Black Hole and Selective Forwarding Attack Detection and Prevention in IoT in Health Care Sector: Hybrid meta-heuristic-based shortest path routing

Abstract

In the current health care scenario, security is the major concern in IoT-WSN with more devices or nodes. Attack or anomaly detection in the IoT infrastructure is increasing distress in the field of medical IoT. With the enormous usage of IoT infrastructure in every province, threats and attacks in these infrastructures are also mounting commensurately. This paper intends to develop a security mechanism to detect and prevent the black hole and selective forwarding attack from medical IoT-WSN. The proposed secure strategy is developed in five stages: First is selecting the cluster heads, second is generating k-routing paths, third is security against black hole attack, fourth is security against the selective forwarding attack, and the last is optimal shortest route path selection. Initially, a topology is developed for finding the cluster heads and discovering the best route. In the next phase, the black hole attacks are detected and prevented by the bait process. For detecting the selective forwarding attacks, the packet validation is done by checking the transmitted packet and the received packet. For promoting the packet security, Elliptic Curve Cryptography (ECC)-based hashing function is deployed. As the main contribution of this paper, optimal shortest route path is determined by the proposed hybrid algorithm with the integration of Deer Hunting Optimization Algorithm (DHOA), and DragonFly Algorithm (DA) termed Dragonfly-based DHOA (D-DHOA) by concerting the parameters like trust, distance, delay or latency and packet loss ratio in the objective model. Hence, the entire phases will be very active in detecting and preventing the two fundamental attacks like a black hole and selective forwarding from IoT-WSN in the health care sector.

Keywords

Internet of Things Wireless Sensor Network health care sector block hole attack selective forwarding attack shortest route path selection hybrid optimization algorithm

Nomenclature

ECC

Elliptic Curve Cryptography

DHOA

Deer Hunting Optimization Algorithm

Dragon Fly Algorithm

D-DHOA

Dragonfly-based DHOA

WSN

Wireless Sensor Network

IoT

Internet of Things

MAC

Medium Access Control

CRS-A

Channel-aware Reputation System by Adaptive detection threshold

Cuckoo Search

RREQ

Route Request

RREP

Route Reply

IDS

Intrusion Detection System

JDICA

Jensen-Shannon Divergence Based Independent Component Analysis

IBFWA

Integrated Bloom Filter with Watchdog Algorithm

Certificate Authority

MAODV

Multicast Ad-hoc On Demand Distance Vector Routing

1. Introduction

WSN is a self-organized system, which consists of low-priced devices. For reducing human interaction, these devices make use of actuators and sensors. In general, the WSN devices are employed in different places like health care, homes, military, etc. [12]. The thing that is considered in the IoT can be smart home appliances, a smartphone, human, smart grid Stations, autonomous cars, a health care monitoring device, smartwatches, and body sensing devices. The WSN devices are extended over a huge geographic surface. Despite its wireless connectivity and vulnerable nature of the network, it is open to various forms of attacks. Black holes, wormhole, selective forwarding, sinkhole, hello flood, jamming, sybil, and grey hole are some of the attacks that affect the WSN. Among the various attacks listed above, the black hole is considered the most hazardous and is the doorway for all remaining attacks [12].

In the WSN devices, the nodes that are attacked by the black hole are named as malicious nodes. The bad node shows the incorrect path as the best and shortest path to the target end in the black hole attacked network. The main aim of these intruder nodes is to cause huge traffic in the network and then to drop the packets without forwarding [12,18]. The black hole attack is a DOS attack that leads to cause humiliation of the network performance [13,17,28]. Thus, the malicious activity of the node that claims to have the minimum route path to the destination is called a black hole [5,22]. Moreover, the selective forwarding attack is another hugely challenging attack. A node affected with a selective forwarding attack purposely rejects the received packets that should be communicated, to obstruct with data transmission between nodes [14,35]. This obstruction of data badly affects the data-centric networks in the case of data fusion and data collection. Based on the research [20], few attacks are extraordinary cases of the selective forwarding attack like a black hole in which the node does not cooperate with data transmission, it drops the received data rather than sending it to the receiver [9,27,31]. Also, another variety of selective forwarding attacks is an On-off attack, in which the node works good and bad alternatively for hiding and avoiding itself from the reorganization during the attack [6]. The impacts on the MAC layer and inbuilt volatility and unreliability of the wireless channel unavoidably leads to normal packet losses which make difficulty in finding the difference between the losses caused by normal and malicious packets. Due to the loss of normal packet, the selective forwarding attacks are hidden causing difficulties in the detection of attacks. Hence it is essential to find selective forwarding attacks and to increase the network performance. In the existing works, more contributions have focused on observing the packet losses and disconnect the node with huge packet losses from the data sending path [2,8,24,32]. The techniques used in the previous works could help in increasing the packet delivery ratio or the throughput of the network but have difficulties in finding the selective forwarding attacks. Mainly, healthcare WSN is vulnerable to various attacks, in which the black hole attack is very difficult to find and resist.

The major part of the study done related to WSN security [7,33] consists of routing security, location security, intrusions key management, and prevention [19]. Conventionally, different approaches have been used to prevent malicious nodes from the clustered network by identifying the alternate routing way with intrusion detection and trust-based systems. Although the above-discussed methods have some advantages, yet the WSN networks are facing trouble in getting the trusted node in the clustered network with high energy efficiency and security.

The key contributions of the paper are as follows:

To detect and prevent the two basic attacks like a black hole and selective forwarding attacks in IoT-WSN using the bait process, and packet validation, respectively, after framing the network topology with selected cluster heads and generated route paths.

To employ the ECC-based packet security model to perform the secure data packet transmission from source to destination.

To frame an optimal shortest route path selection using the novel hybrid algorithms termed as D-DHOA, which focuses the objective parameters like trust, distance, latency, and packet loss ratio.

The entire paper is designed in the following manner: Section 2 specifies the literature review and the features and challenges of existing black hole and selective forwarding attacks detection and prevention methodologies. Architectural representation of proposed black hole and selective forwarding attack detection and prevention is shown in Section 3. Section 4 describes the adopted stages for black hole and selective forwarding attack detection. Selection of secure routing paths: shortest path determination using the proposed hybrid algorithm is given in Section 5. Results and discussions of the paper are given in Section 6. The entire paper conclusion is given in Section 7.

2. Literature review

2.1. Related works

In 2016, Ren et al. [30] offered CRS-A for identifying selective forwarding attacks in WSNs. This method has validated the transmitting behaviours of sensor nodes based on the variation of the loss of packet that was observed and guess the common loss. Also, the optimal threshold was derived from optimizing the detection accuracy of CRS-A. The attack-tolerant information forwarding strategy was introduced for integrating with CRS-A to activate the forwarding assistance of weakens nodes and enhances the information distribution ratio of the network. The simulation outcomes of the proposed system revealed that CRS-A was detecting selective forwarding attacks and recognizing the weaken sensor nodes accurately, where the attacks-tolerant data transmission system was extensively increasing the data distribution ratio of the network.

In 2016, Mathur et al. [23] focused on two risks like a black hole and Selective Forwarding attacks, which caused because of wireless routing among the nodes and access points of medical WSN. To solve the existing challenge, cryptographic hashes were eagerly offered, whereas the last-mentioned problem utilized the neighbourhood watch and threshold-based evaluation for detecting and preventing selective forwarding attacks. Here, the suggested system was able to spot the selective forwarding attack with an accuracy of 96% and effectively recognizing the unauthorized node with an accuracy of 83%.

