennas are eements that coupe eectromagnetic energy between free space and a guiding structure [11]. Antennas may be cassified based on how they radiate and receive energy in different directions. The directionaity or gain of an antenna in a direction d = (θ, φ) is defined as: G( d) = η U( d) U ave (1) Where η is the antenna efficiency, U( d) is the power density in the direction of d, and U ave is the average power density in a directions. An isotropic antenna is a radiator which has uniform gain in a directions ( U( d) = U ave for a directions). An omnidirectiona antenna is defined as a radiator which has (reativey) constant gain in at east one 2-dimensiona pane of directions. A directiona antenna is one which radiates more energy in one or more directions compared to other directions. Antenna Radiation Pattern is the representation of the gain vaues in a or a subset of a directions. The pattern typicay has a main obe in which the gain is at its peak, and some side obes. In this paper, we interchangeaby refer to obes as beams. B. Adaptive Nuing Antennas The circuitry of a beamforming antenna array is depicted in Figure 2. Signas coming from antenna eements consist of the desired signas, interference and noise. The contro process determines individua weights of each signa based on an array response optimization method. In case of Adaptive Nuing Antenna (ANA) arrays, the weights are chosen so that the array response has nus towards the directions of interference sources. Various agorithms for adaptive estimation of Direction of Arriva (DoA) for both the desired and interference signas have been introduced and investigated in the iterature. Such agorithms can be cassified into beamscan agorithms and Fig. 2: Schema for ANA subspace agorithms [12]. Beamscan methods are based on scanning a conventiona beam to cover a region and record the magnitude squared of the output. Exampes of this cass are Minimum Variance Distortioness Response (MVDR) and root MVDR [13]. On the other hand, Subspace agorithms expoit the orthogonaity between the signa and noise subspaces. MUSIC, Root-MUSIC and ESPIRIT are among the most efficient subspace DoA estimation agorithms in antenna arrays [14]. A thorough review and comparison of widey used DoA estimation methods has been provided in [14]. Once the anguar direction of an interference signa is determined, a beamformer cacuates the weight vaues which resut in a nu towards the interference source. Some of the major weight cacuation methods are Doph-Chebyshev weighting, Least Mean Squares (LMS) and Conjugate Gradient Method (CGM) [1]. In the case of mobie ad hoc networks, in which the directions of desired and interference signas are not known and vary, Stochastic Search agorithms are appied [16]. Exampes of such methods are Gradient Search Based Adaptive agorithms [17], [18], [19], Genetic Agorithms [2], [21], [22] and Simuated Anneaing [23], [24]. Thorough reviews and comparison of beamforming methods and agorithms are provided in [16] and [2]. III. METHODOLOGY A. Probem Statement Figure 3 iustrates the effect of adaptive beam nuing in the presence of a moving jammer. In this scenario, the one hop inks between node A and its neighbors B, C, D and E are considered. Node A periodicay scans for the DoA of the jammer s signa (θ m ) in intervas of (τ) seconds. Due to the discontinuous observation of the jammer s DoA, whie cacuating the nu ange, A must take into account the movement of the jammer between two consecutive observations. This cacuation must incude prediction of the jammer s anguar veocity by considering its history of movements. As the mobiity pattern of a jammer becomes more random, the prediction accuracy of its movements decreases. Therefore, the effect of various mobiity patterns of the jammer on a network of beam nuing nodes can provide a practica measure for efficiency of this scheme. Node A uses a modified beam pattern to communicate with its neighbors unti the next sensing period. In Figure 3a,

3 B A θ m+1 D bh b θ m C E t=m θ= t=m+1 B A θ m+1 D b bh θ m C E t=m θ= t=m+1 a disrupting signa on the same frequency channe as the ad hoc network. Even though the introduction of a nu in an omnidirectiona pattern of a beamforming node may be interpreted as changing the mode of communications to directiona transmission, hence necessitating the use of Directiona MAC protocos [26]. However, the higher network ayers can continue to operate under the defaut assumption of omnidirectiona transmission, since the nued direction is aready under jamming and no hidden/exposed termina probem may arise from that direction [27]. (a) narrow nu ange weak jamming signa strong jamming signa jammer's trajectory (b) wide nu ange weak neighbor signa strong neighbor signa notched ange Fig. 3: Depiction of the beam nuing principe. A has a narrower nu ange compared to Figure 3b. With this narrow nu ange, A