ANFIS based Optimal Routing using Group Teaching and Adaptive Equilibrium Optimization based Trust Aware Routing Protocol in MANET

Recently, routing is considered the main problem in MANET due to its dynamic nature. The route discovery and the optimal route selection from the multiple routes are established for the efficient routing in MANET. The major objective of this research is to select the optimal route for packet transmission in MANET. In this paper, four stages namely trust evaluation, route discovery, optmal route selection and route maintanance are elucidated. Initially, the trust evaluation is made by using ANFIS where the primary trust values are evaluated. The next stage is the route discovery scheme, in which the routes are established by Group teaching optimization algorithm (GTA). From the route discovery scheme, multiple routes are found. The optimal route for the transmission is selected with the help of the Adaptive equilibrium optimizer (AO) algorithm. Finally, the route maintenance process is established; if any of the routes fails for the broadcast it immediately selects the alternate optimal route from the multi-zone routing table for efficient packet transmission. The proposed approach is evaluated by various performance measures like throughput, energy consumption, packet delivery ratio, end-to-end delay, packet loss rate, detection rate, and routing overhead. This result describes that the proposed approach outperforms other state-of-art approaches.

 Introducing the ANFIS concept for the evaluation of the trust value of the neighbors for the trust-aware routing protocol.  Establishing the route discovery scheme by the GTA algorithm  Selecting the optimal route by adaptive equilibrium optimizer for efficient packet transmission.
The remaining of the paper is arranged as follows. The summary of the exiting works on the trust aware energy-efficient routing protocol in MANETs is discussed in section 2. The proposed methodology for the trust aware energy-efficient routing protocol is described in section 4. The results and discussions are described in section 5. Section 6 states the conclusion and future scope of the work.

LITERATURE REVIEW
This section describes the review of some of the existing protocols with the disadvantages of those methods and described in Table 1. Nandgave-Usturge S et al. (2020) proposed Water Spider Monkey Optimization (WSMO) for discovering the optimal path depends on the trust and other parameters such as delay, overhead and distance. The performance of the proposed algorithm was evaluated with the measures such as delay, packet delivery ratio, denial of service, throughput with and without considering the denial of service, black hole attack, etc [11]. Mukhedkar MM and Kolekar U (2019) introduced Advanced Encryption Standardenabled Trust-based Secure Routing protocol depends upon the projected Dolphin Cat Optimizer (AES-TDCO) with no accurate centralized monitoring process. The measures such as packet drop, delay, detection rate, and throughput were used for the evaluation. This method was difficult to maintain better trust and security-aware routing protocols when dealing with imaginary thoughts [12]. Sun Z et al. (2019) proposed the Secure Routing Protocol based on Multi-objective Ant-colony-optimization (SRPMA) for obtaining the increased network security through low energy consumption. The simulation results were obtained by using measures such as energy consumption, packet loss rate, routing load. This approach has drawbacks like decreased network lifetime as well as reliability [13]. Mukhedkar MM and Kolekar U (2020) designed the routing protocol called Encrypted trust-based dolphin glow-worm optimization (DGO) (E-TDGO) for providing secure routing in MANET. The experimental results were demonstrated with the measures such as packet drop, delay, detection rate, and throughput with considering the attacks [14].
Kanagasundaram H and Ayyaswamy K (2019) proposed a Multi-objective Ant lion optimizer (MALO) for dealing with the routing protocol with energy efficiency and security. The performance of the method was evaluated using measures such as delay, energy consumption, throughput, and network lifetime [15]. To decrease exposure from the malicious nodes and improves security, Alkhamisi AO et al. (2020) proposed an Incorporated Incentive and Trust-based optimal path identification in Ad hoc On-Demand Multipath Distance Vector (IIT-AOMDV). The simulation results were described using measures such as route selection time, throughput, detection accuracy, and overhead [16]. Mariadas AE and Madhanmohan R (2020) proposed Hybrid PSO and differential algorithm that concentrates on the lifetime energy based on clustering by the fitness value. The effectiveness of the technique was calculated in terms of the measures namely packet delivery ratio, average delay, packet loss ratio, throughput, and energy consumption [17]. Keum D et al. (2020) proposed a trust-based multipath QoS routing protocol (MC_TQR) for the ad hoc network for detecting the malicious nodes and also to assure the reliability of the network. The method was evaluated depends on measures namely end-to-end delay, throughput, and packet delivery ratio. The major difficulty was this method has low reliability [18].
Kumar R and Shekhar S (2020) introduced a trust-based fuzzy bat optimization approach for diminishing the consequences of attacks. The results were demonstrated using measures such as network lifetime, packet delivery ratio, throughput, end-to-end delay and routing overhead [19]. Merlin RT and Ravi R (2019) Trust-based energy-aware routing (TEAR) approach for increasing the resourced restricted issues in the network and provides enhanced route security. The measures such as trust value, energy consumption, and Successful routing ratio utilized for the evaluation purposes. This method increases the lifetime of the network by neglecting the black hole attacks and maximizes the successful routing probability [20].

