An effective attack detection framework using multi-scale depth-wise separable 1DCNN via fused grasshopper-based lemur optimizer in IoT routing system

Abstract

A secured IoT routing model against different attacks has been implemented to detect attacks like replay attacks, version attacks, and rank attacks. These attacks cause certain issues like energy depletion, minimized packet delivery, and loop creation. By mitigating these issues, an advanced attack detection approach for secured IoT routing techniques with a deep structured scheme is promoted to attain an efficient attack detection rate over the routing network. In the starting stage, the aggregation of data is done with the help of IoT networks. Then, the selected weighted features are subjected to the Multiscale Depthwise Separable 1-Dimensional Convolutional Neural Networks (MDS-1DCNN) approach for attack detection, in which the parameters in the 1-DCNN are tuned with the aid of Fused Grasshopper-aided Lemur Optimization Algorithm (FG-LOA). The parameter optimization of the FG-LOA algorithm is used to enlarge the efficacy of the approach. Especially, the MDS-1DCNN model is used to detect different attacks in the detection phase. The attack nodes are mitigated during the routing process using the developed FG-LOA by formulating the fitness function based on certain variables such as shortest distance, energy, path loss and delay, and so on in the routing process. Finally, the performances are examined through the comparison with different traditional methods. From the validation of outcomes, the accuracy value of the developed attack detection model is 96.87%, which seems to be better than other comparative techniques. Also, the delay analysis of the routing model based on FG-LOA is 17.3%, 12.24%, 10.41%, and 15.68% more efficient than the classical techniques like DHOA, HBA, GOA, and LOA, respectively. Hence, the effectualness of the offered approach is more enriched than the baseline approaches and also it has mitigated diverse attacks using secured IoT routing and different attack models.

Keywords

Attack detection secured Iot routing fused grasshopper-based lemur optimization algorithm Multiscale Depthwise Separable 1-Dimensional Convolutional Neural Networks multi-objective constraints

1. Introduction

IoT is the situation where the sensor, things, and objects are offered with network connectivity and computing. So, the objects are easily created and convey the data without human interaction [1]. Moreover, IoT is widely utilized by many applications to monitor farming, the environment, industry, tracking and shipment, smart city, and so on [2]. Thus, the need for IoT is increasing rapidly in human life. In the real-world scenario, highly secure communication is essential for effective installation. It is more complicated to provide secured communication with heterogeneous devices [3]. Most communication devices are more powerful than resource constraints. Furthermore, users required effective communication among the devices with improved security rates [4]. Professionals sometimes work efficiently to design lightweight protocols and techniques for effectual communication in the constrained node on the internet [5]. IoT is the smart object network linked with various nodes and the internet. Based on IoT, smart electric devices gained an effectual progress rate [6]. But, privacy and security-based techniques efficiently utilize IoT techniques. Data exchange is more essential in IoT networks, at the same time, they are highly affected due to breaching attacks. Health applications are called sensitive applications, which need to handle the data related to the service provider and the end-user. Various disruptive attacks take place due to minimal security solutions [7]. Furthermore, IoT needs to fulfill the requirements based on security, integrity, confidentiality, and availability for efficient adaption. The enhancement in IoT secured better improvements in different smart solutions. This leads to the implementation of a wide range of different smart solutions and multiple IoT-based linked objects [8].

Generally, IoT is fused with different small-sized smart devices with minimal resources like energy, storage, and validation [9]. The traditional connection setup inter-connects the components according to energy-efficiency-based wireless communication. Utilizing Low-power and Lossy Networks (LLNs) offers a better performance rate to design an efficient networking solution with an IoT structure [10]. The essential characteristic of LLN effectively supports the efficacy of energy and minimal power communication over the simple wireless structure [11]. LLN requires a minimal-cost IoT network that deploys interconnection over multiple constrained IoT devices. The network routing performed between the IoT devices in the LLN layer is subjected to efficient data communication in IoT [12]. So, an effectual network routing solution for IoT is essential to improve the important features of LLN.