can communicate with B, D and E, whereas with a wider nu ange, A can communicate ony with B and D. By the next sensing period the jammer moves to a new position, faing outside of the narrower nu, which consequenty exposes A to the jammer. As a resut, a inks are disrupted for A. On the other hand, the wider nu ange maintains the jammer inside the nu region for the whoe interva. The trade-off for widening the nu to cover the jammer s probabe movements, is the cost of disabing unaffected inks. Hence, another important factor in efficiency of beam nuing is the choice of optimum nuing ange in dynamic scenarios. The practica imitations of adaptive beam nuing, such as inaccuracy in estimation of DoA, as we as hardware imitations in impementing a desired antenna pattern, ead to introduction of errors in a beamformer s performance. The aforementioned errors are formay defined as foows: the measurement error is the error in DoA estimation. If θactua c is the actua anguar position of the jammer with respect to A, but the observed vaue by A is θobserved c, the measurement error is defined as θactua c θc observed. Simiary, If a node cacuates a nu ange boundary at b c intended, but the impemented boundary is formed at b c impemented, beamforming error is defined as b c intended - bc impemented. For a sensibe study on the efficiency of a practica impementation, investigating the impact of inherent system errors in the simuation is of crucia importance. B. System Assumptions To investigate the effect and feasibiity of adaptive beam nuing as a counter-measure for jamming attacks, a mutihop wireess ad hoc network is considered. The nodes in this network are assumed to be static reative to each other. Each node is considered to be equipped with a beamforming antenna array, capabe of introducing nus in its originay omnidirectiona radiation pattern. A node can not determine the DoA of jammer s signa whie it is communicating with its neighbors. To determine the jammer s DoA it goes through a sensing phase at every τ seconds interva. If the received jamming signa is above an interference threshod it identifies an attack. The jammer is assumed to be a moving node with an omnidirectiona antenna that continuousy transmits C. Adaptive beam nuing Let s consider a mutihop ad hoc network of N nodes. After sensing the presence of a jammer, each node i N observes the anguar position of the jammer or the ange of attack (θa m ) with its reference frame at every sensing phase m {1,..., M}. Node i then adjusts its beamform to attenuate the signa from the jammer. i uses this beamform to communicate with its neighbors unti the next sensing phase (m + 1). In Figure 3, at the m th sensing phase, the jammer is sensed at ange θa m. In the next sensing phase (m+1), i senses the jammer at θa m+1. Since the jammer is moving, it may cross the nu of the beamform and i woud be interfered by the jamming signa. The aim of adaptive beam nuing is to make sure the jammer stays within the nued region for the entire time between sensing periods m and m + 1. i cacuates the anguar veocity of the jammer as (θa m θa m 1 )/τ, and keeps this vaue in a veocity array (v a ). Consider v a and σ(v a ) as the mean and standard deviation of the veocity (v a ), respectivey. Node i constructs a beam nu using an agorithm that considers the history of jammer s movement. A beam nu is defined by two boarders: b m and b m h which are ower and higher anges respectivey. Ceary, θa m + τv a gives the estimated ocation of the jammer at the (m + 1) th sot. Since the actua veocity and direction of the jammer are unknown, the nu shoud be wider in case the jammer changes direction or veocity. Change of veocity of the jammer can be estimated with σ(v a ). If a jammer changes its direction or veocity, σ(v a ) woud be high compared to the case when the jammer moves at the same direction with constant veocity. An estimation for the beam nu ange can be cacuated as: b m h = max(θa m, θa m + τ(v a + ασ(v a ))) (2) b m = min(θa m, θa m + τ(v a ασ(v a ))) (3) m = b m b m (4) Where m is the nu ange constructed at the m th sensing phase, and α is a mutipying factor. Note that the higher the vaue of α, higher the nu ange is. Now, if the nu is wider, chances are that more egitimate neighbors fa in this nued region. A node i cannot communicate with its neighbor j if j is in the nued region of i and vice versa. A higher vaue of α guarantees a higher probabiity that the jammer stays in the nued region unti the next sensing period. A very high vaue of α resuts in more deactivated inks. In Section IV-D1 we can observe that the system performance is a convex function w.r.t. α. Since the jammer s mobiity pattern is not competey observabe by a node, it shoud use adaptive vaue of α. To mitigate this effect, we propose a heuristic that dynamicay adapts the vaue of α based on the observed history of jammer s movements.