SYSTEM DESIGN
This section imagines the important multi-hop MANET in which the nodes are disseminated arbitrarily in the 2D space [21]. Let us consider that there are three cases, where each case comprises of a single channel that are identical to the broadcast range. This paper utilizes directed graph model that are expressed as From Eqn.(1), e V represents the set of the entire nodes and l E represents the set of the entire edges. Every edge indicates that the two nodes are located inside the broadcast range. In this research the hybrid routing protocol is considered in which the graph r G is classified into zones. Every node contains only one zone and every zone contains the middle node and few other nodes. The node that corresponds to the zone is classified into interior as well as the boundary nodes individually. The structure of the hybrid routing protocol consists of trust evaluation and route schemes. Trust evaluation runs for every node to verify the neighbor's behavior in real-time and achieves trust values that are considered as the most significant factors for the hybrid routing protocol. In the majority of the cases, trust contains two sections: direct trust and indirect trust. Due to significant computation cost as well as traffic load, this research considers direct trust. The routing protocol consists of routing within the zone, among the zone, and other fundamental schemes.

Problem definition
Let us consider the MANET with n number of nodes reponsible for sensing and gathering the information related to the environment. Each node lies within the transmission range and the transmission among the source to destination nodes taken place inside the range. If the nodes are moving, they alter their position inside the transmission range, so the new node location is obtained inside the transmission range. Hence to create an effective route for the broadcast, two various factors namely trust and distance are considered. The mobility schemes of the MANETs determine the nodal movement and it depicts the position, acceleration and velocity of the nodes locations where its positions are updated periodically inside the broadcast range. In addition to this, trust is the main parameter which determines the MANETs security and the nodes determines the trust degree of the network. To overcome these problems, trust-aware routing protocol is proposed that further enhances the trust and security.

PROPOSED METHODOLOGY
In this paper, four stages namely trust evaluation, route discovery, optmal route selection and route maintanance are elucidated.Initially, the trust evaluation is made by using ANFIS. In this trust evaluation, the primary trust values are evaluated. The next stage is the route discovery procedure in which the route is discovered by Group teaching optimization algorithm. The third stage is the optimal route selection. Due to the establishment of multiple routes, the optimal route selection is considered as the main problem in the route discovery process, so adaptive equilibrium optimizer is utilized for the optimal route selection procedure. In the route maintenance stage, if any of the routes fails for the broadcast it immediately selects the alternate optimal route from the zone routing table for efficient packet transmission. The route is maintained until the packet reached the destination. The proposed framework for the stable route prediction is described in

Trust evaluation
In the trust aware routing protocol, the analysis of the node's behavior is represented to evaluate whether the node is selfish or the node is acting like the malicious nodes. Each node supervises the behavior of the neighbor nodes in real-time and the trust weights are computed. If the neighbor node j links with node i, then the node j organizes it as the primary trust value. The primary trust value is determined by the ANFIS model.

ANFIS model
The adaptive neuro-fuzzy inference system (ANFIS) is the neural network model that provides the solution of approximation issues. This function provides the mapping link among the input as well as the output data with the deep learning method for determining the best membership function distributions. In this ANFIS structure, five layers are utilized to build the inference system. Every layer composed of various nodes depicted through the node function [22]. The input of the current layers is achieved from the nodes in the preceding layers. The fuzzy inference system is assumed which consists of 5 layers of the adaptive network with two inputs (a, b), and it has only one output c. The ANFIS rule base function is composed of fuzzy if-then rules of Sugeno type [23]. For the first order two-rule Sugeno inference system, the two rules are described as: Where a and b represent the ANFIS inputs, X and Y represent the fuzzy sets F1 (a, b) indicates the first-order polynomial, indicates the output of the first-order Sugeno fuzzy inference system. The structural design of ANFIS is exposed in Fig 2 and the node functions are determined below. The adaptive nodes are represented using squares that indicate the factor sets which are changeable with the help of the nodes [31]. The fixed nodes are represented by the circles, in which the factor sets are permanent in the system.
From Eqns. (2) and (3), the input nodes are represented as input nodes, X and Y represents the linguistic labels, ) (a  and ) (b  represents the membership functions that are typically accepted the bell-shaped function with the highest and lowest value identical to 1 and 0 correspondingly as pursues: From Eqns. (4) and (5), j x , j y and j z represents the factor set. The above-said factors are represented as premise factors.

