WNBLI - A Multifarious Liability Examination for Wireless Sensor Networks

: Wireless communications has completely revolutionized the living environment, just as personal computers forever altered the way of working in the 1980’s and the Internet dramatically changed the obtainment and accessing way of information in the 1990’s. Wireless sensor networks are evolving to a new level. The Failure liability is a very prominent problem in Wireless Sensor Network (WSN) research. Failure liable nodes behave in different form and this faulty behavior is referred to as multifarious faults or heterogeneous faults in WSN. This paper presents a Multifarious Failure liability Investigation protocol called Wireless Nodal Behavior and Liability Investigation (WNBLI) protocol for wireless sensor networks. The proposed WNBLI protocol consists of three vital phases, such as Association phase, Apprehension phase and Assortment phase. The performance of the proposed Wireless Nodal Behavior and Liability Investigation protocol is evaluated using network simulator.


Introduction
Wireless Sensor Networks (WSNs) play a very important part in implementing a technology because they require low deployment funds. The deployment of WSNs requires less effort and it is versatile in nature. The power consumption by the nodes in the WSNs is a major problem. This power constraint issue must be eradicated to deploy fully converged WSNs. The power consumption is attributed to two main factors namely transmission flow and data packet generation. The consumption of power can be minimized by taking control of the transmission and optimizing the data packet size. If the transmission rate is increased, it will have a direct proportion to the bit error rate suppressing factors. On the other hand if more power is utilized, then it will drain the power resources. If the route path involved in the transmission uses less power and can achieve the required rate of transmission, then the utilization of high power rate is unnecessary. The size of the data packet in the transmission is directly related to the power allotment from the power energy source. Thus packet size plays a vital role in the data transmission optimization (Kurt et al., 2017). The nodes in the WSNs consume energy depending upon their utilization. This utilization factor is dependent upon the load of data packets that is transmitted and the node position. If the node is located in the best route path, the transmission factors will be high whereas less interested path require less or sometimes null energy. The power factor of the nodes depends upon the area of deployment. If the nodes have to collect data all day long for everyday, say years, then the power backup for the nodes must be taken into account. If the power supply is negligible, then the nodes operating in such areas will suffer a power consumption defect. The node efficiency is dependent on the power consumption which is directly proportional to the load of transmission. If the nodes are located in areas inaccessible to humans, then it will cause a seizure of node operations, thus making it vulnerable to various layers of attacks (Jan et al., 2017).
Wireless Sensor Networks (WSNs) have attracted lot of attention due to their pervasive nature and their wide deployment in Internet of Things, Cyber Physical Systems and other emerging areas. The limited energy associated with WSNs is a major bottleneck of WSN technologies (Shaikh and Zeadally, 2016). Recent breakthroughs in wireless technologies have greatly spurred the emergence of Industrial Wireless Sensor Networks (IWSNs). To facilitate the adaptation of IWSNs to industrial applications, concerns about networks' full coverage and connectivity must be addressed to fulfill reliability and real-time requirements. Although connected target coverage algorithms in general sensor networks have been extensively studied, little attention has been paid to reveal both the applicability and limitations of different coverage strategies from an industrial viewpoint (Han et al., 2017). The growth in technology paved way for the advancement in Multimedia Wireless Sensor Networks (MWSNs). The main problem in deployment of MWSNs is the security and trust establishment between necessary nodes. The privacy factors also play a very crucial role in the MWSNs. The user authentication and authorization part, that is employed in the MWSNs are much different from the conventional sensor network topology. The limitation factors in MWSNs range from high power consumption to optimal route path selection failure due to the inability to separate the failure node points from the efficient nodes while determining the best route path (Usman et al., 2016). There are a number of security issues in the WSNs and the one that overwhelmingly studied and attracted the researcher in the recent years is the key management issues of network. Though researchers have proposed various key management schemes to enhance the secure information transfer, the existing works lack in the tradeoff between its efficiency and the cost in terms of the factors like structure formulation, energy utilization and key storage overheads (Swaminathan and Vivekanandan, 2017;Rault et al., 2014;Li et al., 2013). Wireless sensor networks can be deployed in any attended or unattended environments like environmental monitoring, agriculture, military, health care etc., where the sensor nodes forward the sensing data to the gateway node. As the sensor node has very limited battery power and cannot be recharged after deployment, it is very important to design a secure, effective and light weight user authentication and key agreement protocol for accessing the sensed data through the gateway node over insecure networks. Most recently, Turkanovic et al. proposed a light weight user authentication and key agreement protocol for accessing the services of the WSNs environment and claimed that the same protocol is efficient in terms of security and complexities than related existing protocols (Gutierrez et al., 2014;Nikolidakis et al., 2013;Li and Lin, 2015;Amin and Biswas, 2016).

