Maximize resource utilization based channel access model with presence of reactive jammer for underwater wireless sensor network

Received Jun 4, 2019 Revised Dec 11, 2019 Accepted Jan 8, 2020 Underwater sensor networks (UWSNs) are vulnerable to jamming attacks. Especially, reactive jamming which emerged as a greatest security threat to UWSNs. Reactive jammer are difficult to be removed, defended and identified. Since reactive jammer can control and regulate (i.e., the duration of the jam signal) the probability of jamming for maintaining high vulnerability with low detection probability. The existing model are generally designed considering terrestrial wireless sensor networks (TWSNs). Further, these models are limited in their ability to detect jamming correctly, distinguish between the corrupted and uncorrupted parts of a packet, and be adaptive with the dynamic environment. Cooperative jamming model has presented in recent times to utilize resource efficiently. However, very limited work is carried out using cooperative jamming detection. For overcoming research challenges, this work present Maximize Resource Utilization based Channel Access (MRUCA). The MRUCA uses cross layer design for mitigating reactive jammer (i.e., MRUCA jointly optimizes the cooperative hopping probabilities and channel accessibility probabilities of authenticated sensor device). Along with channel, load capacity of authenticated sensor device is estimated to utilize (maximize) resource efficiently. Experiment outcome shows the proposed MRUCA model attain superior performance than state-of-art model in terms of packet transmission, BER and Detection rate.


INTRODUCTION
Underwater wireless sensor network play a major role across various application services in offering ubiquitous assess such as weather forecasting, marine safety, environment etc. where sensor devices are placed across environment to offer continuous connectivity and services. Thus, aid in improving quality of humans life. However, traditional wireless network can easily compromised by jamming technology. This is due to exposed nature of wireless links. Jamming can induce attack [1] such as Denial-of-Service (DoS) attack, Sybil attack etc. affecting performance of UWSN [2,3]. Jamming in UWSN can be defined as the interference induced in existing wireless network communication by malicious sensor nodes by decreasing the signal-to-noise ratio (SINR) of the authenticated sensor device (receiver side) by transmitting interfering wireless signals. Jamming is different from regular noise or interference because it is a resultant of deliberate use of wireless signal to degrade network performance whereas as interference is an unintentional forms of noise disrupting performance of UWSN. Unintentional interference caused in network is due to The research contribution of this work are as follows -Firstly, this work presented maximize resource utilization based channel access model for UWSNs. -Presenting a novel cross layer design for cooperative communication (among MAC and physical layer) to detect jammed node and utilizing spectrum efficiently. -Presenting channel load capacity of authenticated sensor for maximizing resource utilization without affecting adjacent contending sensor device. -Experiment outcome shows, the proposed MRUCA model attain superior performance than existing model in terms of bit error rate, detection rate, packet sending ratio, and slot utilization considering grid and random topology deployment.