Layer 2:
The nodes in layer 2 is represented as the permanent node, denoted using circle and marked as  to be multiplied with the input signals to provide the output.   Layer 5: The nodes in layer 5 represent the permanent node using the node function to evaluate the resultant output as Eqn. (9).
From the structure of ANFIS described above, the output is described as the linear arrangement of the subsequent factors. The resultant output is written as The ANFIS model is utilized for the trust assessment process, in this, the two inputs are represented as trust weight and weight of node j, and the output is termed as trust value. The conditions for the trust value assessment are described as follows.
Rule 1: If the trust weight value is high and the weight of node j value is high, then the trust value is high.

Rule 2:
If the trust weight value is low and the weight of node j value is low, then the trust value is low.

Rule 3:
If the trust weight value is medium and the weight of node j value is medium, then the trust value is medium.

Route discovery
The process of route discovery is utilized to establish the optimal route for the packet transfer in the routing protocol in MANET. Here Group teaching optimization algorithm is utilized for the optimal route discovery. The routing process for the trust aware routing protocol consists of two parts: routing within the region and routing between the regions.

Group teaching optimization algorithm
The group teaching optimization algorithm is designed to enhance the entire class knowledge by suggesting the group teaching scheme [24]. To change the group teaching appropriate for employing the optimization technique, first consider the population, fitness value, and decision variables are equivalent to the students, students knowledge, and subjects of the students, correspondingly. This algorithm consists of four phases like teacher allocation phase, teacher phase, capability grouping phase, and student phase. The four phases are explained in detail in the following section.

Capability grouping phase
The information regarding the entire class is considered as the normal distribution. The normal distribution is described as From Eqn. (11), a represents the value of the normal distribution function, p represents the average information of the entire class and  represents the standard deviation. The differences in information between students are represented as the standard deviation  . If the value of the standard deviation is large, then the knowledge differences between the students are also large. The perfect teacher imagines not only enhances the knowledge, also decreases the standard deviation value. For obtaining this objective, the teacher made an appropriate teaching layout for the students.

Teacher phase
The student gained knowledge from the teacher that responds to the described second rule in this phase. The teacher made various layouts for the average group of students and the outstanding group of this algorithm. In the teacher phase I, for the strong capability of the receiving knowledge the teacher aims on enhancing the knowledge of the outstanding student groups as the whole in this algorithm. Particularly, the teacher tries their best for enhancing the mean knowledge of the entire class. Additionally, the differences in receiving knowledge between the students are to be regarded. Hence the outstanding student group attains the knowledge using Eqn. (12).
From Eqn. (14), t represents the present iteration numbers; M represents the number of students, t j a represents the student's knowledge, t K represents the teacher's knowledge at time t, t N represents the average knowledge of the group at time t, f represents the teaching parameter of the teacher.
1 , represents the knowledge of the student j, x,y, and z represents the three arbitrary numbers in the interval [0, 1]. The F value is either 1or 2 as made in [25].
In teacher phase II, the teacher provides more concentration for the poor capability of receiving knowledge to the average group than the outstanding group. Hence the students of the average group attain gain their knowledge is described as From Eqn. (15), r represents the arbitrary number in the interval [0, 1].
Additionally, the student cannot achieve gain by the teacher phase is addressed as

Student phase
The student phase contains the student phase I as well as the student phase II respective to the declared third rule. In the extra period, the student attains the knowledge with two various schemes: one via the self-learning and the other via communication among the students that is described as represents the knowledge of the student k. The second term as well as the third term of Eqn. (17) indicates the right mean learning from the student as well as the self-learning. Additionally, the student cannot achieve knowledge with the student phase that is described as