4 m-1 b h m-1 Node i nu border f h a m at m f a m-1 m-1 k safety fence at m-1 b m-1 safety fence nu border = Fig. 4: Schema for adaptive α heuristics D. Heuristic for dynamic α Agorithm 1 presents a heuristic for adapting the vaue of α at each sensing period m. Figure 4 presents the schema for this adaptation. The beam nu has been created in the previous sensing period m 1. At the m th sensing sot, if the jammer stays inside the nued region ( m 1 ), then the node successfuy avoids the attack. If the jammer is too cose to the nu border, α is increased. The agorithm considers a safety zone defined by two fences: f h and f. We consider a factor k (,.) which defines how defensive the network is. The safety fence is a m 1 /k deviation from the nu border towards the center of the nu. Larger vaues of k owers the probabiity of the jammer being in the safety zone, which consequenty increases α, resuting in a wider nu for the next interva. If the jammer stayed inside the safety zone, α is reduced by a factor of ɛ (, 1). δ is cacuated as the deviation of the jammer from the safety fence. At the m th sensing phase, if the jammer is observed between the nu border and the safety fence, α is increased by a factor of (1 + kδ ). This entais α is doubed if the m 1 jammer is at the nu border. If the jammer crosses the nu border, α is aggressivey increased by a mutipying factor of (1 + ( kδ ) 2 ). m 1 IV. SIMULATION AND RESULTS A. Simuation setup A customized tick based simuator is deveoped to measure the performance of the proposed agorithm. Each tick represents the time interva (τ) between two consecutive sensing periods. The defaut parameters used are isted in Tabe I. Some parameters are varied to observe their effect on the performance. During the sensing phase, at each tick (m), every node checks for the jammer s anguar position (θa m ). Each node then determines its new beamform according to eq. 4 and updates α using Agorithm 1. After the sensing and beamforming phases, communication with neighbors takes pace unti the time interva (τ) ends, when the same cyce is repeated. Each simuation generates the position of each node randomy. The same set of positions are used to measure the performance of the network whie varying other parameters. For simpicity, the simuator considers free space path oss mode to cacuate the received power. The simuator defines the inks between two nodes on each iteration based on the received power from the corresponding neighbor and interference from the jammer at that moment. If the power received is above Agorithm 1: Heuristics for dynamic α 1 m 1 b m 1 h 2 f b m 1 3 f h b m 1 h + m 1 k b m 1 m 1 k 4 if f < θa m < f h then α ɛα 6 ese if θa m > b m 1 then 7 δ θa m f h 8 α α(1 + ( kδ ) 2 ) m 1 9 end 1 ese if θa m < b m 1 then 11 δ f θ m a 12 α α(1 + ( kδ m 1 ) 2 ) 13 end 14 ese 1 if θ m a > f h then 16 δ θ m a f h 17 ese 18 δ f θa m 19 end 2 α α(1 + kδ ) m 1 21 end (a) simuation with 1 nodes (b) simuation with 1 nodes Fig. : Snapshots of simuations the cutoff and neither of the nodes are jammed, the simuator considers the ink to be active. TABLE I: Simuation Parameters Parameters Symbo Vaues Simuation area 1, 1, m 2 Transmission power P t 3 dbm Received Power cutoff P r -78 dbm Communication Frequency 2.4 GHz Communication Radious 3146 m Initia α α DOA error standard deviation. Beam nuing error standard deviation µ bn. Number of nodes simuated N 1 Sensing interva τ ms Simuation Time s mobiity mode Random Wak The simuator considers a scenario of N nodes scattered randomy in an area of 1, 1, m 2. Each node transmits with power of 3 dbm and the average communication radius is cacuated as 3146m. Figure a and Figure b depict snapshots of nodes during sensing intervas. One hop communication inks are represented with yeow ines. The cyan and magenta ines represent the nu borders b and b h

5 Gauss Markov mode Random Wak mode Random Direction mode Predefined Path mode Fig. 6: Time domain sketch of different trajectory modes respectivey. If the jammer is not in the proximity of a node to effectivey jam it, the node does not use beam nuing. With effective beam nuing, the nodes keep some of their inks active. In Figure a we can see 1 nodes present in the area. The network is not fuy connected and there are 3 isands. Figure b shows a snapshot of simuation with 1 nodes. We can see that due to jamming, some inks are deactivated but the network maintains its connectivity and no node is disconnected from the network. The simuator considers the error in DoA estimation and beam nuing. A node observes DoA of the jammer as θobserved c whie the actua DoA is θactua c. For each node, the simuator generates θobserved c randomy based on a norma distribution with mean of θactua c and standard deviation of. Simiary the simuator introduces hardware