Hybrid optimization algorithm for security aware cluster head selection process to aid hierarchical routing in wireless sensor network

In wireless sensor networks, clustering is said to be the most noteworthy technique for increasing the lifetime of network that directly leads a better routing mechanism. This approach involves grouping of sensor nodes to clusters and choosing the appropriate cluster heads for each cluster. In fact, cluster heads gathers data from corresponding nodes in cluster and transmits those aggregated data to base station. However, the major issue in this is the selection of the appropriate cluster head. Till now, many research works have been carried out for solving this issue by considering different constraints. This paper introduces a new cluster-based routing model by selecting the optimal cluster head. Moreover, a novel algorithm known as grey wolf updated whale optimization algorithm is introduced. Here, a new multi-objective function is deﬁned with respect to different constraints like distance, delay, security and energy, respectively. Finally, the performance of security aware clustering with grey wolf updated whale optimization algorithm is evaluated and validated over other conventional works with respect to alive node analysis, throughput and normalized network energy, respectively.

In cluster-based routing, CH [19,20] takes the complete responsibility for data transmission; thereby network prolonging is highly possible via balancing the energy consumption which, is to be noted for the precise or optimal selection of cluster head [21,22]. The number of works has been progressed with cluster-based routing strategy through the utilization of recent metaheuristic and machine learning algorithms [23][24][25][26][27]. Among them, optimization models like particle swarm optimization (PSO) [28] plays a vital role in clustering and CHS [29,30], respectively. Still, the need for advancement in algorithms lags the routing performance with respect to energy balancing and throughput.
The main contributions of the paper are as follows: 1. This paper introduces a new hierarchical routing protocol for WSN via introducing a new variant of the cluster head selection (CHS) model. 2. The proposed CHS model takes the security constraints in addition to the existing constraints like energy, distance, and delay. 3. To assist the optimal CHS process, a new hybridization of grey wolf optimization (GWO) and whale optimization algorithm (WOA), known as grey wolf updated whale optimization algorithm (GU-WOA) is proposed in this paper.
The paper is arranged as shown: Section 2 shows the reviews on CHS. Section 3 offers an enhanced cluster-based routing model for WSN and Section 4 depicts the hybrid optimization algorithm for appropriate cluster head selection. Section 5 elaborates the defined multi-objective model for optimal cluster head selection. Section 6 portrays the results and Section 7 concludes the paper.