In 2019, Mehetre et al. [24] proposed a trustworthy and secure routing algorithm utilizing a two-stage security model, and twin assurance method, to choose the node and protecting the information packet in WSN. The methods were dependent on Active Trust to secure many types of attacks like black hole attack, and selective forwarding attack, throughout the routing. Hence, the proposed system recognized the trusted path and produced the protective routing paths via a trust and CS algorithm. The developed system considered energy as the performance parameter. The test results have verified that the recommended approach produced the guarantee for extending the lifetime of the network as well as the possibility of a secure routing path in the network.

In 2019, Delkesh and Jamali [8] suggested a routing algorithm that depends on the delivery of forged packets to improve the detection accuracy and eliminate the malignant nodes. By the suggested approach, malignant nodes present in the network were identified by transferring forged RREQ and RREP routing packets that consist of the address of an imaginary destination node. By sending an RREP message, the nodes were removed from the routing tables. The developed technique was capable of enhancing the traffic load in the network, recognizing a feasible and secure route, identifying the count of malignant nodes, and optimizing few parameters. Moreover, test outcomes demonstrated that the percentage of the delivered data packets by the suggested method was superior to the IDS algorithm.

In 2019, Sunder and Shanmugam [32] introduced the JDICA method for diagnosing the black hole attack. The JDICA approach recognized the black hole attack by examining the physiological information collected from biomedical sensors. Moreover, attack detection was performed based on the behaviors of the sensor nodes. The divergence result has shown that the suggested JDICA approach has detected the black hole attack with extreme accuracy and also assists to quarantine the malignant node from the network by transmitting the isolation information to entire sensor nodes in the network. Therefore, it has been proved that the JDICA approach improved the detection of black hole attack nodes when compared to conventional techniques, thus improving the packet delivery ratio and decreasing the delay.

In 2012, Arunmozhi and Venkataramani [2] recommended a defence approach to identify black hole node, based on the timely data and objective series numbers preserved in the Neighbourhood Route Monitoring Table that documents the time of reply. The decreased reply time was utilized to find the black hole node. The end series number was verified with the threshold value for enhancing the security that was rationalized vigorously. Hence, the experimental outcomes confirmed that the protocol identifies the black hole attack and enhances the network’s lifetime.

In 2017, Kollati and Somasundaram [16] introduced IBFWA for shunning loss of the packet. At the time of key generation, the node was evaluated by the CA. Later, it examined that the node was either normal or attacker. If it was a black hole node, then that will be jammed and communication through that node was avoided or if the node was undermined and reversed as a normal node, then it was connected to the network and through that, node communication was authorized. Finally, the test results of the suggested approach were validated based on the performance regarding lower node outage, residual energy, end-to-end delay, high detection probability, throughput, packet transmission rate, and packet distribution ratio.

In 2009, Anita and Vasudevan [1] suggested an improved certificate dependent authentication method, in which the nodes validate each other by providing certificates to adjacent nodes and creating a public key without using online authority. Moreover, the MAODV protocol acts as assistance for certification. Therefore, the efficiency of the developed approach was clarified by simulations done via network simulator ns-2.

Table 1
Features and challenges of existing black hole and selective forwarding attack detection and prevention methodologies

Authors [Citations] Methodology Features Challenges

Ren et al. [30] Adaptive detection threshold Validates the long-term forwarding behaviours. Loses more data.

Has the best performance.

Mathur al. [23] Cryptographic hashes Protect the data from unauthorized users. Difficult to access the data even for a valid user.

Digital signatures are used to secure the data.

Mehetre et al. [24] CS algorithm It is simple and easy to implement. It simply falls into the local optimal solution.

Deals with multi-criteria optimization problems. Has a slow rate of convergence.

Delkesh and Jamali [8] AODV Responds fast to the topological alterations that affect the active routes. If the size of the network increases, then different performance metrics start reducing.

Has a lower setup delay for connections. Takes more share of bandwidth.

Sunder and A. Shanmugam [32] JDICA Increases the detection rate. The false alarm rate is not detected effectively.

Reduces the time of detection.

Arunmozhi and Venkataramani [2] AODV It can assist both unicast as well as multicast packet diffusions, even for nodes in constant movement. Takes more time to build a routing table.

The route discovery process is more effective in dynamic nature. Has high processing demand.

Kollati and Somasundaram [16] IBFWA Capable of tracing an attack after a long time of its completion. Requires fine-tuning.

Supports a high degree of node mobility and improves network performance.

Anita and Vasudevan [1] MAODV Reduces the network load when the packets are transferred to the group of nodes. It consumes more memory.

It has high efficiency.

Authors [Citations]	Methodology	Features	Challenges
Ren et al. [30]	Adaptive detection threshold	Validates the long-term forwarding behaviours.	Loses more data.
Has the best performance.
Mathur al. [23]	Cryptographic hashes	Protect the data from unauthorized users.	Difficult to access the data even for a valid user.
Digital signatures are used to secure the data.
Mehetre et al. [24]	CS algorithm	It is simple and easy to implement.	It simply falls into the local optimal solution.
Deals with multi-criteria optimization problems.	Has a slow rate of convergence.
Delkesh and Jamali [8]	AODV	Responds fast to the topological alterations that affect the active routes.	If the size of the network increases, then different performance metrics start reducing.
Has a lower setup delay for connections.	Takes more share of bandwidth.
Sunder and A. Shanmugam [32]	JDICA	Increases the detection rate.	The false alarm rate is not detected effectively.
Reduces the time of detection.
Arunmozhi and Venkataramani [2]	AODV	It can assist both unicast as well as multicast packet diffusions, even for nodes in constant movement.	Takes more time to build a routing table.
The route discovery process is more effective in dynamic nature.	Has high processing demand.
Kollati and Somasundaram [16]	IBFWA	Capable of tracing an attack after a long time of its completion.	Requires fine-tuning.
Supports a high degree of node mobility and improves network performance.
Anita and Vasudevan [1]	MAODV	Reduces the network load when the packets are transferred to the group of nodes.	It consumes more memory.
It has high efficiency.

2.2. Review

Although there would be many detection and prevention methodologies for black hole and selective forwarding attack, there are few challenges that need to be overcome in the future. Table 1 describes some of the pros and cons of conventional methodologies for black hole and selective forwarding attack detection and prevention in IoT using WSN. Among them, the Adaptive detection threshold [30] validates long-term forwarding behaviors, and has the best performance. But, it loses more data. Cryptographic Hashes [23] protects the data from unauthorized users, and digital signatures are used to secure the data. Though, it is difficult to access the data even for a valid user. CS [24] is simple and easy to implement and deals with multi-criteria optimization problems. Still, it is having some defects like it simply falls into the local optimal solution, and has a slow rate of convergence. AODV [2,8] responds fast to the topological alterations that affect the active routes, has lower setup delay for connections, it can support both unicast as well as multicast packet diffusions, even for nodes in constant movement, and route discovery process is more effective in dynamic nature. However, it is having some defects like if the size of the network increases, then different performance metrics start reducing, takes more share of bandwidth, takes more time to build routing table, and has high processing demand. JDICA [32] increases the detection rate and reduces the time of detection. Still, the false alarm rate is not detected effectively. IBFWA [16] is capable of tracing an attack after a long time of its completion and supports a high degree of node mobility and improves the network performance. Yet, it requires fine-tuning. MAODV [1] reduces the network load when the packets are transferred to a group of nodes and have high efficiency, but it consumes more memory. Thence, the above-specified defects are considered and the early detection and prevention of attacks need to be improved in upcoming researches.