imitation for beam nuing. If a node wants to form a beam nu boundary at b c intended, the actua boundary is formed at b c impemented which has a Gaussian distribution with mean of b c intended, and standard deviation µ bn. B. and mobiity mode In this work a moving jammer is considered. Different mobiity modes of the jammer impact differenty on a network. A mobiity mode defines how a node moves or changes its direction with time. The detais of the seected modes (Random Wak, Random Direction, Gauss-Markov, and a predefined path) can be seen in [28] and [29]. Random-based modes are vasty used in the research community but they might not reproduce a reaistic movement. Gauss-Markov is a tempora dependency mode that can be considered more reaistic, where the veocity and direction are correated to the previous vaues, avoiding abrupt changes that occurs in the former modes. A predefined path is aso experimented assuming that a node foows a previousy assigned path. Each mode has its own infuence in the performance of the network. C. Performance Metrics Three performance parameters are defined as foows: Connectivity is defined as the tota number of connected pairs, which refects how we connected a network is. It is defined as the summation of connected nodes. More precisey, connectivity of a network is 1 2 ( i N j N connected(i, j)), where connected(i, j) = 1 if there exists at east one path from i to j and otherwise. The second parameter is average number of active inks. We consider a ink as the one hop communication between two neighbors. A ink may fai if either of the nodes is jammed or fas in the nued region of the other one. The next performance parameter considered is the average number of isands. Some node may not be abe to communicate with its neighbors. This resuts sometimes in a node being isoated from the rest of the network, or a group of nodes isoated from the other groups. We count the number of isand present in the network periodicay. If a network is competey connected, the number of isand is 1. The higher amount of isands is, the more disrupted the network is. The simuator monitors the above mentioned metrics at each iteration. It cacuates the average of these matrices after the fu simuation and record them as the resut. D. Resuts and Discussion 1) Discrete fixed α: In the initia phase of the simuation, the effect of α on system s performance is investigated. In this case the network is simuated without adaptive α, i.e. nodes do not use Agorithm 1. Figure Figure 7 presents the simuation resuts when α is fixed. The x-axis of these pots represent the discrete vaues of α that form the beam nu in eq. 4. Nine different scenarios are considered: one benchmark scenario with no jamming and for each mobiity mode we simuated the network once with omnidirectiona antenna, and once with the proposed beam nuing agorithm. The worst case scenario occurs when there is a jammer in the fied and the nodes use omnidirectiona antenna, consequenty the performance is heaviy affected by the presence of the jammer. The top benchmark resut is obtained simiary to the worst case but with no jammer present, therefore the communications are not affected by any adversary. It can be seen from the resuts that when there is no jammer, the network is competey connected as the number of isands is 1. For a competey connected network with n nodes, the connectivity vaue is n(n 1) 2. Therefore, in a network of 1 nodes with no jammer, the connectivity is 49, supporting the simuated resut. When nodes do not use beam nuing isands are created, resuting in a poor connectivity vaue. Aso it is observed that in the presence of a jammer, adaptive beam nuing significanty improves the overa performance in terms of a the metrics considered. In addition, when a jammer is present and the nodes do not appy beam nuing, the network is heaviy affected, and a arger number of isands is created. However, when nodes appy adaptive beam nuing, different trajectory modes perform differenty with respect to the vaues of α. It is noteworthy that for higher vaues of α, the number of average inks may fa beow the benchmark case of omnidirectiona nodes in the presence of a jammer. This is because a higher vaue of α creates a wider nu that resuts in deactivation of more inks. A node may reduce this shortcoming by sensing the jammer more frequenty but this aso reduces the data communication window. In addition, it can be observed that as α increases, the average number of isands decreases, whie the number of active inks begin to deteriorate after a peak. This phenomenon can be interpreted