Related Work
Research works pertaining to the concept of improvising the performance and efficiency factors of the wireless sensor networks have been published. In all the works, the study was focused on two main aspects which is network topology and best route path discovery.
All the works have their own delicate point of supremacy which excels them for future study. Amin and Biswas (2016) designed a novel architecture for the Wireless Sensor Network environment and basing upon which a scheme has been proposed and presented for user authentication and key agreement scheme. The security validation carried out in the proposed work follows a session management and key authorization methodology. Zhang et al. (2016) proposed an asymptotically optimal algorithm based on the dual decomposition method and a sub optimal algorithm. This sub optimal algorithm was done with lower complexity. The working nature of the algorithm differs in the operational key management status which ensures maximum performance. Baroutis and Younis (2017) presented a technique for preserving privacy with location attribute, which is the in-situ base-station. The technique injects deceptive transmissions aiming to even the traffic abundance across the network. The highlight of the proposed work is the trade-off between location privacy and network's performance. Prabhu et al. (2016) proposed a detailed study of various distributed clustering approaches in the wireless sensor networks. This detailed research study is made on optimized cluster initialization based on jumping ant approach in order to avoid random cluster development. The operational status of the cluster heads is given more importance implicating its performance. Gu et al. (2016) research work aims to address the gaping issues on the sink mobility problem occuring in the mobile networks. The way of depiction of the work is similar to the bisection of wireless sensor networks. The schemes proposed in this research work are divided into four categories: Uncontrollable Mobility (UMM), Path-Restricted Mobility (PRM), Location-Restricted Mobility (LRM) and Unrestricted Mobility (URM).
Weber (2017) proposed a multi-round Aloha-based protocol which illustrates the network topology as the sending and receiving node ends. The receiving nodes collect data from the transmitting nodes. The protocol takes the working part of the nodes performance into account and generates metrics which acts as a base for the node performance evaluation. The concentration of the nodes existence and the performance attribute collectively form the metric analyzing factor. Prathima et al. (2016) proposed a Secure Data Aggregation for Multiple Queries (SDAMQ) in Wireless Sensor Networks. In this proposed work many links of nodes are formed into a aggregate queries from the depression node and are authenticated, then listed to the sensor nodes. The sensor nodes respond by collecting these data and making them into a single packet, thereby reducing the transmit metric factor. The proposed model consumes less energy. Njoya et al. (2017) proposed and implemented an algorithm which is performance efficient and scalable. The proposed algorithm creates virtual sensors which reduces the need of physical sensors. This deployment of virtual sensors creates an array of functional nodes which carries the data from the source to the destination in a controlled way depicting a secure transmission process. The performance of the virtual sensors is similar to the physical nodes employed and the working part of the data transmission unit of the virtual nodes is scalable in both operations and fail safe nature.

Understanding the Structure of the Model
The following structure describes the Inferences, Network representation and Flow of Communication.