3. Architectural representation of proposed black hole and selective forwarding attack detection and prevention

3.1. Proposed model

The proposed architecture of the clustered IoT-WSN model in the health care sector is shown in Fig. 1. The main intention of the proposed model is to accomplish the shortest route path for node communication, which leads to maximize the lifetime of sensor nodes and networks as well. Hence, a trustable secure routing scheme is implemented in clustered IoT-WSN with attack detection and prevention mechanisms to attain secure communication. The proposed model involves five stages: (1) Cluster head selection (2) k-paths generation (3) Detection and prevention of black hole attack (4) Detection and prevention of selective forwarding attack, and (5) Optimal shortest route path selection. Initially, a set of nodes in the network are randomly selected as cluster heads. Once the cluster heads are selected, the next step is to select the k-paths between the cluster heads to the base station or sink node. Before finding the optimal path from generated k-paths, the third stage undergoes the detection and prevention of a black hole attack. The bait process is used to detect the black hole attack, and the concerned malicious node is eliminated by the system. In the bait process, the attack node is detected from the feedback packets replied from all destination nodes. The next stage process with the identification of selective forwarding attacks, which is detected by packet validation. The malicious node by selective forwarding attack may drop some of the data packets and forward remaining to the next node. These types of nodes will also be prohibited by the system after detection and even use ECC-based packet security further to perform efficient communication without any packet loss. After confirming the security of the network with proper detection and prevention of attack mechanisms, the optimal shortest path is selected in the entire path transmission using the new hybrid D-DHOA model, which considers the objective constraints like the trust of the node, the distance between the nodes, delay of transmission, and packet loss ratio. Finally, energy dissipation has to be computed for the cluster heads involved in communication, and the number of alive nodes will be calculated in each round. Hence, the optimal shortest route path is accomplished in IoT-WSN by protecting the network from the black hole and selective forwarding attacks.

Fig. 1.

Proposed IoT-WSN model in health care environment: contribution towards shortest route path selection by detecting and preventing black hole and selective forwarding attack.

3.2. System model of cluster head formation

This section discusses the cluster head formation model in IoT-WSN, in which the system model is diagrammatically shown in Fig. 2. In general, WSN involves multiple numbers of sensor nodes represented as j, from which the data is transmitted to only one sink node or base station denoted as $J_{B}$ . The wireless connection indicates direct data transmission between the nodes in the radio range. The transmission of data packets in WSN is performed reliably by every node, which is happened within the dimension X and Y in meters. The optimal gathering of sensors forms a cluster, with each sensor having its own ID. In general, the base station has to be located at such a point, where the overall data transmission of the network is smooth, which is obviously $(0.5 X, 0.5 Y)$ . Using efficient routing protocols, the sensor nodes in the network transmits data to the cluster head, and then cluster head to the sink node. In Fig. 2, $N_{O}$ indicates the normal nodes and $M_{q}$ indicates the nodes that are activated as cluster head.

Fig. 2.

Model of clustered IoT-WSN with cluster head selection.

If $N_{o}^{j}$ is the set of sensor nodes present in the cluster, and $M^{*}$ is the number of clusters or cluster heads, then the computation of normal sensor nodes is equivalent to $j - M^{*}$ . Thus, the data packet transmission within the network is carried out from every normal sensor node $N_{o}^{i}$ to its corresponding cluster head $M_{q}$ . Further, the cluster heads collect the entire received data packets and transmit it to the base station $J_{B}$ . Once after the optimal placements of entire sensor nodes in the network, the distance between the $u^{t h}$ normal node to the $v^{t h}$ cluster head is represented as $K_{u v}$ , and the distance between the $v^{t h}$ cluster head $J_{B}$ is represented as $Q_{v J_{B}}$ .

4. Adopted stages for black hole and selective forwarding attack detection and prevention

4.1. K-paths generation

When the cluster head selection is ready in the IoT-WSN, the selection of the source node $n_{s}$ and destination node $n_{d}$ has to be selected using a graph structure. It is done in such a way that the user can select the source node and destination node along with ID. Further, the process continues with selecting the k-paths for determining the shortest path in the network. This k-paths routing process is depended on the correlation of determining the minimum count of possible bandwidth or shortest path between two cluster heads in the transmission fields, which confirms that the entire edge in the network has attained the minimum bandwidth value. Here, different shortest route paths are selected randomly, from which the optimal path is determined by the proposed D-DHOA, which will be considered as the best route. However, before finding the optimal route, the proposed model tries to identify the black hole and selective forwarding attack, and prevent it using bait process, and packet validation, respectively. The structural frame of k-paths route selection and attack node removal is shown in Fig. 3.

Fig. 3.

K-paths route selection and attack node removal.

In Fig. 3, assume, there are 7 cluster heads, in which different k-paths are selected. Herein, A indicates the attack node, which tries to attack the whole network.

4.2. Blackhole attack detection by the bait process

A black hole problem means that “one malicious node utilizes the routing protocol to claim itself of being the shortest path to the destination node, and drops the routing packets by not forwarding the packets to its neighbours”. The key objective of the bait process [4] is to detect the malicious node or black hole attack node while trying to send a reply $RREP$ for the request ${RREQ}^{'}$ from the source node. Assume an adjacent node $n_{A}$ , which is randomly selected by the source node. Initially, the source node sends a request message ${RREQ}^{'}$ which contains the information like source ID ${IDn}_{s}$ , destination ID ${IDn}_{d}$ , and path length $pl$ based on Eq. (1). $\begin{matrix} (1) & {RREQ}^{'} = {{IDn}_{s}, {IDn}_{d}, pl} \end{matrix}$

Here, the path length $pl$ provides the information about the count of hops that forward the transmitted request from the source to the destination node. Once the request is reached to every node, they will send the feedback in the form of a reply $RREP$ based on Eq. (2) that signifies that the request successfully reached the final node. $\begin{matrix} (2) & RREP = {{IDn}_{s}, {IDn}_{d}, pl} \end{matrix}$

Once the feedback packet or $RREP$ is reached by the source node, it compares with ${RREQ}^{'}$ . In the case of a malicious node, the path length $pl = 0$ , this signifies that the particular node acts as the destination. Since the destination and path length information is already stored in ${RREQ}^{'}$ , it can easily find the malicious node. If ${IDn}_{d}$ in ${RREQ}^{'}$ is equal to ${IDn}_{d}$ , and $pl = 0$ , then the corresponding node is considered as the genuine destination node. Moreover, the network discards the attack node if such a case occurs. Hence the black hole attack is detected by the bait process and is successfully removed from the system, and selects an alternate path for communication.

4.3. Selective forwarding attack detection by packet validation

Selective forwarding [10,34] attack means “the attack node tries to stop the packets in the network by refusing to forward or drop the messages passing through them”, which is also termed grey hole attack. The packet Validation process is done for diagnosing the selective forwarding attack and the routing protocol removes such attack from the network, once it is detected. Consider the packet information as $OP$ , which have information like data type $DT$ , and data size $DS$ based on Eq. (3). $\begin{matrix} (3) & OP = {DT, DS} \end{matrix}$

If a node transmits the data packet to another node, the receiver node checks the data type and data size. If the data type is not matched with the data size, it is confirmed that the transmitted node drops some packets, considered as the malicious node. Further, this node will be removed from the system and thus could minimize the packet loss ratio of communication.