Related works
In 2019, Toor and Jain [31] have presented a narrative framework for hierarchical heterogeneous WSNs on considering mobile sensor nodes named MEACBM routing protocol. Depending on this novel probability equation, the CHs were selected optimally. Here, only the sensor nodes were chosen as CH that offered larger energy among the other sensor nodes. This approach had considerably minimized the energy utilization of sensor nodes for transferring the information to BS. The experimental outcome had shown the betterment over the other models regarding the throughput, count of CHs, network lifetime and count of dead nodes. In 2018, Kaur and Mahajan [32] have considered the energy efficiency of WSN as the major issue in this work. These sensor nodes turned out to be lifeless over a course of time as they were charged using the battery. Network lifespan of WSN was improved by means of the tree and clustering-based data aggregation schemes for sensor networks. This work had implemented a tree oriented routing protocol along with hybrid ant colony optimization (ACO) and PSO based techniques. At first, depending on the residual energy, clusters were formed, and subsequently, the hybrid ACOPSO framework was developed for further enhancement of inter-cluster data aggregation. The experimental analysis stated the substantial enhancement in network lifetime.
In 2015, Rani et al. [33] have analysed the minimization of energy utilization and total transmission time of WSN by means of multi-hop data aggregation, which formed synchronization within hierarchical clustering. A new transmission algorithm was proposed for routing depending on the predefined path. Based on the relay nodes, the minimization of transmission distance was made. The simulation output revealed the better performance of the implemented model regarding time and energy.
In 2018, Sodairi and Ouni [34] have analysed the efficacy of LEACH and LEACH-based protocols on energy-constrained WSNs for prolonging the lifetime. For reducing the energy consumption and for enhanced packet delivery and network lifetime in WSNs, a new enhanced LEACH clustering protocol was introduced in this work. Further, the sturdiness of the LEACH protocol was also presented in this paper. A new set of rules was developed for CH selection on the basis of residual energy. Also, the multi-hop communication technique was integrated into the WSN by means of operating processes namely generic and levelling multi-hop routing.
In 2018, Huang et al. [35] have implemented the annulus sector grid aided routing protocol (ASGRP) model, where the improved energy efficiency and extended network lifetime in WSN were attained using an annulus sector grid clustering scheme. Depending on arithmetic progression, the annulus sector grid clustering model was implemented that split the nodes into several clusters. Further, network area was split into diverse levels based on the equivalent diverse distance. Finally, the experimental analysis thus proved that the proposed model had extended network lifetime and reduced the energy consumption significantly.
In 2016, Aslam et al. [36] have implemented two protocols concerning two-hop heterogeneity-aware centralized energy efficient clustering (THCEEC) and advanced heterogeneityaware CEEC (ACEEC) frameworks. Further, the intellectual distribution of energy resources was required by the implemented model to evenly spread over the heterogeneous WSNs on the prime phase of cluster formation. The simulation outcome demonstrated that the ACEEC and THECCE have superior performance over the other models and it provided a better lifespan of network and better data delivery with optimal stability time.
In 2018, Tianshu et al. [37] have presented a routing model based on genetic algorithm-based energy-efficient clustering and routing (GECR) and GA-based energy-efficient clustering for prolonging the life cycle of network and enhancing energy efficiency. The optimal solution attained from the previous network round was added with the initial population for the present round and thus the search efficiency was enhanced. Further, while modelling the fitness function, the load balancing factor was also considered that balanced the energy consumption between the nodes. The simulation outcomes have revealed the betterment of the proposed model in terms of load balancing with low variance and improved energy efficiency.
In 2016, Ahmed et al. [38] have examined a novel routing protocol using the sleep-awake energy-efficient distributed (SEED) clustering approach. When comparing the high energy region CHs with low energy region CHs, the communication with BS in high energy region had required a more distance and also pays the additional energy cost. The sub-clusters were formed for reducing the count of transmissions headed for BS. From these sub-clustered nodes, one node in every round was awakened and the data and the remaining nodes were transferred for saving the existing resources. The experimental investigation had shown that SEED model gained a larger throughput and extended lifespan of network.
In 2018, Shankar et al. [39] offered the GGWSO algorithm for security-aware CHS in WSN. Also, the offered method was compared over other conventional algorithms such as fractional group search optimization, artificial bee colony (ABC), as well as GWO-based CH selection. At last, the experimental results show that there was a requirement for a hybrid model to accomplish better results.
In 2016, Bhatia et al. [40] have presented a CHS algorithm in WSN by hybridizing the GADA-LEACH to diminish energy dissipation and enhance network lifetime. Also they utilize the GA for selecting the CH optimally and relay node for distance aware (DA) routing. Finally, the result shows that the presented model was proficient in terms of network lifetime.
In 2019, Tabatabaei et al. [41] have presented an energy-aware clustering approach by using the LPO algorithm and fuzzy logic. Here, clustering was done based on two-parameter nodes. Finally, the results demonstrate that the proposed scheme was better in terms of input packet, power consumption enhanced network lifetime as well as average delay.
In 2017, Miglani et al. [42] presented a trust-aware and energy-efficient system to secure routing in LEACH which enhances LEACH protocol. The presented approach did the arrangement of trust management and trust-based routing module, which works collectively to choose trusted CH. Finally, the result shows that the presented method was superior in terms of PDR and network lifetime. Table 1 shows the reviews on CHS in WSN. MEACBM [31] offered a superior performance by minimizing the consumption of energy and rises network lifetime, throughput and count of dead sensor nodes. However, it needs simulation over the realtime experiments and requires the consideration of the scalability of sensor nodes within every cluster. ACOPSO [32] poses reduced packet size, enhances network lifetime and conserves energy in an effective manner. But, critical to define the initial design parameters and cannot define the issues of scattering. CBCCP [33] offered an elongated network lifetime and speeds up data communication. Further, checks on the possible condition of feasibility in the control room, which was the major challenge of this method. EM-LEACH [34] provided an improved performance regarding the packet delivery rate and extended network lifetime. The main demerits are fault tolerance issues and the nodes are homogenous in nature with the same energy. ASGRP [35] prolongs the lifetime of network and offers better scalability. Network performance needs improvement and learning on clustering of heuristic algorithms was necessary. THCEEC [36] had reduced the packet drop and avoids retransmission. It still lacks in real-time applications for learning and optimizing. GECR [37] poses a better network life cycle, and improved energy efficiency. Yet, future work needs on verifying and applying the appropriate metaheuristic algorithms. SEED [38] had improved network stability, throughput and packet drop rate and further significantly reduced energy consumption. Two major disadvantages are energy harvesting scheme needs to be considered and network lifetime further needs improvement.