Different complications presented in the IoT devices are resolved using RPL to minimize the routing attack in the system and enhance the efficacy of routing in RLP [13]. Moreover, they are not developed in real-world applications and approaches. Later, more research works are implemented for detecting different kinds of attacks over the RPL model [14]. Deep structured methods like Recurrent Neural Networks (RNNs), as well as Long-Short Term Memory Networks (LSTMs), are utilized to study the patterns presented in the long sections. LSTM effectively performs pattern analysis because they are improved from RNN [15]. It is feasible to utilize the LSTM to study the characteristics and pattern of the network data, and they are employed to categorize whether the attack is normal or not [16]. Deep learning techniques didn’t need to perform more procedures as they work efficiently with the features, and also, the traditional machine learning approaches are widely utilized with the raw data. Different research works have been designed for the RPL to enhance the cryptographic defense attack and security features over the rank attack [17]. Yet, they resulted in validation overhead and utilized more energy, which is against over RPL policy. Thus, the basic aim of the recommended technique is to enhance the RPL technique to identify and isolate the rank attackers. Among all those implemented techniques, this paper provides a new concept for better performance for Secured IoT routing against different attacks.

Multiple contributions linked with the initiated attack detection with malicious node mitigation framework for secured IoT routing are explained below.

•
To implement an innovative attack detection model for secured IoT routing technique with deep learning strategy to obtain effective attack detection rate over the routing network to offer better attack detection rate.
•
To obtain the efficient weighted features, the weights presented in the selected features are tuned by recommended FG-LOA to improve the efficacy of attack recognition.
•
To construct the novelty in attack detection technique in secured IoT framework named MDS-1DCNN to resolve the complications by modifying the parameters of 1DCNN by initiated FG-LOA to enhance the rate of attack identification by maximizing the accuracy.
•
To mitigate the malicious node in the IoT network to minimize constraints like energy consumption, path loss, delay, and shortest distance by developed FG-LOA.
•
To design a novel heuristic technique named FG-LOA to alter the constraints of 1DCNN like learning rate, epochs, and the number of suitable hidden neuron counts and improve the weights and essential features of the recommended system.
•
To validate the efficacy of the constructed MDS-1DCNN-based attack detection framework over multiple observations with conventional heuristic technique and attack detection strategy.

The leftover parts of the implemented attack detection framework for secured IoT routing systems are elaborated as follows. Multiple literature works connected with the conventional attack detection framework are detailed in Section 2. Secured IoT routing over different attacks is explained with the deep learning strategy provided in Section 3. Multi-attack detection frameworks with the developed MDS-1DCNN-based attack detection are detailed in Section 4. Attack detection with several multi-objective constraints is discussed in Section 5. Different experimental observations performed over the initiated framework over the conventional technique are provided in Part VI. Finally, the conclusion parts of the offered framework are provided in part VII.
2. Literature survey

2.1 Related works

In 2022, Nandhini et al. [18] proposed an Enhanced-Rank Attack Detection (E-RAD) framework to minimize the production of DODAG Information Object (DIO), and also they utilized the message solicitation model DODAG Information Solicitation (DIO) for isolating the rank attackers. The attackers can be identified using the Destination Advertisement Object (DAO) message. Moreover, alarm models were linked with the control message but not the packet to identify the attackers. Finally, the designed model attained an enriched performance rate regarding accuracy and minimal overhead issues in constructing the control packet.

In 2022, Muzammal et al. [19] proposed a protocol based on Secured Mobility Trust named SMTrust by validating the trust measures associated with mobility-based measures in IoT. The major usage of SMTrust was to offer better security over the blackhole and RPL rank attacks. Moreover, the recommended protocol was validated over three various scenarios like mobile nodes, static, and dynamic in the IoT network. Later, the developed SMTrust was contrasted over different objective functions and ranking schemes regarding throughput, packet loss rate, and topology stability.

In 2020, Alsukayti and Singh [20] introduced a lightweight as well as efficient recognition and fusion resolution over RPL VN attack. Here, minimal changes were performed in the functionality of RPL. Next, a security technique named the collaborative and distributive model was fused with the protocol structure. The implemented protocol offered an accurate and efficient recognition rate with a better convergence rate on the attempted attack. They efficiently maintained traffic overhead, consumption of energy, efficacy of QoS, and stability rate of the network in various VN attacks. The recommended framework secured better efficacy outcomes when contrasted with the existing techniques with minimal communication overhead, no additional entities, and simple local node processing.