4.4. Packet security model using ECC

Even though the malicious nodes are detected by the packet validation process, ECC-based encryption [15] is also deployed here for providing more security to the data packets. When some information is transferred from one place to another place in the network, the information file is divided by the protocol into tiny sizes for a well-organized routine. Moreover, each packet is independently recognized by number, and it consists of the address of the target node in the network. The isolated data packets are transferred from different routes present in the network. The packets are recreated to a regular file once the overall data packets reached the destination. In the meanwhile, the loss of a packet has happened if there is any attacker in the node. Thus, the network model has offered the security technique for transferring the data packet without losing the information for keeping the network secure. Here, the security approach protects the information from the intruder using the ECC algorithm. The information is encrypted by the algorithm before routing. To verify the data integrity at the sensor node, the SHA-1 hashing technique is employed, which is functioned as follows.

The data present in the source node is transformed into packets.

Later, every packet is encrypted by the ECC encryption algorithm.

For the obtained encrypted packet, the hash value is generated by the SHA-1 hashing method.

Consequently, the network forwards the encrypted packet and its related hash value in the chosen reliable path.

At every receiving node, the hash value is computed for the received encrypted packet.

The data is known to be secured when new hash values are the same as the received hash values, which represents that the node is safe from the intruder.

5. Selection of secure routing paths: Shortest path determination using proposed hybrid algorithm

5.1. Objective model for shortest path selection

The shortest route path selection is performed after detecting and preventing the black hole and selective forwarding attacks from the network. The main objective model of the secured IoT-WSN model considers trust, energy, latency, and packet loss ratio.

Trust: Trusty communication [21] between the nodes is validated by determining the relationship between two nodes. The value of trust can be defined as the scale that one node expects a certain service from another node. During the routing, trust computation is deployed to estimate the trust node for avoiding the black hole and selective forwarding attacks. Assume two nodes $N_{1}$ and $N_{2}$ are communicating in a particular duration $B_{k}$ , and the trust node with successful routing is indicated as $Δ_{N_{1}}^{N_{2}} (B_{k})$ , whereas the other trust node is defined as $Λ_{N_{1}}^{N_{2}} (B_{k})$ . Thus the computational formula for trust is depicted in Eq. (4). $\begin{matrix} (4) & TN = \sum_{k = 1}^{K N} {Δ_{N_{1}}^{N_{2}} (B_{k}) / Λ_{N_{1}}^{N_{2}} (B_{k})} h e (k) / K N \end{matrix}$

Here, $h e (k)$ denotes the attenuation function, whose value ranges between 0 and 1, and $K N$ indicates the total number of interactions between the nodes.

Energy: The average energy of the cluster heads present for communication after its energy dissipation is computed here based on Eq. (5), where the cluster heads $u = 1, 2, \dots, M^{*}$ . $\begin{matrix} (5) & E n_{a v} = \frac{1}{M^{*}} \sum_{1}^{M^{*}} E n_{a v}^{u} \end{matrix}$

Latency: Latency is defined as “the duration among the response and stimulation or a delay of time between the source and the target of the physical modification in the system being examined”. Energy loss can be mostly avoided by computing the latency. The mathematical formula for calculating latency is given in Eq. (6). $\begin{matrix} (6) & LA = t i^{s} \times t i^{d} \end{matrix}$

In Eq. (6), $t i^{s}$ represents the time duration for packets transmitted from the source node and $t i^{d}$ represents the time duration for packets received the destination node.

Packet Loss Ratio: It is defined as “the ratio of data received at the destination to the data transmitted from the source”. The formulation is depicted in Eq. (7) $\begin{matrix} (7) & PL = \frac{{DP}_{D}}{{DP}_{S}} \times 100 \end{matrix}$

Here, ${DP}_{D}$ indicates the transmitted data, and ${DP}_{S}$ indicates the received data.

The overall objective function for the trusty shortest path computation is given below. $\begin{array}{l} (8) & q_{1} = α {Fit}_{1} + (1 - α) {Fit}_{2} \\ (9) & q_{2} = β / {Fit}_{3} + (1 - β) q_{1} \\ (10) & q_{3} = λ / {Fit}_{4} + (1 - λ) q_{2} \\ (11) & Final Fit = max (q_{3}) \end{array}$

Here, ${Fit}_{1}$ is calculated based on Eq. (12), where $ED$ is the Euclidean distance between the nodes, which is defined based on Eq. (13), in which e and f are the cluster heads taken for communication and $y e z e$ are the coordinates of those nodes. $\begin{array}{l} (12) & {Fit}_{1} = \frac{TN}{ED} \\ (13) & ED = \sqrt{(e_{y e} - f_{y e}) + (e_{z e} - f_{z e})} \end{array}$

In addition, the other functions ${Fit}_{2} = E n_{a v}$ , ${Fit}_{3} = LA$ , and ${Fit}_{4} = PL$ . This multi-objective problem is solved by the proposed D-DHOA in the current secured IoT-WSN model.

5.2. Solution encoding

In order to find the secured shortest route path, the developed D-DHOA is used, which selects the best between the cluster heads from the IoT-WSN system. The solution pattern before encoding is shown in Fig. 4, in which $M_{q}$ denotes the cluster head, and $M^{*}$ denotes the number of clusters. After detection and prevention of black hole and selective forwarding attacks, the cluster heads which are in progress are taken for selecting the shortest route path. Consider $M_{q}^{1}$ as the source node, and $M_{1}^{8}$ as the destination node. While encoding, the source node has to be placed at the first element of the solution, and the nodes after the destination node have to be removed. Accordingly, the pattern of a solution after encoding is shown in Fig. 5.

Fig. 4.

Solution pattern before encoding.

Fig. 5.

Solution pattern after encoding.

5.3. Conventional dragonfly algorithm

The DA algorithm [11] takes the inspiration from the static and dynamic swarming characteristics of the dragonfly. The exploitation and exploration are the two main steps in optimization which are alike to the two swarming behaviours of the dragonfly. In exploration (static swarm), the local movements along with abrupt changes in the flying pathway are the most important features. In exploitation (dynamic swarm), large numbers of dragonflies form a group for moving in a single direction for elongated distance. The important goal of any swarm is to survive so that it has to get attracted to the food and to repel from the enemies. Distraction, separation, alignment, attraction, and control cohesion are the five important factors seen during updating the position of individuals in the swarm when considering the two swarming behaviours of the dragonfly. The derivation that denotes the separation of the $i_{t h}$ dragonfly from its neighbours is given in the Eq. (14). $\begin{matrix} (14) & P_{i} = \sum_{j = 1}^{L^{'}} (D^{'} - D_{j}^{'}) \end{matrix}$

Here, the position of the current individual is denoted by $D^{'}$ and $D_{j}^{'}$ denotes the position of the $j_{t h}$ neighbouring individuals. The term $P_{i}$ denotes the separation of the $i_{t h}$ dragonfly from its neighbour. The total number of neighbouring individual is denoted by $L^{'}$ . The mathematical formulation of alignment is given in the Eq. (15). $\begin{matrix} (15) & G_{i} = \frac{\sum_{j = 1}^{L^{'}} C_{j}^{'}}{L^{'}} \end{matrix}$

In the Eq. (2) $C_{j}^{'}$ denotes the $j_{t h}$ neighbouring individual. $G_{i}$ denotes the alignment of the $i_{t h}$ individual. The mathematical formulation of cohesion is given in the Eq. (16). $\begin{matrix} (16) & H_{i} = \frac{\sum_{j = 1}^{L^{'}} D_{i}^{'}}{L^{'}} - D^{'} \end{matrix}$

The term $H_{i}$ indicates the cohesion of the $i_{t h}$ individual. The mathematical formulation for the attraction of dragonfly towards food is given in the Eq. (17). $\begin{matrix} (17) & O_{i} = F - D^{'} \end{matrix}$

In Eq. (4) F denotes the position of the food resource. $O_{i}$ denotes the food resource of the $i_{t h}$ individual. The mathematical formulation for the distraction of dragonflies from the enemy is given in the Eq. (18). $\begin{matrix} (18) & M_{i} = E + D^{'} \end{matrix}$

In the Eq. (18) E denotes the position of the enemy. $M_{i}$ denotes the position of the enemy of the $i_{t h}$ individual.