Review
Most of the CHS model considers the renowned physical layer parameters such as distance, energy and delay, this paper is one of the few notable works that consider security constraints in the CHS model. Moreover, the implementation of the CHS model in networking environment, while the existing works have been implemented only for optimizing the CHS model. To be more specific, this paper has come out with a novel variant of an existing optimization algorithm, termed here as GU-WOA. The proposed algorithm hybridizes the GWO and WOA in a unique manner to enhance the CHS process.

Network model
WSN comprises of numerous stationary sensor nodes, symbolized by M M , in which all the sensors has its equivalent competences. During the data transmission, a node could act as a CH and also as an active sensor. Normally, the WSN is associated with the allocation of sensors, topology features, data sensing, radio communication, and energy utilization. The sensors are positioned by random or manual mode in application areas. The mixture of different sensor nodes carries out clustering process. It is a superior technique for expanding the lifespan of WSN. During clustering process, clusters are formed by gathering the sensor nodes in which a CH is chosen and its count is indicated by M c . This CHS is made for all clusters. The nodes in a specific cluster are created such that, it should have a minimal distance from CH. The entire sensor nodes collect information from the target area and pass it to the CH during the entire process. Moreover, the particular CH passes the gathered information to the BS. In the presented work, the CHS of WSN with numerous sensor nodes and centralized BS are shown in Figure 1.
The WSN concerns with transmitting the data from one node to another. Therefore, by recognizing the shortest path, the process of data transmission could be improved. Also, the energy utilized by the nodes is regarded as another main issue. Here, the major problem is to share the data with the shortest path and reduced energy. Numerous analysts have introduced techniques for distributing data packets among the BS and nodes by employing various improved routing protocols. In many routing protocols (specifically hierarchical routing), the desired CHS in terms of energy and position is regarded as the main aspect. Usually, a node needs high energy for transferring a huge amount of data. The optimal positioning of CH could decrease the energy, thus enabling a certain CH to transmit more data. Hence, a node chosen as the CH helps in better positioning of sensor nodes with reduced energy utilization. Along with this, the maintenance of secured data transmission is a vital consideration as there exist different risk factors. In this work, distance, delay, energy consumption and security are considered as the major aspects of selecting the appropriate CH.