Teacher allocation phase
Depends upon the described fourth rule, the teacher allocation method is more significant for enhancing the student's knowledge. The teacher allocation scheme is described as a represents the first, second, and third best students. To speed up the algorithms convergence, the outstanding group, as well as the average group shares the same teacher. By using the value of the teacher phase and student phase, two populations are constructed. These two populations together construct the new population which is the optimal value. So this optimal value is utilized for the efficient route discovery process. Thus by using the algorithm, route discovery is established for efficient data transmission between sources to destination. If the routing protocol is proactive, the routing protocol is classified into two categories: routing inside the zones and routing between the zones.
In the route discovery inside the zones, the central node transmits the route request message from the source to the destination and it discovers the new route or the middle node that routes the packet to the destination node. After the route is established, the destination node transmits the route confirmation message to the source. The route now established is to maintain the optimal path for forwarding the packets from source to destination.
In the route discovery between the zones, when the source sends the data packets to the destination, the node first checks the zone routing table for the route. If there is no existing route, then the route discovery between the zones are triggered. Initially, the source node checks for the multi-zone route table to evaluate if it has any unexpired routes. If the node is established, then the node sets the time and forwards the route request message for establishing the new route. When the node reaches the intermediate node, the node further checks the unexpired route in the multi-zone route table. If there is more than one route, the optimal route is selected and the message is transmitted back to the source node in the optimal path. When the destination node arrives at the intermediate node on the way to the source node, the new route is added to the multi-zone routing table. Hence the multi-zone routing tab contains the new route. When no route is established with the fixed time or threshold, the routing process fails. If there establishes multiple routes, then the adaptive equilibrium optimizer is utilized to find the optimal new route for the transmission of packets.

Optimal route selection procedure
The route discovery process utilizes the group teaching optimization algorithm. Due to the establishment of multiple routes, the optimal route selection is the main problem in the route discovery process, so adaptive equilibrium optimizer is utilized for the optimal route selection procedure.

Adaptive equilibrium optimizer
The adaptive equilibrium optimizer is proposed for the identification of the optimal route for the packet transmission. The adaptive equilibrium optimizer [26] is the advancement of the equilibrium optimizer [27]. This algorithm depends upon the arbitrary spreading of the search agents within the search space that is chosen from the fitness value. The equilibrium optimizer is encouraged by the equilibrium as well as the dynamic condition that depends upon the mass preservation law through leaving, generating, and entering the control volume.
Let us consider the search agents linked with the search space concentration and the iteration is initialized at 1  itr as pursues: These fitness values are organized in ascending order: Now the equilibrium candidates, as well as the fitness values, are expressed as: Ultimately, the pool of the equilibrium is described as: avg eq eq eq eq eq pool eq The exponential factor j E  is to help the equilibrium optimizer for the exploitation as well as the exploration is computed for the jth search agents as The j g  is described as From the above equations, grc represents the control factor of the generation rate and represents the arbitrary numbers in the range 0 and 1. The adaptive decision is made by using the present fitness and average fitness of the entire search agents and the minimization issue is described as The adaptive equilibrium optimizer takes over the concepts of memory storage from the equilibrium optimizer. So compare both the fitness values of the present iterations as well as the earlier iteration and then updated once it obtained the optimal fitness value and it is formulated as The node passing through the multiple zones are described in the route selection process. Let us consider that the source nodes tend to forward the data packets from the source to the destination and the route is selected for the transmission purpose. Initially, the route checks in the zone route tab to know whether there is any local route that matches with the earlier nodes after the source node. If there is more than one local route in the zone then AO is utilized to replace the existing route. This procedure is iterated until the packet arrives at the destination. Thus by the adaptive equilibrium optimizer (AO), the optimal route is explored for the efficient packet data transmission from source to destination.

Route maintenance procedure
Route maintenance is the process that manages the failures of the route particularly caused due to node mobility or faulty nodes that is more frequent in MANETs. If the damaged route links with the nodes inside the zone, several alternative routes are established in the zone routing table and an alternative route is selected that depends upon the optimization technique. When there is no alternative path, then the path is launched after some time due to the adoption of the proactive routing protocol.
If the damaged path links with the multiple zones, then the upstream node triggers the restricted repair method to establish the new route that is identical to the process of route discovery between the zones. If the alternate route is established, then the communication persists and the warning is forwarded to the source node for updating the multi-zone routing tab for every node on the path. If the error message is received but the source requires transmitting the data packets, then the source selects the alternate unexpired route present in the multi-zone route table and also starts the process of route discovery between the zones.