Two vectors are considered for updating the position of the dragonflies in the search space and for simulating their movements: step vector $Δ D^{'}$ and position $D^{'}$ . The direction of the movement of the dragonflies is given by the step vector which is derived in the Eq. (19). $\begin{matrix} (19) & Δ D_{t + 1}^{'} = (p^{'} P_{i} + g^{'} G_{i} + h^{'} H_{i} + o^{'} O_{i} + m^{'} M_{i}) + δ . Δ D_{t}^{'} \end{matrix}$

In the Eq. (19) $p^{'}$ denotes the weight of the separation, $g^{'}$ denotes the weight of the alignment, $h^{'}$ denotes the weight of the cohesion, $o^{'}$ denotes the food factor, $m^{'}$ denotes the enemy factor, δ is the weight of the inertia, and t represents the number of iterations. The mathematical formula for calculating the position vector is given in the Eq. (20). $\begin{matrix} (20) & D_{t + 1}^{'} = D_{t}^{'} + Δ D_{t + 1}^{'} \end{matrix}$

In the DA algorithm, during exploration, of search space, the low cohesion and high alignment weights are considered and during exploitation, the search space the high cohesion and low alignment weights are considered. A random walk (Levy flight) and the stochastic behaviour and exploration are used to fly around the search space when there is no nearby solution obtained. The Levy flight is used to improve the randomness. The mathematical derivation used to update the position of the dragonflies without any neighbours is given in the Eq. (8). $\begin{array}{l} (21) & D_{t + 1}^{'} = D_{t}^{'} + Levy (x) \times D_{t}^{'} \\ (22) & Levy (x) = 0.01 \times \frac{n_{1} \times Φ}{| n_{2} |^{1 / g}} \\ (23) & Φ = {(\frac{Γ (1 + ξ) \times sin (\frac{π ξ}{2})}{Γ (\frac{1 + ξ}{2}) \times ξ \times 2^{(\frac{ξ - 1}{2})}})}^{\frac{1}{ξ}} \\ Γ (x) = (x - 1)! \end{array}$

In the above Eq. (21), Eq. (22) and Eq. (23), the dimension of the position vector is denoted by g, the random numbers within the interval $[0, 1]$ is represented by $n_{1}$ and $n_{2}$ , the value ξ is constant. To determine the neighbourhood of each dragonfly, the Euclidean distance between all the dragonflies is calculated and $L^{'}$ number of dragonflies is selected, for updating $Δ D^{'}$ and $D^{'}$ vectors. The process of updating goes on till the required criteria are obtained. The algorithmic representation of conventional DA is shown in Algorithm 1.

Algorithm 1

Pseudo code for DA [11]

5.4. Conventional deer hunting optimization algorithm

The main goal of the DHOA [3] is to obtain the best position of the human to capture the deer. The deer also known as buck has some special characteristics, which help them in escaping from the predator like it has good visual sense, good smelling ability and can detect ultra-high-frequency sounds. The process of DHOA is carried out in four steps. The first step in the DHOA is to initialize the population of the hunters is shown in the Eq. (24). $\begin{matrix} (24) & D = {D_{1}, D_{2}, \dots, D_{r}}; 1 < s ⩽ r \end{matrix}$

In the Eq. (24) the total number of hunters is represented by r. The population of the hunters is represented by D.

The second step to be followed in the DHOA is to initialize the two important parameters that help in determining the optimal positions of the hunters: wind angle and position angle. The formula for wind angle is based on the circumference of the circle, which is given in the Eq. (25). $\begin{matrix} (25) & θ_{t} = 2 π d \end{matrix}$

The wind angle is denoted by θ. The random number whose value ranges from $[0, 1]$ is denoted by d and t denotes the current iteration. The mathematical formula for the position angle is given in the Eq. (26). $\begin{matrix} (26) & φ_{t} = θ + π \end{matrix}$

The position angle is denoted by φ in the Eq. (13). The third step in the DHOA is position propagation. Initially, the position of the optimal space is not known, so it is assumed to be near the best solution that is found through the fitness function. Two position are considered that is leader position ( $D^{lead}$ ) and successor position ( $D^{lead}$ ).

Propagation through a leader’s position: The entire solution or individual is updated to attain the best position as reached by the leader. The mathematical formula for the encircling behaviour is given in the Eq. (27). $\begin{matrix} (27) & D_{t + 1} = D^{lead} - Y . v . | T \times X^{lead} - D_{t} | \end{matrix}$

In the Eq. (27) $X_{t}$ represents the position in the current iteration, $X_{t + 1}$ represents the position in the next iteration, the random variable that is formed based on the wind speed which ranges from 0 to 2 is represented by v and Y, T are the coefficient vectors, which is calculated by the formula given in the Eq. (28) and Eq. (29) respectively. $\begin{array}{l} (28) & Y = 1 / 4 log (t + \frac{1}{t_{max}}) a \\ (29) & T = 2 . b \end{array}$

Here, the term $t_{max}$ represents the maximum number of iteration, the value of the parameter a used lies between −1 and 1. In the Eq. (16) b is a random number between the interval $[0, 1]$ .

Propagation all the way through position angle: The DHOA is expanded by considering the position angle in the update process for improving the search space. To make the hunting process effective, consideration of position angle is essential. The mathematical formula for the angle of visualization $c_{t}$ of the prey is given in the Eq. (30). $\begin{matrix} (30) & c_{t} = \frac{π}{8} \times d \end{matrix}$

For updating the position angle, a new parameter $l_{t}$ is derived from the difference between the wind angle and the angle of visualization is given in the Eq. (31). $\begin{matrix} (31) & l_{t} = θ_{t} - c_{t} \end{matrix}$

The formula for the next iteration in updating the position angle is given in the Eq. (32) $\begin{matrix} (32) & φ_{t + 1} = φ_{t} + l_{t} \end{matrix}$

The position update formula derived with considering the position angle is given in the Eq. (33). $\begin{matrix} (33) & D_{t + 1} = D^{lead} - v . | cos (w) \times D^{lead} - D_{t} | \end{matrix}$

Propagation all the way through the position of the successor: Hereby adjusting the vector T, the encircling idea is been adopted in the exploration stage. Though random search is assumed initially, the value of the vector T is considered less than 1. The position update based on the position of the successor is given in the Eq. (34). $\begin{matrix} (34) & D_{t + 1} = D^{successor} - Y . v . | T \times D^{successor} - D_{t} | \end{matrix}$

The search agent is selected randomly when the value of T is less than 1 and the best solution is selected when the value T is greater than or equal to 1, to update the agent’s position.