6 average connectivity average no. active inks (α) Omnidirectiona, no jammer Omnidirectiona, jammer: random direction Proposed mechanism, jammer: random direction Omnidirectiona, jammer: random wak Proposed mechanism, jammer: random wak Omnidirectiona, jammer: gauss-markov Proposed mechanism, jammer: gauss-markov Omnidirectiona, jammer: predefined path Proposed mechanism, jammer: predefined path Fig. 7: Simuation resuts with fixed α as a rise in congestion. Another concusion that can be derived from the resuts is that a fixed vaue of α does not guarantee the optima performance, since the mobiity pattern of the jammer is not known to any node. A node estimates the veocity of the jammer through periodic sensing. Therefore, the vaue of α must be dynamicay updated based on the history of the jammer s movements. 2) Effect of s mobiity mode: Four different mobiity modes for the jammer are considered. Figure 8a iustrates the impact of these modes on the defending network. It can be seen that the Random Direction and Random Wak modes adversey affect the performances of the network, since the direction of the jammer undergoes abrupt changes in random intervas. For the predefined path and Gauss-Markov modes, the direction and veocity are constant for the majority of the time, which aows adaptive beam nuing system to accuratey estimate the jammer s movement. It is observed that for 1 nodes, the proposed mechanism achieves an improvement in connectivity of up to 7.27% reative to the omnidirectiona under attack. 3) Effect of node density: Figure 8b iustrates the effect of varying number of nodes in the network which constitutes a change in the node density. It is observed that when a network is not we connected, the number of isands increases. As the number of nodes increases, connectivity is we preserved in the no jamming scenario. The jammer succeeds in disabing more inks when the node density higher. Even though the number of ink faiures is on a simiar eve as the worst benchmark of omnidirectiona with jamming, connectivity and number of isands demonstrate a better performance. In the benchmark scenario with omnidirectiona antennas, the number of isands increases greaty with an increase in the number of nodes, since the density is higher and the attacker has more inks in its jamming range. The proposed adaptive beam nuing approach succeeds in keeping the connectivity and number of isands cose to the scenario of the benchmark with no jamming. 4) Effect of errors in beam nuing: As discussed earier, errors are introduced in the simuator to account for the practica inaccuracies in beam nuing and DoA estimation. The effective beam nu border is a random function with the mean of intended boarder ange and standard deviation of µ bn. Simiary for each node the observed DoA is a random function of mean at the actua DoA and standard deviation of. Figure 8c pots the performance of the network w.r.t. the error in the beam nuing. The X-axis is µ bn, whie the simuations are repeated with severa different vaues of. With a of.1 that entais an error of.7 o in DoA measurement the connectivity sti remains cose to that of no jamming scenario. The pots refect that both the error decreases the network performances significanty as the jammer is not tracked propery. However, with a higher vaue of error in measurement the proposed beam nuing mechanism sti performs better than the omnidirectiona antenna mode. V. CONCLUSION This paper investigates the performance of adaptive beam nuing in mutihop ad hoc networks under attack from a moving jammer. To iustrate the effectiveness of the proposed mechanism against jamming attacks, connectivity of various network topoogies with different mobiity patterns of the jammer are studied through simuations. Aso, to increase the accuracy of the simuated modes for practica impementations, effects of varying inherent errors on the performance of a beam nuing ad hoc network is considered. The resuts demonstrate a significant improvement in survivabiity of connectivity when adaptive nuing is used with our proposed mechanism. REFERENCES [1] R. Saranyadevi, M. Shobana, and D. Prabakar, Artice: A survey on preventing jamming attacks in wireess communication, Internationa Journa of Computer Appications, vo. 7, pp. 1 3, November 212. Fu text avaiabe. [2] A. Yahya, O. Sidek, J. Mohamad-Saeh, and C. M. D. E. Center, Performance anayses of fast frequency hopping spread spectrum and jamming systems., Int. Arab J. Inf. Techno., vo., no. 2, pp , 28. [3] L. B. Mistein, Interference rejection techniques in spread spectrum communications, Proceedings of the IEEE, vo. 76, no. 6, pp , [4] B. D. Van Veen and K. M. Buckey, Beamforming: A versatie approach to spatia fitering, IEEE assp magazine, vo., no. 2, pp. 4 24, [] W. C. Cummings, Anadaptive nuing antenna for miitary, Lincon Laboratory Journa, vo., no. 2, [6] G. Okamoto, nuing via ow compexity bind beamforming agorithm, in Antennas and Propagation Society Internationa Symposium, 27 IEEE, pp. 2 28, IEEE, 27.