Inferences
The following inferences must be taken into account: • The wireless sensor nodes are static and analogous in nature • Each nodes in the WSN has an Unique Identification (Node-UID) • There are supervisory nodes which are fail safe and are more efficient than other nodes • The deployment of the supervisory nodes is controlled in the initial state but after the flow of communication happens, the node positions and their contributions are dependent upon their working functionality • There are virtual links created to maintain loop free communication

Network Representation
The wireless sensor network can be represented as a graph GR (nn,bin) where nn is the number of nodes and bin is the bay area in the network. The node a , node b Є GR(nn) can communicate, if there is a virual link existing bin a,b Є GR(bin) between the two nodes. In the proposed methodology, NUM numbers of nodes are distributed in a region of REG 1 x REG 2 which is greater than the transmission (Tran dis ) range of the wireless sensor nodes. The virtual link bin a,b between the two nodes node a and node b exist, if the Euclidean distance between them is less than their transmission (Tran dis ) range of the wireless sens Li, S., Da Xu, L. and Wang, X. (2013). Compressed sensing signal and data acquisition in wireless sensor networks and internet of things. IEEE Transactions on Industrial Informatics,9(4), 2177-2186. Let (j a ,k a ,l a ) and (j b .k b .l b ) are the Cartesian co ordiantes of the sensor nodes node a and node b respectively. Therefore, the Euclidean Distance (EU-DIS(.)) between the two wireless sensor nodes are calculated by: From the above equation, virtual link bin a,b exists between the two nodes node a and node b , if EU-DIS(node a, node b ) ≤ Tran dis , where Tran dis is the transmission range of the static and analogous nodes. Again all the virtual links of this network graph GR(nn,bin) are undirected and it is represented as the following equations: Lapsing Period

Flow of Communication
The entire flow of communication between the nodes is clearly depicted in the following Fig. 1. Initially the nodes are aggregated to form a cluster. This cluster then organizes itself in a certain period of time. After the cluster organization, the nodes begin to communicate between them. This nodal communication phase is prominent and the communication flow of the nodes are detailed and initiated to the aggregate communication. The flow of information happens between aggregates and also within aggregates to provide enough information on the information flow.

Proposed Work
The proposed Wireless Nodal Behavior and Liability Investigation (WNBLI) protocol for wireless sensor networks is accomplished mainly in three major phases, such as: (i) Association phase, (ii) Apprehension phase and (iii) Assortment phase.

Association Phase
In this phase the nodes involved in the transmission process is analyzed deeply and is given a node identity which is unique in nature. The nodes are grouped together into a group which contains all the active nodes in them like NN g = {node 1 , node 2 , node 3 ,..node n }. There are numerous clusters that are formed as group of association head nodes where the number of nodes is always more than the association head nodes count like NH g = {nh 1 ,nh 2 ,nh 3 ,….nh m }. The nodes which falls under the transmission range of the Base supervisory node is taken into account and the actual count of the nodes participating in the transmission is noted down. The Base transmission node will always communicate with the supervisory node and the number of nodes involved with the supervisory nodes is taken in as an unit and this unit is unique to that region of transmission. The active nodes are associated with a supervisory node and this count is given by NH g = {nh 1 } where NH g is the count of association head nodes which contains the first set of associated nodes nh 1 .This process of Association will take place till all the active nodes are assigned to a supervisory node. The node distribution is depicted clearly in the Fig. 2 given below. The working nature of the nodes with the supervisory nodes is directly proportional to the transmission range of the nodes with the supervisory node and also the range up to which the supervisory node can communicate with the base transmission node. (1) Initialized group of nodes: NN g = {node 1 , node 2 , node 3 ,….node n } (2) Initialized group of association head nodes: NH g = {nh 1 ,nh 2 ,nh 3 ,….nh m } where n > m (3) Initialized nh(node u ): Set of all nodes which falls within the transmission range of the node node u ∈ NN g (4) Initialized dis (u,v) which gives the distance between node u and node association head nh v (5) Assign the node node u ∈ Nn g to corresponding node association head group nh v ∈ nh(node u ) based on dis(u,v) (6) Perform the Association technique based on allotment made with association head node groups NH g = {nh 1 ,nh 2 ,nh 3 ,….nh m } (7) while NN g ≠ Null (8) Choose a association head node group nh v from association from NH g (9) Assign the nodes node u to nl v such that nh v ∈ nh(node u ) and node u is closer to nh v (10) Delete node u from NN g (11) Update the association array NH g so that the load is minimum to the nh v which is the next association head node group element (12) end while (13) Stop