In 2022, AliSeyfollahi et al. [21] proposed an optimization technique named Moth-Flame Optimization-based secure scheme for RPL (MFO-RPL) to modify the routing procedure and the attack identification in RPL. Moreover, MFO-RPL utilized the petal approaches to choose the next hop nodes and design an optimal way among the root and the source in a graph. Further, the rank attacks presented in RLP were identified by MFO to secure from the malicious nodes chosen by the parents. Experimental analysis performed under various analyses displayed that the suggested framework lost minimal packets and used decreased convergence time and attack than the classical techniques.

In 2020, Sahay et al. [22] recommended a layered framework to perform efficient security routing for observing the attacks related to every stage of the routing procedure. They effectively explored inherent characteristics presented in blockchain to increase the routing security in IoT LLNs. To conclude this developed framework, a novel blockchain-based framework, and smart contracts were designed to produce real-world alerts for improving the efficacy of the sensor node included in the LLN tampering.

In 2022, Ragesh and Kumar [23] has proposed securing data for IoT-based communications. At first, Fuzzy and Particle Swarm Optimization (F-PSO) was employed to define a secured path. Later, cryptographic models were utilized with a learning-based encryption framework for effective data encryption. Here, the trust values of the entire node presented in the network were validated by trustworthiness observation. Based on true value, the malicious nodes were collected then the features of secure nodes were used to identify highly secured paths among the destination and source by F-PSO. Once the secured paths were identified, the essential data were encrypted effectively. Thus, the suggested structure secured a better throughput rate with a minimal delay than the classical techniques.

In 2022, Janani and Ramamoorthy [24] presented a deep structure architecture according to LSTM and Adaptive Mayfly Optimization Algorithm, and this combination is termed (LAMOA) to perform efficient categorization in IoT attacks. The adaptive optimization model altered the weights over different layers in LSTM and fully connected layers for categorizing. In different examination procedures, the offered approach achieved an enriched rate than the standard techniques.

In 2020, Lalit Kumar and Pradeep Kumar [25] addressed the energy problems presented in Wireless Sensor Networks (WSN) to attain better routing rates. The main stages of the implemented framework were transmission and clustering, and also the energy-efficient cluster head is performed according to various criteria like distance, packet delivery ratio, throughput, security, and residual energy. Later, different classical optimization models were used to employ efficient cluster head selection in the multi-objective function. Finally, the recommended model attained a better routing energy efficacy rate.

2.2 Summary

The major advantages and limitations of secured IoT routing against different attacks are listed in Table 1. The e-RAD [18] model has effectively detected the newly joined attackers in the networking model. It consumes lower energy for performing the overall performance in the networking model. Incorporating node mobility as well as the detection of other IoT attackers needs to be developed. SMTrust [19] technique has been adopted by most IoT applications and is prone to various attacks. It has also provided better outcomes for throughput, topology stability as well as packet loss rate. There is a requirement for a power consumption system. CDRPL [20] method has incorporated the distribution and collaborative model against the VN attacks. It also can provide fast and accurate detection of composite and simple VN attacks. The effective mitigation of hybrid attacks is lacking in this model. The MFO-RPL [21] model has a better balance between the exploitation and the exploration phase, which improves the performance of this model. It has also provided a better convergence rate while performing the cluster head function. But, this technique has failed to protect the routing information. IoT-LLNs [22] model has assured the most secure and effective routing process. It has included the alert signal in this model, which has improved the performance of this model. The growing size of the blockchain and the volume of the generated data have a major impact on this model. F-PSO [23] model can potentially optimize the parameters of several learning models and is also used to resolve various issues. It shows better computational efficiency when assimilated over other models. It has acquired a low convergence rate. The ensemble [24] technique has been utilized to get rid of the characteristics that are redundant and irrelevant features. It has shown better outcomes over the performance metrics when compared to other models. There is a lack of a better understanding of dataset and pattern classification and predictions. The ensemble [25] method has provided secure-energy-aware multi-hop routing that has effectively increased the model’s performance. But, it is more expensive in terms of both time and space.