The last step in the DHOA is the termination of the process. Here the position update is done for all iteration till the best position is obtained. Algorithm 2 shows the algorithmic representation of conventional DHOA.

Algorithm 2

Pseudo code for DHOA [3]

5.5. Hybrid algorithm for shortest path routing

Although lots of researchers have been done on a heuristic concept with several beneficial parts for solving the different application-oriented optimization problems, each optimization concept has to be enhanced for getting adaptable under various conditions. DHOA is an advanced well-performing optimization algorithm. However, it imposes limitations on some non-convex and unsmooth nature of the problems. To further improvise the performance of DHOA, another renowned algorithm called DA is merged with it in the current proposed model for finding the secured shortest route path. The main advantageous part of DA is its high probability of selecting the best solution from a set of solutions. In the conventional DHOA, the position is updated by Eq. (33) for the condition ( $v ⩾ 1$ ). Moreover, the proposed D-DHOA updates the solution for the same condition using Eq. (20) of DA. Accordingly, the pseudo-code of proposed D-DHOA is illustrated in Algorithm 3.

Algorithm 3

Pseudocode for proposed D-DHOA

The proposed system flow is shown in Fig. 6. The main objective of the developed IoT-WSN overview in the health care sector is to find the secure shortest route path based on the constraints like trust, energy, latency, and packet loss ratio of the cluster heads taken for communication to extend the network lifetime.

Fig. 6.

Flow chart representation of proposed system model.

5.6. Energy dissipation of communicated nodes

Initially, every sensor node contains some initial energy, which is denoted as ${Eg}_{ini}$ . Radio electronics and power amplifiers are utilized by the sender for dissipating the energy. Two energy dissipation models are followed for the energy dissipation of each packet transfer of size ${pc}_{sz}$ depending on the nature and the distance of the sensor node whether the node is a normal node or cluster head. While transferring the packet size ${pc}_{sz}$ , the energy dissipation of the normal node is described in Eq. (35), and Eq. (36). For computing the distance ${dis}_{0}$ , the equation is shown in Eq. (37), which ${Eg}_{fs}$ is the energy required in the utilization of free space, and ${Eg}_{am}$ be the energy of the power amplifier. $\begin{array}{l} (35) & {Eg}_{dis} (N^{u}) = {Eg}_{ee} * {pc}_{sz} + {Eg}_{am} * {pc}_{sz} * {‖ N^{u} - M^{v} ‖}^{4}; if ‖ N^{u} - M^{v} ‖ ⩾ {dis}_{0} \\ (36) & {Eg}_{dis} (N^{u}) = {Eg}_{ee} * {pc}_{sz} + {Eg}_{am} * {pc}_{sz} * {‖ N^{u} - M^{v} ‖}^{4}; if ‖ N^{u} - M^{v} ‖ < {dis}_{0} \\ (37) & {dis}_{0} = \sqrt{\frac{{Eg}_{fs}}{{Eg}_{am}}} \end{array}$

Electronic energy depending on distinct factors such as modulation, amplifier, filtering, spreading, and digital coding is denoted as ${Eg}_{ee}$ that is represented in Eq. (38), where transmitter energy is indicated as ${Eg}_{te}$ , whereas data aggregation is denoted as ${Eg}_{da}$ , and the distance among the $u^{t h}$ normal sensor node and $v^{t h}$ cluster head is indicated as $‖ N^{u} - M^{v} ‖$ . To define the energy dissipation of cluster head for transferring ${pc}_{sz}$ bytes of data, the mathematical formula is based on Eq. (39). Then, the energy value of each node is updated when the packet is transferred or received. The transferring process is done until the node is said to be dead, which means the node is containing the energy below zero. $\begin{array}{l} (38) & {Eg}_{ee} = {Eg}_{te} + {Eg}_{da} \\ (39) & {Eg}_{dis} (M^{v}) = {Eg}_{ee} * {pc}_{sz} \end{array}$

Moreover, the numerical equation to ensure the energy value of each node after transferring and receiving of data is shown in Eq. (40), and Eq. (41). $\begin{array}{l} (40) & {Eg}_{en + 1} (N^{u}) = {Eg}_{en} (N^{u}) - {Eg}_{dis} (N^{u}) \\ (41) & {Eg}_{en + 1} (M^{v}) = {Eg}_{en} (M^{v}) - {Eg}_{dis} (M^{v}) \end{array}$

6. Results and discussions

6.1. Experimental setup

The proposed detection and prevention of black hole and selective forwarding attacks in medical IoT-WSN was implemented in MATLAB 2018a, and the performance evaluation was performed. In the developed IoT-WSN system, 35 nodes are assigned as cluster heads, the initial energy of every node was fixed as 0.02 Joule, and the total number of rounds was 800. Once after the implementation, the performance evaluation of the developed D-DHOA model was compared over conventional PSO [29], GWO [26], WOA [25], DA [11], and DHOA [3] models. The convergence analysis was provided for ensuring the impact of the implemented model over existing models. Moreover, parameter analysis in terms of normalized energy, latency, and length of the shortest path was performed with respect to the count of nodes. Along with this, latency evaluation, shortest path evaluation, and normalized energy evaluation were also analyzed for the developed model with a prevented black hole and selective forwarding attacks of IoT-WSN in the healthcare sector.

6.2. Simulation results

The simulation results of attack detection and prevention in IoT-WSN is shown in Fig. 7. Here, Fig. 7(a) shows the sensor deployment with k-path selection. If no attack is present, the network model will be based on Fig. 7(b), and the shortest route path without any attack is shown in Fig. 7(c). Moreover, network with the attack, which is node 6 is detected as per Fig. 7(d), and the network with the shortest route path after removing the attack is shown in Fig. 7(e).

Fig. 7.

Simulation resuts for attack detection and prevention in medical-based IoT-WSN system (a) k-path generation (b) no attack (c) shorest path selection for no attack (d) network with attack (e) shorest path selection after removing attac.

6.3. Convergence analysis

The convergence analysis of the introduced D-DHOA-based IoT-WSN model was contrasted with conventional algorithms as given in Fig. 8. This analysis shows the effect of finding the shortest path after detecting and preventing the black hole and selective forwarding attacks on data communication in IoT-WSN. Figure 8 shows the cost functions of the proposed and the state-of-the-art models for various iterations. At 1st iteration, DHOA is minimum, next to that DA, followed by WOA, PSO, and D-DHOA model. In the 5th iteration, the proposed D-DHOA system is having the minimum cost function. DHOA is occupying the second position. Later, DA and WOA is having the minimum cost function. Based on the degree of improvement, at 10th iteration, the developed D-DHOA is 50% improved than DHOA, and 66.6% improved than WOA. Thus, the overall convergence analysis of determining the shortest path without any black hole and selective forwarding attacks in IoT-WSN using the proposed D-DHOA algorithm outperforms the conventional algorithms.

Fig. 8.

Convergence analysis of proposed and conventional IoT-WSN model.

Fig. 9.

Parameters analysis in IoT-WSN with respect to number of nodes (a) normalized energy (b) latency, and (c) length of shortest path.