RESULTS AND DISCUSSIONS
The proposed approach is analyzed using the network simulator ns-3 and compare the simulation results with the other existing approaches like B-iHTRP [21], HyphaNet [28], QNewBee [29], and Crowwhale-ETR [30]. In the simulation environment, 100 nodes are consistently distributed in the area of 1200 x 1200 sq.m. For this scheme, every node moves in an arbitrary path at a speed uniformly distributed. When the node arrives at the target location, it stops the node for 5s rest and then moves identically. Here in this paper, malicious node detection is considered. The simulation is performed for the pause time ranging from 0 to 50 seconds. Table 2 describes the simulation parameters.

Performance metrics
The metrics utilized for the examination are packet delivery rate, throughput, packet drop rate, detection rate, routing overhead, delay, and energy consumption [32]. The description of the measures are described below. Detection rate: It is described as the percentage of successfully identified attackers in the network.
Routing overhead: It is the excess data forwarded employed for control. It is described as the proportion of the amount of control information to the total amount of transmitted information.
Energy consumption: The energy consumed by the network is described as the total energy necessitated by the nodes in accepting, broadcasting, and sending the information. Initially, every node assigned the starting energy and each energy level is calculated every time as the simulation measures.

Performance Analysis
The comparative analyses of the approaches are described in this segment, and the analysis is evaluated that depends upon the performance measures. This analysis is performed in the presence of malicious nodes to express the efficiency of the proposed approach. This section describes the comparative analysis with the help of 100 nodes with the existence of attacks. Fig 3 shows the comparative analysis of the throughput of the proposed approach with other existing approaches like B-iHTRP, HyphaNet, QNewBee, and Crowwhale-ETR. It is concluded that the throughput of the proposed approach is high when compared to other existing approaches. routing approaches. When the pause time is maximized, the packet delivery ratio is also increased. The proposed approach achieved better performance concerning the packet delivery ratio than other routing approaches. The results describe that the proposed approach achieved a high packet delivery rate than other state-of-art approaches.  The comparative analysis of the detection rate is demonstrated in Fig 6. The efficient approach must provide the largest detection rate that calculates efficient attack detection. From the figure, it is clear that the detection rate decreases with an increase in time but the proposed approach provides a high detection rate when compared to other existing approaches likeB-iHTRP, HyphaNet, QNewBee, and Crowwhale-ETR. The detection rate launches the opinion that differentiates the normal nodes from the malicious nodes, it detects the malicious nodes with more accuracy.

Fig 7:
Comparative analysis of the packet loss rate Figure 7 demonstrates that the comparative analysis of the packet loss rate for the proposed approach with other approaches such asB-iHTRP, HyphaNet, QNewBee, and Crowwhale-ETR. From the figure, it is proved that the proposed approach obtained a low packet loss rate when compared with state-of-art approaches. For this simulation, the routing overhead was computed for the proposed approach, B-iHTRP, HyphaNet, QNewBee, and Crowwhale-ETR by changing the time. Figure 8 depicts the routing overhead that happened during the routing process. The proposed approach has the ratio of routing overhead from 29.4% to 34.6%, B-iHTRP achieves from 34.3% to 36.2%, HyphaNet achieves from 34.3% to 38.5%, QNewBee achieves overhead from 38.7% to 61.4% and Crowwhale achieves from 43.4% to 64.4%. It clearly states that the routing overhead for the proposed approach is less when compared to other approaches.

CONCLUSION
The optimal route selection for efficient packet transmission is the most significant issue in MANET. In this paper, the ANFIS model is utilized for the determination of the trust evaluation. The group teaching optimization algorithm (GTA) is proposed for the route discovery scheme. By using this GTA scheme several routes are established. So the optimal route is selected with the help of the adaptive equilibrium optimizer (AO) algorithm. Then the route is maintained for effective packet transmission. This approach increases the trust and security in the packet transmission.Simulation results exposed that the proposed approach outperforms other approaches such as B-iHTRP, HyphaNet, QNewBee, and Crowwhale-ETR in terms of throughput, end-to-end delay, energy consumption, routing overhead, detection rate, packet loss rate, and packet delivery ratio. This approach improves the network throughput by accurate packet distribution, detection rate, decreases the delay, increases the packet delivery rate, reduced energy consumption as well as the packet loss rate via various