LITTERATURE SURVEY
This section present extensive survey on provisioning security and addressing security issues in underwater wireless sensor network and identified research issues to model an enhanced secure and efficient resource allocation model for UWSN. In [19], showed MAC protocol is a key element in UWSN similar to terrestrial network. However, UWSN has unique feature such as, low channel reliability, very small channel capacity, high dynamics of channel quality, and long propagation delay. Thus, MAC design modelled for terrestrial network cannot work well for UWSN. Here they conducted extensive survey of various MAC design proposed in recent times for building enhanced MAC model. Further, major remaining issues and possible research directions are also discussed. In [20], showed for prolonging lifetime of UWSN, two factor such packet size and transmission power plays vital factor. At one hand, smaller packet are more robust to packet error when compared with larger packets. Thus, using smaller packets aid in reducing bit error. However, it requires larger frame for transmission and hence, induce energy and network overhead. For minimizing frame error, transmission power can be increased. However, this will result in unnecessary energy dissipation in the network. Thus, it is important to consider both packet size and transmission power for enhancing lifetime of network. Here, the presented optimization model using integer linear programming to maximize lifetime of network considering both packet size and transmission power. Along with, a realistic link-layer energy dissipation model is presented using physical layer features of UWSNs.
In [21], showed that channel shared among contending sensor device to utilize resource efficiently. However, sharing channel are prone to impersonation and various other kind of attacks [22]. In [21] presented a spatial reuse based resource allocation model for UWSN for avoiding destructive collision. Major cause of such collision is due to near-far effect [18] where sensor device placed faraway from receiver is jammed by a closer sensor device. Here, they considered spatial reuse time-division multiple access (TDMA) for increasing throughput. They adopted both opportunistic and contention free. Their main objective is to guarantee per-node packet transmission rate and maximize time slot (resource) allocation. Their model increases contention free packet transmission, and decrease scheduling delay of opportunistic packets. However, it induces collision among neighboring contending device. In [15], showed that UWSN packets rarely include encryption due to physical and performance limitations. Thus, UWSN is exposed to various kind of security attack breaching legitimate message. Here they presented an algorithm for message authentication in an UWSN environment. Further, observed that an attacker can impersonate the channel associated with the authenticated sensor device only for a single or certain set of receiving sensor device. This is due to strong spatial dependency of the UWSN channel. Considering these observation, they presented a model using cooperative strategy among trusted sensor device toward base station or sink. For each incoming message, the sink fuses beliefs computed by the trusted sensor device to reach an authentication decision. These beliefs are computed by estimating statistical channel parameters, preferred to be the most sensitive to the communicating device movement. Outcome shows accurate identification of an attacker's packet.
In [17], outlined a hybrid design that is composed software defined network, physical layer security, cross-layer design, cognition, node cooperation and context-awareness. They envisioned a security model at both network as well as at the node level that adapt to dynamic environmental condition, the status of the network, and possible wide range of attacks or security breaches. Here they discussed several kinds of attacks, security breaches and countermeasures along with implementation, deployment and functionality issues and challenges of building hybrid security model for UWSN. The main focusses of their model design is to suggest future research direction to research community or organization working on UWSN. In [18], presented topology-efficient discovery model for UWSN. Here they used network information of source and destination sensor device for performing routing and scheduling packet transmission. They aimed to assure better convergence time in completing topology discovery and the network transforms to its steady-state scheduling design. For meeting, they aimed to assess the link reliability and to identify acoustic link. Their method allow sensor device to share time slots while minimizing/controlling the potential collision to reduce overhead and delay in topology discovery process. Further, it offers power control mechanism among near-far node pairs (NFNPs) to improve spectrum utilization (.i.e. offers spatial reuse). Though, their method offers better spectrum utilization with minimal topology discovery time. However, without proper scheduling and delays in transmissions, many packets still collide. From extensive survey it can be seen the spatial reuse mechanism has been widely applied across state-of-art method to utilize resource efficiently. However, it induces various security issues. Further, sensor node cooperatively transmit among adjacent sensor device to minimize energy consumption and utilize spectrum resource efficiently. However, these model do not consider physical layer information into consideration. Thus, affecting network performance. Along with, considers scheduling of channel to utilize resource efficiently. However, when user are selfish it incurs collision overhead among adjacent sensor device. Further, very limited work is carried out for detecting reactive jammer. Therefore, there is a requirement for new model that that detect jamming effectively and at the same time utilize resource efficiently. This work present a maximize resource allocation based channel allocation model with presence of jammer node for underwater sensor network.

MAXIMIZE RESOURCE UTILIATION BASED CHANNEL ACCESS MODEL FOR UNDERWATER WIRELESS SENSOR NETWORK
This work present Maximize Resource Utilization based channel access (MRUCA) model for underwater wireless sensor network (UWSN) with the presence of jamming sensor device. For maximizing resource utilization without affecting adjacent contending device a channel access model adopting cross layer design (i.e., cooperative communication among physical and MAC layer) is presented. Firstly, the system model is defined. Then, the reactive jamming model adopted for research work is described. Then, the cross layer design of physical and MAC layer is presented. Further, the method for identifying authenticated sensor device is presented. Along with, cross layer based channel access model is defined. Lastly, channel load capacity is estimated to maximize resource utilization using either direct or through hop based transmission.

System model
Let's consider an underwater wireless sensor network that is composed of set of authentic sensor device. Further, each sensor device is composed of source-destination pair of devices. Considering this scenario, the wireless sensor network can be seen as a set of concurrent device-to-device communications. This work assumes that the communicating sensor device queues are always flooded/backlogged, then we can describe each transceiver pairs as a session. We describe the transmitting and receiving devices of each session ∈ as ( ) and ( ), respectively. Further, for performing transmission by authentic sensor device there are set of orthogonal frequency channels.