7 average connectivity average no. active inks Gauss Markov Random Direction Random Wak Predefined Path Omnidirectiona, no jammer Omnidirectiona, jammer Proposed mechanism, jammer average connectivity average no. active inks no. of nodes (N) Omnidirectiona, no jammer Omnidirectiona, jammer Proposed mechanism, jammer average connectivity average no. active inks beam nuing error standard deviation(µ bn ) =.1 =.2 =.3 =.4 =. (a) different trajectory modes (b) varying number of Nodes (c) varying error rate Fig. 8: Simuation Resuts varying different simuation parameters [7] W. Zhu, B. Daneshrad, J. Bhatia, H.-S. Kim, D. Liu, K. Mohammed, R. Prabhu, S. Sasi, and A. Shah, Mimo systems for miitary communications, in Miitary Communications Conference, 26. MILCOM 26. IEEE, pp. 1 7, IEEE, 26. [8] D. Mingjie, P. Xinjian, Y. Fang, and L. Jianghong, Research on the technoogy of adaptive nuing antenna used in anti-jam gps, in Radar, 21 CIE Internationa Conference on, Proceedings, pp , IEEE, 21. [9] G. K. Rao and R. S. H. Rao, Status study on sustainabiity of sateite communication systems under hostie jamming environment, in India Conference (INDICON), 211 Annua IEEE, pp. 1 7, IEEE, 211. [1] J. C. Liberti and T. S. Rappaport, Smart antennas for wireess communications: IS-9 and third generation CDMA appications. Prentice Ha PTR, [11] C. A. Baanis, Antenna theory: anaysis and design. John Wiey & Sons, 212. [12] H. L. Van Trees, Detection, estimation, and moduation theory. John Wiey & Sons, 24. [13] C. Vaidyanathan and K. M. Buckey, Performance anaysis of the mvdr spatia spectrum estimator, Signa Processing, IEEE Transactions on, vo. 43, no. 6, pp , 199. [14] V. Krishnaveni, T. Kesavamurthy, et a., Beamforming for direction-ofarriva (doa) estimation: A survey, Internationa Journa of Computer Appications, vo. 61, no. 11, pp. 4 11, 213. [1] J. Voakis, Antenna Engineering Handbook, Fourth Edition. McGraw- Hi Companies,Incorporated, 27. [16] J. Becker, Dynamic beamforming optimization for anti-jamming and hardware faut recovery, 214. [17] L. Lei, X. Rongqing, and L. Gaopeng, Robust adaptive beamforming based on generaized sideobe canceation, in Radar, 26. CIE 6. Internationa Conference on, pp. 1 4, IEEE, 26. [18] Y. Xu and Z. Liu, Noncircuarity-rate maximization: a new approach to adaptive bind beamforming, in Wireess Communications, Networking and Mobie Computing, 29. WiCom 9. th Internationa Conference on, pp. 1 4, IEEE, 29. [19] S. Chen, L. Hanzo, N. N. Ahmad, and A. Wofgang, Adaptive minimum bit error rate beamforming assisted receiver for qpsk wireess communication, Digita Signa Processing, vo. 1, no. 6, pp. 4 67, 2. [2] R. Haupt and H. Southa, Experimenta adaptive nuing with a genetic agorithm, MICROWAVE JOURNAL-EUROGLOBAL EDITION-, vo. 42, pp , [21] A. Massa, M. Donei, F. G. De Natae, S. Caorsi, and A. Lommi, Panar antenna array contro with genetic agorithms and adaptive array theory, Antennas and Propagation, IEEE Transactions on, vo. 2, no. 11, pp , 24. [22] Y.-j. Lee, J.-W. Seo, J.-K. Ha, and D.-c. Park, Nu steering of inear phased array antenna using genetic agorithm, in Microwave Conference, 29. APMC 29. Asia Pacific, pp , IEEE, 29. [23] D. J. Sader, Panar array design for ow ambiguity, in Antennas & Propagation Conference, 29. LAPC 29. Loughborough, pp , IEEE, 29. [24] H. Evans, P. Gae, B. Ajibouri, E. Lim, E. Korokwiwicz, and A. Sambe, Appication of simuated anneaing to design of seria feed sequentiay rotated 2 2 antenna array, Eectronics Letters, vo. 36, no. 24, pp , 2. [2] S. Ram, A Study of Adaptive Beamforming Techniques Using Smart Antenna For Mobie Communication. PhD thesis, Nationa Institute of Technoogy Rourkea, 27. [26] O. Bazan and M. Jaseemuddin, A survey on mac protocos for wireess adhoc networks with beamforming antennas, Communications Surveys & Tutorias, IEEE, vo. 14, no. 2, pp , 212. [27] A. A. Ansari, Z. Hasan, M. J. Aam, K. Mohammad, and A. Siddique, Performance comparison for omnidirectiona and directiona mac protocos for ad hoc network, [28] F. Bai and A. Hemy, A survey of mobiity modes, Wireess Adhoc Networks. University of Southern Caifornia, USA, vo. 26, 24. [29] O. Bouachir, A. Abrassart, F. Garcia, and N. Larrieu, A mobiity mode for uav ad hoc network, in ICUAS 214, Internationa Conference on Unmanned Aircraft Systems, pp. pp 383, 214.