Apprehension Phase
After the Association phase, the nodes that are defective are identified and suitable measures are undertaken by reacting to the imprecise node activity. In the Apprehension phase, the inaccurate nodes are identified. This phase is divided into two states depending upon the imprecise node characteristics. They are dense defective state and Elastic defective state.

a) Dense Defective State
In this state the permanently erroneous nodes are identified. In a node communication topology, there will be a base transmission node and this node communicates with all the supervisory nodes. The supervisor nodes present in the topology maintains a table called as Node Administration Table [NADT]. This NADT maintains all the information about the nodes connected to its domain like node state (active or passive), node activity (transmitting or dormant), node positions, etc.,. The status of the node (node i ) belonging to the group (NN g ) at the time (tp) is denoted as (nod-stat tp i ). The possible values are 1,2 and 3 where 1 is node functioning normally, 2 is nodes that are likely to be unreliable and 3 is node that are completely erroneous.
Initially the nod-stat tp i value is 1 as all the nodes belonging to that domain are considered to be functioning normally. The supervisor nodes will send Hello messages to all the participating nodes in its domain for each time period of tp. After receiving the Hello message, the participating nodes send Acknowledgment message (n-ACK) back to the supervisory nodes. The Acknowledgment messages (n-ACK) are received within a time period tp out . If the n-ACK messages is n not received within the time period tp out , then the node status value nod-stat tp i of the NADT of the supervisory node is changed to 2 which means the node is likely to be unreliable and can become erroneous some time soon. The total time instance (tp) depends upon various aspects. The time out period tp out is the total time taken for the complete exchange of a data packet and its corresponding acknowledgment packet. The time out period tp out for node node i belonging to the group NN g [node i ∈ NN g ] is given by the formula: Where: tp propd = Data proliferation delay tp trand = Transmission delay-aware period tp queued = Queinng delay-aware period tp processd = Processing delay period

b) Data Proliferation Delay
The data proliferation delay tp propd for node i in the data transmission between node i and node k given by: where, m is the data proliferation speed in the given data transaction.

c) Transmission Delay-Aware Period
The transmission delay-aware periodtp trand for node i in the data transmission between node i and node k given by: where, ln(pack i ) denotes the data packet length for node i and cbw denotes the channel bandwidth of the propagation that is happening.

d) Queinng Delay-Aware Period
The queinng delay-aware period is the time taken by the transmitted data packet to wait till it awaits its turn to get processed.

e) Processing Delay Period
The processing delay period is the time period taken for processing the data in the router.
At any given point of time, the value obtained for Queinng delay-aware period (tp queued ) and Processing delay period (tp processd ) is very negligible. So the values for tp queued and tp processd are considered to be zero. So taking these considerations, Equation 4 can be reprocessed as: The channel access scheme of Time Division Multiple Access based Media Access Protocol is used where the slot duration is set as tp out . The nod-stat tp i value of node (node i ) for time instance tp = {1,2,3,4,5….m….tp} is computed and the status of the node (node i ) is given by: Let the node node i is considered as unreliable node by the Node Administration table [NADT] which is obtained by checking the nod-stat tp i value that is computed. If the value is 50% or more than 50% in failing to respond, then it is considered as completely erroneous node with tp denoting the total time slot is given by the equation:

f) Elastic Defective State
In this state the elastic nodes, that is nodes with little faults are identified and the operational efficiency of the nodes are taken into account. Initially each sensor node node i belonging to the group NN g [node i ∈ NN g ] sends its sensed value sen j , j = {1,2,3….m} to its association head node nh k ∈ NH g in its transmission range. The supervisor nodes detects the presence of erroneous nodes in its domain. A statistical based analysis method namely Analysis of Variance (ANOVA) method (O'brien, 1979;Cvijović et al., 2005;Zhang et al., 2012;Ozcelik and Erzurumlu, 2006;Azadeh et al., 2007) is used to analyze the actual sensor values and the erroneous sensor values and the unreliable nodes are picked out. The ANOVA method follows two types of hypothesis for testing namely Null Hypothesis (HY 0 ) and Alternative Hypothesis (HY 1 ). The Null Hypothesis (HY 0 ) states that there is no significant difference between the values given by the transmitting nodes. The Alternative Hypothesis (HY 1 ) states that there is at least one significant difference between the values given by the transmitting nodes. The ANOVA method steps are discussed as follows: Calculate the mean node values of each nodes node i ∈ NN g where node i having sensor value {sen 1 ,sen 2 ,sen 3 ,…..s m }. So the mean sensor value for the node node i is denoted as µ i and is calculated in equation (10) Let the sensor nodes set NN g = {node 1 , node 2 , node 3 ,…. node n } have means µ i ={µ 1, µ 2, µ 3 ,… .. µ i }, The overall mean µ for n number of nodes is calculated as Finally the Erroneous node ratio which is given as EN-ratio is calculated, which is defined by the ratio between mean squared value between nodes (msq b ) and mean squared value within nodes(msqw b ). These values are already computed in Equation 14 and 17: From this, the error critical value (Er critic = (deg fr , degw fr ) is calculated where the significance level α varies from 5 to 90%. The Er critic = (deg fr , degw fr ) is compared with the EN-ratio value for finding the significant level α. If the EN-ratio > Er critic = (deg fr , degw fr ) condition is satisfied then, the Null Hypothesis (HY 0 ) is rejected and concluded that there is a significant difference between the sensor values of the sensor nodes. If the EN-ratio < Er critic = (deg fr , degw fr ) condition is satisfied then, the alternative Hypothesis (HY 1 ) is rejected and it is concluded that there is no significant difference between the sensor values of the sensor nodes.

g) Algorithm: Apprehension Algorithm
(1) Initialized the Node Administration table [NADT] for each node association lead node (2) Initialized nod-stat tp i value for each sensor node with time out period tp out for node node i belonging to the group NN g [node i ∈ NN g ] (3) For each time period moment = {1,2,3,….tp} do (4) Supervisor nodes will send Hello messages to all the participating nodes in its domain for each time period of tp. (5) After receiving the Hello message, the participating nodes send Acknowledgment message (n-ACK) back to the supervisory nodes. (6) The nod-stat tp i of the NADT of the supervisory node is changed according to the values (possible values are 1,2 and 3 where 1 is node functioning normally, 2 is nodes that are likely to be unreliable and 3 is node that are completely erroneous.)

Probabilistic Neural Network
The nodes that are identified as erroneous are classified in the assortment phase. Probabilistic Neural Network (PNN) (Wu et al., 2007;Mishra et al., 2008;Specht, 1988;Xiao et al., 2017;Silva et al., 2018) is used to assort the erroneous nodes. PNN is used for analyzing the probability of different erroneous situations for unreliable node assortment process. PNN is generally a classifier, which takes the input that is feed, analyze its pattern and then turn them into corresponding output for assortment. Compared to the operational efficiency of neural networks, the PNN mode of operation and performance is much higher in network training like back propagation technique. PNN lacks the capacity of including minimal problem which makes it a better player in assorting the nodes after the pattern is analyzed.