6.4. Parameter analysis with respect to number of nodes

The parameter analysis of the developed D-DHOA is evaluated based on the number of nodes regarding the normalized energy, latency, and the length of the shortest path is shown in Fig. 9. Figure 9(a) depicts the performance of normalized energy for the number of nodes. For 35 nodes, the proposed D-DHOA is having more energy when compared to all the other algorithms. Next to that, DA is having more energy followed by PSO, WOA, GWO, and DHOA. For 50 nodes, the implemented D-DHOA is still having more energy. After that, GWO is having more energy followed by PSO. Next, DA, WOA, and DHOA are having the least energy. In the 65th node, the proposed D-DHOA model tops all the other models. Here, the DHOA model is having more energy next to the proposed one followed by GWO, WOA, and PSO. Moreover, DA is having the least energy. Based on improvement, the proposed model is having more energy continuously from the 30th node when contrasted over other models. The D-DHOA is 1% better than DHOA, 5.8% better than WOA, 6.9% better than PSO, 8.1% better than GWO, and 13% better than DA. Therefore, it is concluded that the developed model is having more energy compared to conventional models. On evaluating the latency performance with respect to the number of nodes from Fig. 9(b), the proposed D-DHOA is 16.6% superior to WOA and DHOA, 30% superior to DA, 41.6% superior to PSO, and 50% superior to GWO at 35th node. Moreover, the length of the shortest path of the developed and the state-of-the-art algorithms is shown in Fig. 9(c). At 50th node, the proposed D-DHOA is 40% enhanced than WOA, 46.4% enhanced than DA, 53.1% enhanced than GWO, 60.5% enhanced than PSO, and 69.3% enhanced than DHOA. Thus, it is confirmed that the proposed D-DHOA model is having less latency and shortest path over existing models. Finally, from all the parameters it is proved that the developed D-DHOA algorithm is superior to conventional algorithms.

At 700th round, the length of the shortest path of the implemented D-DHOA model is 83.8% superior to DHOA, 85.7% superior to WOA, 33.3% superior to GWO, and 37.5% superior to PSO is shown in Fig. 10(c). Moreover, the length of the shortest path of the developed D-DHOA model at 500th round is 60% enhanced than DHOA, 58.3% enhanced than WOA, 33.3% enhanced than GWO, and 80% enhanced than PSO is given in Fig. 10(d). Finally, from the above calculations, it is confirmed that the presented D-DHOA model is having the minimum shortest path over the state-of-the-art models.

Fig. 10.

Analysis on shortest path length in IoT-WSN for varied network configurations for (a) number of cluster heads = 35 (b) number of cluster heads = 50 (c) number of cluster heads = 65 and number of cluster heads = 80.

6.5. Latency evaluation

With respect to the number of rounds, the evaluation of latency is given in Fig. 11 for the different number of cluster heads. The latency of the proposed D-DHOA is 62.5% improved than DA, 40% improved than WOA and PSO, and 50% improved than GWO that is given in Fig. 11(a) for 35 number of cluster heads. From Fig. 11(b), the latency at 800th round by the developed D-DHOA is 53.3% better than DHOA, 56.2% better than DA, 12.5% better than WOA, and 65% better than GWO for 50 number of cluster heads. The latency of the implemented D-DHOA at 700th round is 75% enhanced than DHOA, 14.2% enhanced than DA, 76.9% enhanced than WOA, and 25% enhanced than GWO, which is given in Fig. 11(c). In Fig. 11(d) for 80 number of cluster heads, at 100th round, the latency of the offered D-DHOA is 46.6% superior to DHOA, 38.4% superior to DA, 36.5% superior to WOA, 50% superior to GWO, and 55.5% superior to PSO. Thus, it is proved that the improved D-DHOA is performing well in detecting and preventing black hole and selective forwarding attacks in IoT-WSN, and continues data communication with less latency.

Fig. 11.

Analysis on latency in IoT-WSN for varied network configurations for (a) number of cluster heads = 35 (b) number of cluster heads = 50 (c) number of cluster heads = 65 and number of cluster heads = 80.

6.6. Normalized energy evaluation

Figure 8 shows the normalized energy analysis of the proposed and the conventional algorithms for the count of the rounds. From Fig. 12(a), at 800th round the normalized energy of the introduced D-DHOA model is 2.9% better than DA, 6% better than PSO, 10.7% better than WOA, 12.9% better than GWO, and 25% better than DHOA. The normalized energy at 600th round, the proposed D-DHOA is 1.7%, 2.9%, 4.1%, 6%, and 6.7% improved than GWO, PSO, DA, WOA, and DHOA is given in Fig. 12(b). At 800th round, the normalized energy from Fig. 12(c) of the developed D-DHOA is 0.6% superior to GWO, 1.2% superior to GWO, WOA, and PSO, and 11.2% superior to DA. In Fig. 12(d), the normalized energy of the implemented D-DHOA is 0.4% enhanced than DHOA, 3.9% enhanced than WOA, PSO, and GWO, and 5% enhanced than DA. Thus, it is confirmed that the proposed method is having maximum energy when compared over the other algorithms in transferring the data on the IoT-WSN network.

Fig. 12.

Analysis on normalized energy in IoT-WSN for varied network configurations for (a) number of cluster heads = 35 (b) number of cluster heads = 50 (c) number of cluster heads = 65 and number of cluster heads = 80.

7. Conclusion

The aim of the paper was to introduce a security approach for detecting and preventing the black hole and selective forwarding attacks in medical IoT-WSN. Here, the developed security mechanism consisted of five stages: a selection of cluster heads, k-routing paths generation, security against black hole attack, security against selective forwarding attack, and finally minimum shortest path selection. At first, for determining the cluster heads, a topology was introduced, and further continued with choosing the k-routing paths. Later, detection and prevention of black hole attacks was done using bait procedure. To detect selective forwarding attacks, packet evaluation was performed by verifying the forwarded packet and the received packet. To promote packet security, the ECC-based hashing function was also established. The key role of the paper was to find the optimal shortest path using the proposed D-DHOA by considering the objective functions like trust, distance, delay or latency, and packet loss ratio. The convergence results of the proposed D-DHOA were 50% improved than DHOA, and 66.6% improved than WOA. The limitation includes that it is difficult to define initial parameters, difficult to solve problems, it can lead to abusive practices, and cost-prohibitive. Thus, it is confirmed that the proposed D-DHOA model was efficient for promoting the shortest path of communication after detecting and preventing the black hole and selective forwarding attacks in medical IoT-WSN.

Conflict of interest

None to report.

References

A.M.

Anita and

Vasudevan, Prevention of Black hole attack in multicast routing protocols for mobile Ad-hoc networks using a self-organized public key infrastructure, Information Security Journal: A Global Perspective 18(5) (2009), 248–256.

Arunmozhi and

Venkataramani, Black hole attack detection and performance improvement in mobile Ad-hoc network, Information Security Journal: A Global Perspective 21(3) (2012), 150–158.

Brammya,

Praveena,

N.S.N.

Preetha,

Ramya,

B.R.

Rajakumar and

Binu, Deer Hunting Optimization Algorithm: A New Nature-Inspired Meta-Heuristic Paradigm, the Computer Journal, 2019.

Chang ,

Tsou,

Woungang,

Chao and

Lai, Defending against collaborative attacks by malicious nodes in MANETs: A cooperative bait detection approach, IEEE Systems Journal 9(1) (2015), 65–75. doi:10.1109/JSYST.2013.2296197.

Chhabra,

Vashishth and

D.K.