Distance model
Initially, the whole CH's selected in the network transmit the message to a state, in which they function as the CH. Each sensor node evaluates its distance from CH. A node is placed in a particular cluster only if its distance from CH is minimal. Accordingly, the sensor node transmits the messages directly to the BS, if the distance between CH and node is higher than the distance between the node and BS. Depending on the nearer distances, it creates a set-up for cluster formation. Thus, the nodes in network are clustered again with the chosen CH using distance matrix DM (q × p) as mentioned by Equation (1), where e M c indicates the Euclidean distance among CH and normal node with the position information M c as well as z 1 , z 2 , … .z n indicates the sensor nodes. Consider two sensor nodes ath and bth and let their positions be y and w, respectively. The Euclidean distance is computed as per Equation (2). The distance matrix of the entire elements indicates the distance that relies on the CH of ath node and bth node as revealed in Equation (1). Consider a component e M c2 ,z 1 that fills the first column of the matrix with minimal distance. Here, the CH M c2 and node z 1 are correlated to one another.
Furthermore, a time slot is allotted by M c to each sensor node during the transmission of data. The most important function of all M c was to store the transmitted data from all sensor nodes in clusters. Following the accomplishment of data from all sensor nodes in a specific cluster, M c passes the related data to the BS or sink. When M c remains at active mode, the sensor node remains at sleep mode. The data transmission and reclustering are attained from several cycles until the sensor nodes turn inactive. The free space and multi-path fading channels are exploited depending on the distance between the transmitter and receiver.

Radio model
Even if nodes are placed at a known position, its radio distance may vary with several parameters. Euclidian distance is not therefore a reliable parameter to be considered when addressing routing protocols or any other wireless network-related problem [43]. Hence, by considering the energy parameters the radio model assumed in the paper is given in Equation (3). Where, E pw specifies the power amplifier energy and E fs indicates the required energy when deploying the free space model [9,13]. The energy model is discussed below.

Energy model
Energy utilization is considered as the most significant concern in WSN. In fact, when there is no power supply, the battery could not be re-energized. More energy is necessary for passing the data to BS from the entire sensor nodes. Moreover, more energy is required as network carries out various functions such as sensing, transmission, aggregation and reception. Therefore, the entire requirement of energy during message transmission is given by Equation (3), where E el specifies the electronic energy that includes spreading, filtering, digital coding, etc. and E TX (M ∶ e) denote the total energy utilized for transferring N bytes of packets at a distance e. Equation (4) demonstrates the model of electronic energy, in which E ae indicates the energy utilized during data aggregation. The total energy required for attaining M bytes of packets at distance d is given by Equation (5). Equation (6) illustrates the energy necessary for amplification. E pw specifies the power amplifier energy and E fs indicates the required energy when deploying the free space model.
The total network energy is specified by Equation (8), in which E 1 signifies the energy required in the idle state and E S indicates the cost of energy at the time of sensing. Thus, it is crucial to lessen the entire energy as indicated in Equation (8).

Security awareness
In the security mode, risky mode and risky modes are considered as major aspects that ensure the security in WSN routing. Security mode: This mode chooses the CH [44] that satisfies the security requirements. In Equation (9), o r and o d refers to the security rank and security needs associated with the CHS. If o d ≤ o r , the node is said to be the better CH.
Risky mode: This mode takes all the risks for enabling the optimal CHS by electing an existing CH. Thus, the risky mode is considered as an insistent mode during the CHS processing.
-risky mode: The CH which can hold at most -risk is preferred in the assortment process depending on -risky mode. Consequently, refers to the probability measure with values, = 0 and = 1 (i.e., 100%) similar to the risky and security mode.
Among the three modes, the secure mode is said to be the challenging and economical one. However, the much-exploited one is the -risky mode. Here, five security levels are offered by the qualitative fuzzy scale such as, "very low (assigned as 1), low (assigned as 2), medium (assigned as 3), high (assigned as 4) and very high (assigned as 5)". If the proper CH is elected during the selection process, [44] then it is said to be the secure process. The security constraint model probability is specified by Equation (9).
Accordingly, if the chosen CH achieves the state o d > o r the risks have to be lesser than 50%. If the condition is 0 < o d − o r ≤ 1, the selection process would be executed, and if the state is 1 < o d − o r ≤ 2, delay arises in the selection process. Yet, the CHS process could not be completed, and it should be continued for the state 2 < o d − o r ≤ 5 [44].