Fig. 3: Probabilistic Neural Network Architecture
The architecture of the Probabilistic Neural Network (PNN) has four layers namely (a) Input layer, (b) Pattern layer, (c) Classes layer and (d) Output layer which is depicted in Fig. 3. The PNN has its operations which link to Bayes-Parzen (Nandagopal et al., 2017;Fernandes et al., 2016;Syahputra et al., 2017) window probability density function (pdf) estimator. PNN uses the Parzen's method of calculating the density function for variables that are occurring spontaneously. According to the Bayes' theorem PNN is classified an unknown sample using the following: where, CL is the class of the i th population and p denfunci defines the density function of the sample x of the i th population. Input layer contains the values that are got from the sensor nodes. These inputs are occurring spontaneously and they have to be assorted into any one type of population. The pattern layer will obtain the sensor values and bisect them into various classes which depends on the inaccuracy of the nodes. The classes layer is directly related to the pattern layer and the functionality of this layer depends upon the count of patterns of the pattern layer and make them to various classes with each classes having unique attribute of identification. This layer sums up all the patterns from each layer and fed that to the output layer. The output layer obtains all the patterns from the classes layer and this layer is responsible for the decision making process. Then the associated class label is determined by the following ways: (i) Sample values that are close to the unknown values should have a large contribution (increase the probability of classifying the unknown as that population) and (ii) Sample values that are present far away from the unknown should have a small contribution (decrease the probability of classifying the unknown as that population). So in every population the average value is calculated and then the unknown value is classified. The main advantage of employing PNN is that the unknown value measurements can be correlated and assorted using the probabilistic features which employs a polynomial mapping process.

Assortment Phase
After the Apprehension phase, each node association lead node performs the Probabilistic Neural Network computation. The collected values from the Apprehension phase is taken in to account and the erroneous nodes are identified and assorted. The Pattern layer collects the sensor values and these values can be normal functioning, likely to be unreliable or completely erroneous. This is assorted into three classes with each class representing one of the sensor values. The input layer collects the values and the pattern layer groups them based on the values obtained. Then each dendrites present in the patterns are watched and the average distance of the nodes are calculated. This distance subjected to Gaussian function gives the activation of the pattern layer dendrites. Each of the pattern layer values in the i th class corresponds to a Gaussian function related to the i th function. The activation of each class is taken by the dendrites associated with it. The classes layer adds all these values and finds the probability density function (pdf): where, x is the spontaneously occurring value and xi is the vector. N denotes the total number of assorting values and α is the balancing parameter. The main aspect of the assortment phase is that all the values must belong to only one class. All the three classes are computed for pdf value and are assigned a unique attribute namely gc 1 , gc 2 and gc 3 respectively for class 1, class 2 and class 3. Therefore all the values will be evenly distributed among all the available classes which make spontaneous sample representation mandatory in PNN computation. All the three classes along with their sensor values are taken into account and the total value of computation is calculated as an standard average. The iteration aspect is absent in PNN so it can handle large memory loads and since computational weights are not mandatory, the execution process will be slow but the output will be unbiased. x 2 , x 3 ,….x n } (7) for each class j = 1 to k do (8) for each sample value i = 1 to n do (9) Compute Gaussian function value for GF i (10) end for (11) end for (12) for each class j = 1 to k do (13)