Sharma, A game theory based secure model against black hole attacks in opportunistic networks, in: 2017 51st Annual Conference on Information Sciences and Systems (CISS), Baltimore, MD, 2017, pp. 1–6.

Y.-T.

Chuang and

Y.-F.

Lee, Defense mechanism for malicious and selective forwarding attacks in large and mobile wireless networks, The Computer Journal 61(12) (2018), 1862–1879. doi:10.1093/comjnl/bxy095.

Deepak,

Mehetre,

Roslin and

S.J.

Wagh, Detection and prevention of black hole and selective forwarding attack in clustered WSN with active trust, Cluster Computing 22 (2019), 1313–1328. doi:10.1007/s10586-017-1622-9.

Delkesh and

M.A.J.

Jamali, EAODV: Detection and removal of multiple black hole attacks through sending forged packets in MANETs, Journal of Ambient Intelligence and Humanized Computing 10(5) (2019), 1897–1914. doi:10.1007/s12652-018-0782-7.

Hammamouche,

Omar,

Djebari and

Tari, Lightweight reputation-based approach against simple and cooperative black-hole attacks for MANET, Journal of Information Security and Applications 43 (2018), 12–20. doi:10.1016/j.jisa.2018.10.004.

10.

Hao,

Liao,

Adhikari,

Sakurai and

Yokoo, A repeated game approach for analyzing the collusion on selective forwarding in multihop wireless networks, Computer Communications 35(17) (2012), 2125–2137. doi:10.1016/j.comcom.2012.07.006.

11.

Jafari and

M.H.B.

Chaleshtari, Using dragonfly algorithm for optimization of orthotropic infinite plates with a quasi-triangular cut-out, European Journal of Mechanics A/Solids 66 (2017), 1–14. doi:10.1016/j.euromechsol.2017.06.003.

12.

Kaurav and

K.A.

Kumar, Detection and prevention of blackhole attack in wireless sensor network using Ns-2.35 simulator, International Journal of Scientific Research in Computer Science, Engineering and Information Technology 2(3) (2017).

13.

Y.M.

Khamayseh,

S.A.

Aljawarneh and

A.E.

Asaad, Ensuring survivability against Black Hole Attacks in MANETS for preserving energy efficiency, Sustainable Computing: Informatics and Systems 18 (2018), 90–100.

14.

W.Z.

Khan,

M.Y.

Xiang,

Arshad and

M.Y.

Aalsalem, Comprehensive study of selective forwarding attack in wireless sensor networks, International Journal of Computer Network and Information Security 1 (2011), 1–10.

15.

M.S.

Khoirom,

D.S.

Laiphrakpam and

Themrichon, Cryptanalysis of multimedia encryption using elliptic curve cryptography, Optik. 168 (2018), 370–375. doi:10.1016/j.ijleo.2018.04.068.

16.

Kollati and

Somasundaram, IBFWA: Integrated bloom filter in watchdog algorithm for hybrid black hole attack detection in MANET, Information Security Journal: A Global Perspective 26(1) (2017), 49–60.

17.

Kumar and

Sinha, Design and development of new framework for detection and mitigation of wormhole and black hole attacks in VANET, Journal of Statistics and Management Systems 22(4) (2019), 753–776. doi:10.1080/09720510.2019.1609555.

18.

Kumar,

Jain,

U.S.

Bisht,

Jain and

Gupta, Evaluation of black hole attack with avoidance scheme using AODV protocol in VANET, Journal of Discrete Mathematical Sciences and Cryptography 22(2) (2019), 277–291. doi:10.1080/09720529.2019.1585635.

19.

Latha and

Kuppusamy, Secure routing through trusted nodes in wirels sensor networks – a survey, International Journal of Advanced Research in Computer Engineering and Technology 3 (2015).

20.

Liu,

Dong,

Ota and

Long, PHACK: An efficient scheme for selective forwarding attack detection in WSNs, Sensors (Basel) 15(12) (2015), 30942–30963. doi:10.3390/s151229835.

21.

Liu,

Dong,

Ota and

Liu, Active trust: Secure and trustable routing in wireless sensor networks, IEEE Trans. Inf. Forensics Secur 11(9) (2016), 2013–2027. doi:10.1109/TIFS.2016.2570740.

22.

Manoranjini,

Chandrasekar and

Jothi, Improved QoS and avoidance of black hole attacks in MANET using trust detection framework, Journal for Control, Measurement, Electronics, Computing and Communications 60(3) (2019), 274–284.

23.

Mathur ,

Newe and

Rao, Defence against black hole and selective forwarding attacks for medical WSNs in the IoT, Sensors (Basel) 16(1) (2016).

24.

Mehetre,

S.E.

Roslin and

S.J.

Wagh, Detection and prevention of black hole and selective forwarding attack in clustered WSN with active trust, Cluster Computing 22 (2019), 1313–1328. doi:10.1007/s10586-017-1622-9.

25.

Mirjalili and

Lewis, The whale optimization algorithm, Advances in Engineering Software 95 (2016), 51–67. doi:10.1016/j.advengsoft.2016.01.008.

26.

Mirjalili,

S.M.

Mirjalili,

Lewis and

G.W.

Optimizer, Advances in Engineering Software 69 (2014), 46–61. doi:10.1016/j.advengsoft.2013.12.007.

27.

Moudni,

Er-rouidi,

Mouncif and

ElHadadi, Black hole attack detection using fuzzy based intrusion detection systems in MANET, Procedia Computer Science 151 (2019), 1176–1181. doi:10.1016/j.procs.2019.04.168.

28.

Otoum,

Kantarci and

H.T.

Mouftah, Hierarchical trust-based black-hole detection in WSN-based smart grid monitoring, in: 2017 IEEE International Conference on Communications (ICC), Paris, 2017, pp. 1–6.

29.

M.E.H.

Pedersen and

A.J.

Chipperfield, Simplifying particle swarm optimization, Applied Soft Computing 10(2) (2010), 618–628. doi:10.1016/j.asoc.2009.08.029.

30.

Ren,

Zhang,

Zhang and

Shen, Adaptive and channel-aware detection of selective forwarding attacks in wireless sensor networks, IEEE Transactions on Wireless Communications 15(5) (2016), 3718–3731. doi:10.1109/TWC.2016.2526601.

31.

Shashwat,

Kumar,

Pandey and

K.V.

Arya, A modified AODV protocol for preventing blackhole attack in MANETs, Information Security Journal: A Global Perspective 26(5) (2017), 240–248.

32.

J.C.

Sunder and

Shanmugam, Jensen–Shannon divergence based independent component analysis to detect and prevent black hole attacks in healthcare WSN, Wireless Personal Communications 107(4) (2019), 1607–1623. doi:10.1007/s11277-019-06347-5.

33.

Tian,

Y.-X.

Yang,

Li,

Li and

Xin, A security framework for wireless sensor networks, The Journal of China Universities of Posts and Telecommunications 17(2) (2010), 118–122. doi:10.1016/S1005-8885(09)60571-9.

34.

Xiao,

Yu and

Gao, CHEMAS: Identify suspect nodes in selective forwarding attacks, Journal of Parallel and Distributed Computing 67(11) (2007), 1218–1230. doi:10.1016/j.jpdc.2007.04.014.

35.

Zhu,

Zhang,

Du,

Luo and

Xin, Detection of selective forwarding attacks based on adaptive learning automata and communication quality in wireless sensor networks, International Journal of Distributed Sensor Networks 14(11) (2018). doi:10.1177/1550147718815046.