Objective model
The major objective of the presented research work on CHS is to reduce the distance among the selected CH and node. It also aims at reducing the delay to transfer the data from one node to another. On the contrary, network energy has to be high, that is, it must exploit only a minimal amount of energy during data transmission. Finally, the node should tolerate the risks obtained in network. The objective function of the adopted CH model is shown by Equation (10), in which the value of should rely on 0 < < 1. Here, v m and v n point out the operations as shown by Equations (11) and (12), correspondingly. The constraints on delay, energy, distance and security are specified by 1 , 2 , 3 and 4 . The condition of these constraints is indicated: 1 + 2 + 3 + 4 = 1. In Equation (12), Y z − S s refers to the distance amongst the normal and sink node.
Equation (13) shows the fitness function for distance, in which v dis (m) is associated with packet transmission from the normal node to CH and then from CH to BS. Usually v i dis lies among [0, 1] and its value goes higher if the distance between CH and the normal node is high. Equations (14) and (15) demonstrates v dis (m) and v dis (n) correspondingly, in which Y z refers to the normal node in zth cluster, G z portrays the CH of zth cluster, the distance among the BS and CH is specified by G z − S s , G z − Y z denotes the distance among the normal node and CH and Y z − Y y signifies the distance amongst two normal nodes, M z and M y symbolizes the node count, which excludes the zth and yth cluster.
The fitness function of energy is mentioned by Equation (16). The value v i ene will be higher than one and the entire CH cumulative v ene (m) and v ene (n) concerns the highest energy value and the higher count of CH.
Equation (17) depicts the fitness function of delay v del i that lies within [0,1]. v del i is directly proportional to all the nodes residing in cluster. Therefore, a delay gets minimized if the CH has a reduced number of nodes. The denominator M M refers to the whole count of nodes in WSN, and the numerator refers to the higher count of CH.