Reactive jamming model
Let's consider that there is one jammer device which has limited power constraint and tries to degrade the throughput performance of authentic sensor device by generating interference on the accessible channel. Further, this work considers that the jammer can emit wideband interference simultaneously across all the accessible channel. The jammer power allocation strategy is denoted as follows: where is the power given on channel , thus we obtain, where 1 depicts an 1 * | | vector of '1', and ↑ is the maximum power of the jammer. Due to the diversity of frequency channels, the jammer must assign its power constraint in a way that aid in attaining good jamming effect.

MAC layer model
The cooperative (mutual) jamming detection model is designed based on mutual functional computation among physical and MAC layers. In MAC layer, sensor device controls their channel accessible probability to attain higher chances to communicate to devices that are being jammed. For satisfying, an opportunistic spectrum access (OSA) model with adjustable spectrum access probability is required at MAC layer. For such consideration, slotted multichannel CSMA is considered where sensor device can adjust the channel access probability. Our MAC model is similar to state-of-art model (i.e., it is based on contention) where communication time is divided into set of time slots, and all sensor device is considered to be synchronized. At each time slots, a sensor device at most select only one channel at a time for performing transmission, similar to frequency hopping methodology. However, the channel is arbitrarily selected (i.e., a sensor device select each channel with certain probability). A sensor device is also permitted to select none and in this scenario the sensor devices acts as hop device for other contending sensor device. If a sensor device select channel ∈ , it initially sense the channel at the start of the time slots, to decide if the channel is accessible. Contention may arise as there may be multiple contender (sensor device) selecting the same channel. Similar to CSAM, a contending sensor device set arbitrary back off time and initialize counter, and the first device counter turns to zero win the contention for channel access. For easiness and attain negligible collision probability, the contention window is kept feasibly large. The proposed MAC design, allows the sensor device to alter it channel accessible probability by just fine-tuning the channel sensing probability of different channel. Thus, let , ∈ depicts the channel sensing probabilities of sensor device on channel . Further, the sensor device may induces delay of its transmission for serving as hop device for other senor device due to nonzero probability of sensor device . Thus we have:

Physical layer model
Physical layer information is received through hopping. Rather than transmitting its own traffic, a sensor device can behave as a hop device and cooperatively communicate a packet on behalf of another sensor device. Cooperative communication is attained by distributing the accessible communication time into two stage. Firstly, the source (sender) broadcast the packet (information) to both the hop device and the receiver (destination). Secondly, the hop device forward the obtained packet to the receiver, which then cumulate these packets and perform decoding. In this work, we consider decode-and-forward based cooperative transmission, under which the hop device forward the message only when the packet collected from the source devices can be decoded successfully. To handle with such dynamic behavior of the jammer, we consider a dynamic hop device selection method policy to allow the sensor device form virtual MISO links. At each time slots, a sensor device that decide not to perform sensing any frequency channel will behave as a hop device for other sensor device if there is an optimistic cooperative gain.

Strategy of authenticated sensor device
The behavior of authentic sensor device is obtained as follows. When a sensor device is flooded, it choses its channel sensing probability for each channel. The sensor device will serve as a potential hop or cooperator for other authenticated sensor devices if it decide not to sense any channel. Naturally, the cooperative hoping transmission probability is a function representation of the channel sensing probability , of the policies of jammer , and the wireless sensor network topography. The sensor device that decide to sense the same channel by initializing an arbitrary back off and the sensor device with maximum probability will obtain contention for transmission. The feature describing the cooperative transmission at both layers, i.e., channel accessible probability and cooperative hoping probability, are both functional representation of the channel accessible probability, network topography, and for a respective jammer policies. Thus, we can utilize channel accessible probability as the policy space of an authenticated sensor device and can be represented as follows: with, where depicts the sensing probability of channel , and 0 depicts the probability that doesn't sense any of the channel. Then, it must satisfy following condition: ≤ 1, ∀ ∈ , ∀ ∈̃ (6) 1 = 1, ∀ ∈ Further, the sensing probability strategy of all sensor device in is expressed as follows: and similarly, the sensing probability strategy of all user except for can be obtained as follows:

Channel accessible probability evaluation
Let us assume that, the estimated size of a authenticated sensor device ∈ , is expressed as follows: where ( , ) is the probability that sensor device is able to resourcefully access the channel, is the probability that sensor device senses frequency channel, and ( , ) is the attainable size on that channel (through either using a cooperative hop device or by using direct transmission), for a given jamming power strategy and sensing probability strategy . As described above, MAC layer cooperative transmission is attained through stochastic channel access. That is, the sensor device ∈ is able to resourcefully access frequency channel ∈ if session wins the channel accessible competition and the channel sensed is idle at session 's transmitting sensor device ( ). Let ̂( ) depicted as the probability that session wins the contention game and ̃( ) be represented as the probability that channel is idle, the channel accessibility probability ( , ) in (10) can be computed as follows: If we consider ℎ as power threshold below which a frequency channel is sensed idle, thus ̃( ) can be expressed as follows: where ( ) , ( ) is the capturing fading and path loss component of the link among jammer and session 's transmitting sensor device ( ) on frequency channel , respectively, and ( ( ) ) 2 is the noise power.
As ( ) Rayleigh distributed with fading component ( ) , the ̃( ) in (11) can be rewritten as follows: Using (12), ↑ can be computed as follows: We further compute the probability that a sensor device ∈ wins the medium accessible game post performing sensing the frequency channel ∈ to be idle. Let consider set of sensor device contending with sensor device on frequency channel represented as ⊂ ⁄ , the winning probability for sensor device can be expressed as follows: where | | is the number of sensor device in . Since each probable contending sensor device ∈ joins the access contention game with probability ̃( ), the cardinality of , that is | | is considered to be Poisson distributed with mean and is expressed as follows: (| |) = ∑̃( ).
∈ ⁄ Then, the cumulated probability of establishing a channel access contention game for sensor device , that is, ̂( ) in (11), can be computed as follows:

Estimated channel load capacity
Let us assume that ∈ sensor device won the channel access game to perform transmission on channel . Now we express the expected channel load capacity attainable using direct communication i.e., ( , in (10) is computed as follows: where is the data rate of each channel, and , 2 ( ) is computed as follows:  (10) can be computed. Let us assume that each source device ∈ ⁄ functions as a possible hop device with probability 0 . Thus, with respect to certain probability, sensor device will obtain cooperative gain by one of the potential cooperative sensor devices. In such case, a sensor device selects ( ) as the hop device, then, the resultant cooperative capacity can be expressed as follows: where , 2 = , 2 ( ) and , 2 = , 2 ( ) depicts the signal to noise ratio of the link among transmitter to hop device and hop device to receiver, respectively. An important thing to be noted here is, from (18) and (19) the cooperative capacity can be lower or higher than the direct capacity. This is due to ½ coefficient consideration in (19). Thus, the overall estimated capacity attainable by sensor device over channel can be computed as follows: Form (20) it can be seen, the above equation can be satisfied only when the probability that more than one cooperative sensor device joins in cooperative transmission is very low. Otherwise, the capacity function will be computed as a summation of the estimated cooperative capacities offered by different cooperative hop device. Further, this work aims to utilize resource efficiently without affecting other contending sensor device. Thus, the utility parameter of each sensor device can be expressed as follows: and the proposed objective parameter to maximize the resource utilization of sensor device without affecting other legitimate sensor device can be expressed as follows: : Maximize : (5), (6), (7) This work present a distributed strategy to meet proposed resource utilization objectives of (22). This work adopt an iterative fine-grained (best response) model using cost parameter. At every iteration, each session tries to maximize its objective parameter minus a cost factor that acts as a penalty incurred/levied to each contending session for being too selfish in selecting its own policies and thus affecting other contending sessions. Since this work assumes that for each authenticated sensor device for Int J Elec & Comp Eng ISSN: 2088-8708  which the policy of the jammer, that is, is a given parameter, for easiness, it is not considered in from the objective strategy. The proposed channel access model attain superior performance than state-of-art model which is experimentally proved in next section below.