Simulation Parameters
In this section, the performance of the proposed Wireless Nodal Behavior and Liability Investigation (WNBLI) protocol for wireless sensor networks is evaluated using Network simulator 2 (Issariyakul and Hossain, 2011). To get a clear working knowledge of the proposed protocol, it is simulated in the Network Simulator with the parameters suggested in the following Table 1. When a research work is conducted, it is always advisable to give a clear concise view of the operating scenario. It will help the experiment to take new forms when it comes to the evaluation part. The below listed parameters in Table 1 can be used to set up the operational environment which will express the proposed routing protocol better.
Initially 700 fault free sensor nodes are deployed in the network area. The heterogeneous faulty environment is created by adding different percentage of composite errors such as completely erroneous, soft unreliable nodes, intermittent errors and transient errors in the network. In this simulation, the soft unreliable nodes gives 81 to 100% erroneous results, the intermediate faulty node gives erroneous results for some random interval having 31 to 80% faulty values and the temporary error node gives erroneous results for a spike time interval having 5 to 30% faulty values. The composite error percentage are gradually increased from 5 to 40% in the network having 700 sensor nodes. The different faulty nodes percentage are shown in the Table 2.   5  50  5  41  3  1  10  75  8  62  3  2  15  100  9  81  6  4  20  125  9  100  7  9  25  150  20  119  2  9  30  175  16  144  3  12  35  200  18  156  8  18  40  225  20  176 9 20

Results and Discussion
The proposed routing protocol is compared with the existing protocols in two ways. Firstly the characteristics of the proposed routing protocol are compared with the already existing protocols. Secondly the proposed protocol is compared with other protocols of same nature for its operational efficiency in the wireless sensor network environment along with their energy consumption nature. The operational efficiency of the proposed protocol is tested for its faulty node identification and self rectification. This self rectification is made possible by dissecting the working nature of the proposed algorithm into three distinct feature sets of dependent nature algorithms namely association, apprehension and assortment.

Comparison Based on the Routing Characteristics
Characteristics like their adjunct characteristics, routing path, route position and packet generation methodology is taken into account. The following table (Table 3) gives a clear comparison between various other protocols belonging to the same feature set as the proposed routing protocol.

Comparison Based on Fault Detection Accuracy and Energy Consumption
The proposed routing protocol is compared with two other algorithms namely Threshold sensitive Energy Efficient sensor Network (TEEN) protocol which is proposed by Manjeshwar and Agrawal (2001) and Base-station Controlled Dynamic Clustering Protocol (BCDCP) which is proposed by Muruganathan et al. (2005). The TEEN routing protocol is targeted towards reactive networks as the nodes operating with this routing protocol sense the network environment continuously. It can be applied to real time applications like intrusion detection and network explosion detection.    (Yao et al., 2015) Yes Yes Yes Yes ALBA-R (Petrioli et al., 2012) Yes Yes Partial Yes R3E (Niu et al., 2014) Yes Yes Yes Yes WNBLI Yes Yes Yes Yes The BCDCP routing protocol distributes the energy abundance evenly among all sensor nodes to improve network lifetime and improve the average energy savings by the operating nodes. The proposed routing protocol is compared with TEEN and BCDCP routing protocols for Faulty node detection. The graph which compares the Fault detection accuracy among the routing protocols is depicted clearly in the Fig. 4. The energy consumption criteria of the routing protocols are also taken into account for analysis. This energy consumption is compared with the number of faulty nodes existing in the wireless sensor networks. The graph which illustrates the energy comparison difference among the routing protocols is depicted in the Fig. 5.

Conclusion
In this research work, a static and analogous routing protocol with error detecting capability for wireless sensor network is proposed. The proposed Wireless Nodal Behavior and Liability Investigation (WNBLI) protocol is accomplished mainly in three major phases, such as Association phase, Apprehension phase and Assortment phase. A time division based multi channel media access control protocol is designed for data communication. The architecture of the proposed protocol is designed for a clustering method which is load balanced by nature. This working structure makes the proposed protocol consume less energy while operating in a network. The main thing to be noted in this research work is the point that the energy consumption factors play a very vital role in determining the outcome of a research. The node behavior is explained on the context of topology in which they belong so that the node operations can be analyzed for their utility. Once the node behavior is analyzed and an output is obtained, it is very easy to understand its operational functionality in the present topology. The proposed protocol is simulated in the network simulator and it is compared with other protocols operating at the same environment. The result shows that the proposed protocol is energy efficient and performance oriented for wireless sensor networks. By employing the proposed routing protocol, many prospects of the wireless sensor networks can be benefited.