HYBRID OPTIMIZATION ALGORITHM FOR APPROPRIATE CLUSTER HEAD SELECTION
Even though the conventional WOA model offers satisfactory outcomes, it still suffers from premature convergence and it also gets trapped by local optima that affect the optimization process. Hence, the concept of GWO is integrated with the existing WOA model that ensures better convergence and optimal outcomes. Hybrid optimization algorithms have been reported to be promising for certain search problems [45]. The most significant motivating property concerning the humpback whales [30] is the amazing process of hunting. They discover the prey's position and surround them. This approach presumes that the current optimal solution is the one that is nearer to the targeted prey. This behaviour was denoted in Equations (18) and (19), where it specifies the present iteration, ⃗ A and ⃗ E indicates the coefficient vectors and J * symbolizes the best solution attained till now, || points out the absolute value, and '⋅' is an "elementby-element multiplication".
It is significant to observe that J * has to be updated in the existence of a better solution. The vectors B and E are computed as specified in Equations (20) and (21), where ⃗ o is minimized from 2 to 0 for varied iterations and ⃗ f indicates the random vector in [0, 1].
Exploitation phase: "Shrinking encircling model": This phase could be achieved by minimizing ⃗ o value in Equation (20). Note that the variation of ⃗ A is lessened by ⃗ o, that is ⃗ A lies between [−o, o] in which o is reduced from 2 to 0 for more iterations.
"Spiral updating position": Primarily, this system computes the distance positioned at (J , V ) and prey at (J * , V * ). A spiral formula is further produced among the location of prey and whale as in Equation (22), in which ⃗ D = | ⃗ J * (it ) − ⃗ J (it )| and it denotes the distance found between the ith whale and prey, the constant for defining the form of the spiral is denoted by b, and l is an alternate value lying in the interval between [−1,1]. Equation (22) can be rewritten as in Equation (23), where p is a random integer lying among [0,1].  energy, and throughput, and the betterment of the adopted scheme was evaluated over other traditional methods namely, glowworm swarm optimization (GSO) [46], ACO [47], PSO [48], and ACI-GSO [49]. Here, the analysis was done under two experiments: Experiment 1 with 50 nodes and experiment 2 with 100 nodes by varying the count of rounds to 20, 30, 40, 50,  60, 70 and 80, respectively. During implementation, the initial energy E i was set as 25 J, and receiving power energy E RX was set as 0.0648 nJ∕bit∕m 2 . Consequently, the transmission power energyE TX was allotted as 0.744 nJ∕bit∕m 2 and the idle power and sense power were fixed as 0.05 and 0.0175 nJ∕bit∕m 2 correspondingly. The simulation results of the suggested model in terms of 50 and 100 nodes are given in Figure 3. The simulation parameters are depicted in Table 2. The algorithm parameters is depicted in Table 3.   Figure 4 shows the analysis of the adopted GU-WOA scheme with respect to the number of alive nodes. From the attained outcomes, it could be observed that the number of alive nodes has reduced with an increase in the number of rounds. However, when compared to other conventional schemes, the proposed model makes the nodes alive even in round 80. This proves the efficiency of the proposed model over other models. However, the nodes may die quickly in the middle, if the energy consumption is not balanced properly. More particularly, from Figure 4(a), the adopted GU-WOA scheme for 50 nodes is 15.79%, 46.67%, 29.41%, 22.22%, 6.82%, 9.09% and 15.90% better than ACI-GSO, GSO, ACO, PSO algorithms and algorithms proposed by Shankar, et al. [39], Vinitha and Rukmini [50], and Sharma, et al. [12] with more alive nodes at 80th round. Also, from Figure 4(b), the presented system for 100 nodes is 5.77%, 10%, 14.58%, 7.84%,10.90%, 9.09% and 5.45% better than ACI-GSO, GSO, ACO, PSO algorithms and algorithms by Shankar, et al. [39], Vinitha and Rukmini [50], and Sharma, et al. [12] at 80th round with an increased count of alive nodes. Thus, the superiority of the presented GU-WOA model has been proved over other existing models with a higher count of alive nodes.  Network energy remains higher with the proposed clustering process, whereas the remaining models could not balance the energy consumption while data transmission. This is drastically proved through the curves shown in the graphical representation. Specifically from Figure 5( [50], and Sharma, et al. [12] algorithms. Therefore, the enhancements of the adopted GU-WOA for optimal CHS in WSN have been validated effectively.

Analysis of throughput
The throughput analysis of the proposed GU-WOA model over other existing models is specified by Figure 6. From the simulated outcomes, the adopted model has proved the increased throughput of proposed work over the existing schemes almost in all rounds. The increase in throughput directly means the proposed model has less overhead, less delay, uniform distribution of energy among nodes as well. However, the conventional models failed to attain these results with poor throughput performance. From Figure 6(a), the adopted model for 50 nodes has acquired more throughput that is 1.75%, 14.71%, 29.68%, 19.39%, 4.09%, 15.68% and 12.36% better than ACI-GSO, GSO, ACO, PSO algorithms and algorithms by Shankar, et al. [39], Vinitha and Rukmini [50], and Sharma, et al. [12] at the 80th round. Also, from Figure 6(b), the throughput of the implemented GU-WOA for 100 nodes is 1.86%, 15.66%, 6.93%,11.68% 8.31%,13.15% and 3.74% and better than ACI-GSO, GSO, ACO, PSO algorithms, and algorithms by Shankar, et al. [39], Vinitha and Rukmini [50], and Sharma, et al. [12] in the 80th round. Thus, the attained outcomes prove the superiority of the adopted GU-WOA over other schemes.

6.6
Analysis of security Figure 8 illustrates the security analysis of the proposed GU-WOA model with respect to the number of rounds. On observing, Figure 8 it can be seen that when the number of rounds increased the security level also get increases. Figure 9 illustrates the communication cost of the adopted GU-WOA scheme with respect to the number of nodes. From  Figure9(a), it can be observed that the adopted model for 50 nodes has obtained minimum communication cost that   [50], and Sharma, et al. [12] at the 80th round. Also, on observing it is clear that the Figure 9( [12] in the 80th round for the implemented GU-WOA for 100 nodes. Therefore, the attained outcomes prove the supremacy of the proposed GU-WOA over existing schemes.