SIMULATION RESULT AND ANNALYSIS
This section present experiment analysis of proposed maximize resource utilization based channel access (MRUCA) model with and without presence of jammer nodes and compares the outcome with state-of-art model [18]. For conducting experiment analysis, this work used MAcoSim [23][24][25]. MAcoSim is operated by both GUI and MATLAB command line interface and is written on top of NS2 simulator. Additionally, centralized parameter manger was employed to acquire easy configuration. The output trace is written in NAM file which can be used later to analyze the simulation experiment. The parameter considered for experiment analysis is similar to [18]. This work considered a topology with 8, 12, and 16 sensor device placed randomly in a region of 16m*16m with one jammer node. Along with experiment are conducted considering topology with 28, 32, and 36 sensor devices placed in grid region of 16m*16m with presence of singe jammer node. Jammer will send 8-bit packet in each time slot and each node will generate traffic (transmit) 3 bit of data. Experiment is carried out considering 100 simulation cycles and the outcome is logged in terms of packet received, contention packet received, amount of packet being dropped, bit error rate, packet sending ratio and bit error rate.  and 16 nodes, respectively. Out of which is 76, 80, and 78 packet is identified and dropped from the network considering 8, 12, and 16 nodes, respectively. Thus, the MRUCA model attain a detection accuracy of 96.69%. In Figure 4 packet sending ration performance is shown. From result, it can be seen without jamming our model attains 100% packet sending ratio. However, with presence of jammer node, an average of 97.02% packet sending ratio is attained considering varied nodes. From overall result attain it shows that the proposed MRUCA model attain superior performance (less collision) when compared with existing model [18].   Figure 5 shows the packet transmission performance attained by proposed MRUCA model considering with and without jamming under varied nodes (i.e., 28, 32, and 36). From Figure 5 it can be seen without jammer no packet drop can be seen. However, with presence of jammer it can be seen 25.33 packets been dropped. However, these packet are been corrupted by the jammer node. As a result, are identified and eliminated from network. Thus, will aid in reducing congestion in network. Further, with presence of jammer only 2 packet is been retransmitted. However, in case of no jammer 4.66 packet has been retransmitted. Further, from Figure 6 the proposed model attain 92.26% and 100% packet sending ratio performance with and without presence of jammer, respectively. An average drop or underutilization of spectrum of 7.74% is induced due to presence of jammer nodes considering varied nodes size. Further, experiment is conducted to evaluate the Bit Error Rate (BER) performance. The outcome shows proposed model with presence of jammer attain BER of 0.0588, 0.0588, and 0.0571 for 28, 32, and 36 sensor device, respectively with signal to noise ratio (SNR) = 4dB.

Result and discussion
This section present performance evaluation discussion of proposed MRUCA over existing model [18]. Table 1, shows result attained by proposed MRUCA over existing model. In [18] evaluated their model using utility function (i.e., slot utilization), near far node pair (NFNP) detection accuracy (i.e., detection rate) and packet collision (drop rate). However, packet transmission performance is not evaluated by them. From overall result attained by existing model shows they reduce topology discovery time. Further, adoption of spatial reuse aided in attaining better resource utilization. However, packet still get collided due to improper scheduling and transmission delay. In other side the proposed model attain better slot utilization, packet drop rate, and detection rate which was experimentally proven above. The significant result attained by proposed model is due adoption of cross layer where the model jointly optimizes the cooperative hopping probabilities and channel accessibility probabilities of authenticated sensor device). Along with channel load capacity of authenticated sensor is estimated for maximizing resource allocation without affecting neighbouring sensor device. Table 1. Performance comparison of proposed model MRUCA over existing model [18] Proposed

CONCLUSION
Firstly, this work conducted extensive survey to identify issues and challenges in designing efficient channel access model under presence of jammer node. Further, identified security and performance venerability of UWSNs with presence of jamming node. Among jamming attack, reactive jamming attack is considered to be very difficult to identify and remove it from the network. Number of approaches has been presented to prevent such attack in UWSN. However, the existing model failed to distinguish between the corrupted and uncorrupted parts of a packet. As a result are not efficient in identifying jamming nodes. Thus affecting network performance (i.e., induce bandwidth wastage). For overcoming research challenges, this work presented maximize resource utilization based channel access model adopting cross layer design. A novel cooperative scheduling mechanism is presented that mitigate jamming nodes and aid in better resource utilization by jointly optimizing the cooperative hopping probabilities and channel accessibility probabilities. Further, for maximizing resource utilization without affecting performance of adjacent contending sensor device a channel load capacity is estimated using either direct or through hop based transmission. Experiment are conducted to evaluate performance of MRUCA over existing model considering both random and grid topology deployment. The outcome shows the MRUCA attains good bit error rate performance with 6.14% and 7.74% packet drop due to presence of jammer nodes considering random and grid deployment, respectively. Similarly, 3.22% contention packet drop is attained due to presence of jammer node under random deployment. The MRUCA attain 97.02% and 93.86% packet sending ratio and slot utilization efficiency, respectively with detection accuracy of 96.69% under deployment. Similarly, under grid environment with presence of jammer model the proposed model attain 92.26% packet sending ratio performance. The overall result attained shows superior performance than state-of-art model. The future work would consider presenting distributed resource allocation scheme to support diverse real-time application under UWSNs.