Computational time
The computational time of the adopted GU-WOA scheme and the conventional models, with respect to the number of nodes, has been shown in Figure 10(a) 50 nodes and Figure 10(b) 100 nodes. From Figure 10, it can be observed that the adopted model takes less time to compute for 50 nodes that are 56.14%, 5.81%, 12.31%, 61.95%, 57.15%, 26.86% and 5.55% better than ACI-GSO, GSO, ACO, PSO algorithms and algorithms proposed by Shankar, et al. [39], Vinitha and Rukmini [50], and Sharma, et al. [12] at the 80th round.

6.9
Overall performance of proposed model Tables 4 and 5 show the overall analysis of the presented GU-WOA model by varying population sizes to 10, 20, 30, and 40 with 50 nodes and 100 nodes. Here, the number of alive nodes, normalized network energy, and throughput are evaluated for diverse rounds namely 20, 40, 60 and 80. Table 4 shows the presented system for 50 nodes, which shows better performance in the 20th round for energy, alive nodes for all the population sizes. The throughput is found to be better at the 80th round and it is exhibited in kbps. Similarly, the adopted approach for 100 nodes is revealed in Table 5, where the better performance of energy and alive nodes are attained at the 20th round, whereas the throughput is found to be better at the 80th round. Moreover, on considering Tables 4 and 5 the presented model with 100 nodes attains superior performance than 50 nodes in the 20th round for energy, alive nodes for all the population sizes. In addition, the throughput of 100 nodes is better at around 80 than 50 nodes. These outcomes have shown the enhancement of the adopted scheme in terms of alive nodes, network energy, throughput and PDR. Figure 11 depicts the lifetime of the nodes with respect to the number of nodes. Figure 12 shows the convergence analysis of the presented GU-WOA model over the existing schemes by varying the number

FIGURE 11
Node lifetime with respect to the number of nodes of rounds. Here, the cost values of the proposed and conventional schemes are gradually minimized with an increase in the number of rounds, however, the presented GU-WOA model has revealed a minimal cost value when evaluated over the conventional schemes. Specifically, in the first round, the adopted scheme is 7.62%, 6.85%, 5.74% and 4.97% superior to ACI-GSO, GSO, ACO and PSO algorithms. Similarly, in the second round, the implemented approach is 7.01%, 6.28%, 4.83% and 0.1% superior to ACI-GSO, GSO, ACO and PSO algorithms. Thus, the superior performance of the adopted scheme has been verified from the analysis.

CONCLUSION
This paper has presented a new cluster-based routing model by choosing the optimal CH. Moreover, a hybrid version of GU-WOA was introduced in this paper. In this work, a novel

FIGURE 12
Convergence analysis of the proposed model over the traditional models multi-objective function is portrayed with respect to different parameters such as distance, delay, security and energy. In the end, the performance of the implemented work was validated over the existing schemes in terms of alive nodes, throughput, and network energy. From the analysis, the adopted GU-WOA scheme for 50 nodes was 8%, 24%, 32% and 48% better than ACI-GSO, GSO, ACO and PSO algorithms with more alive nodes at 80th round. Moreover, the presented system for 100 nodes was 4.48%, 43.28%, 65.67% and 79.1% better than ACI-GSO, GSO, ACO and PSO algorithms at 80th round with an increased count of alive nodes. Thus the betterment of the presented GU-WOA based CHS in WSN was validated effectively. Future work can be expanded with a new technique for the energy issue and provides an energy proficient multi-hop routing with the intention to obtain high performance. Besides, to handle the incomplete knowledge more effectively the use of an intrusion detection facility can be added as an additional feature by using fuzzy constraints.