Table 1
Uppercomings and Lowercomings of secured IoT routing against different attacks model

Author [citation]	Framework	Upper comings	Lower comings
Nandhini et al. [18]	E-RAD	This approach has effectively detected the newly joined attackers in the networking model. It consumes lower energy for performing the overall performance in the networking model.	Incorporating node mobility as well as the detection of other IoT attackers needs to be developed.
Muzammal et al. [19]	SMTrust	This technique has been regarded as accepted by most IoT applications and is prone to various attacks. It has also provided better outcomes for throughput, topology stability as well as packet loss rate.	There is a requirement for a power consumption system.
Alsukayti and Singh [20]	CDRPL	This method has incorporated the distribution and collaborative model against the VN attacks. It also can provide fast and accurate detection of composite and simple VN attacks.	The effective mitigation of hybrid attacks is lacking in this model.
AliSeyfollahi et al. [21]	MFO-RPL	This model has a better balance between the exploitation and the exploration phase, which improves the performance of this model. It has also provided a better convergence rate while performing the cluster head function.	But, this technique has failed to protect the routing information.
Sahay et al. [22]	IoT-LLNs	This model has ensured the most secure and effective routing process. It has included the alert signal in this model, which has improved the performance of this model.	The growing size of the blockchain as well as the volume of the generated data, has a major impact on this model.
Ragesh and Ajay Kumar [23]	F-PSO	This model can potentially optimize the parameters of several learning models and is also used to resolve various issues. It shows better computational efficiency when assimilated over other models.	It has acquired a low convergence rate.
Janani and Ramamoorthy [24]	Ensemble	This technique has been utilized to get rid of the features that are redundant and irrelevant features. It has shown better outcomes over the performance metrics when compared to other models.	There is a lack of a better understanding of dataset and pattern classification and predictions.
Lalit Kumar and Pradeep Kumar [25]	Ensemble	This method has provided secure-energy-aware multi-hop routing that has effectively increased the model’s performance.	But, it is more expensive in terms of both time and space.

3. Proposed methodology

3.1 Proposed energy efficient attack detection-based secured IoT routing model

IoT is widely employed to fuse different devices, which are operated and communicate continuously to allow devices, peoples, and services to link and interchange the data in huge applications utilized in industrial and healthcare environments. Such devices are termed potential vectors for the attack from the malicious node. Moreover, malware and botnet-related attacks create more threats to the protection of the device and the hosting network, like IoT devices. Based on these attacks, more demand is generated to recognize and classify the compromised device in time to integrate the risk with the network and IoT devices. According to this previous outcome, more need is generated to perform dynamic detection as well as recognition in the compromised devices on time to resolve the mitigation risk in the network and also a successive time limit in the IoT devices. The key elements are more needed for recent and upcoming security in sensor-cloud systems. So, there is a need to analyze the efficient solution for countering the attacks. Moreover, an Intrusion Detection System (IDS) became an essential technique to resolve the complications presented in network security. Further, these techniques are employed to monitor the traffic on the network and also to detect dangerous activities. Generally, the IDS are categorized as anomaly-based and signature-based defense methodologies. The signature-aided IDS effectively identifies the instructions and protects the network from well-known attacks. Next, the anomaly-aided IDS monitoring network effectively monitors and compares the traffic over the previously obtained patterns to detect malicious activities. Finally, the anomaly recognition model offered a better recognition rate in detecting the new attacks, at the same time; it requires some recurrent updating to identify the new attack. So, several complications presented in the standard energy-efficient attack detection models must be determined. Hence, a novel energy-efficient attack detection framework with malicious node mitigation is designed by considering the limitations of the existing techniques, and their pictorial presentation of the developed framework is given in Fig. 1.

Figure 1.

Pictorial view of developed energy efficient attack detection framework with malicious node mitigation.

A novel energy-efficient attack detection framework with malicious node mitigation is designed to recognize different attacks like version attacks, replay attacks, and rank attacks in a secured IoT-based routing model. Initially, fundamental data for the validation are garnered from the IoT networks and forwarded to the data cleaning and transformation region. Then, the data cleaned and transformed are subjected to a weighted feature selection region. Here, the weights of the essential feature are tuned by the offered FG-LOA. Next, the weighted features are input to the attack detection phase. In this phase, the initiated MDS-1DCNN model is utilized to detect several attacks and also their parameters like learning rate, number of suitable hidden neuron count, and epochs, and are modified by developed FG-LOA to maximize the accuracy of attack detection. Later, efficient routing is performed over the selected node by mitigating the malicious node in the IoT network, and also, they efficiently validate and minimize diverse variables like delay, energy consumption, path loss, and shortest distance in the suggested framework. Therefore, the recommended attack detection structure attained an improved attack recognition rate.

3.2 IoT-aided attack detection dataset

Different data utilized for the investigation are gathered from the IoT-based intrusion detection dataset and the hyperlink https://drive.google.com/file/d/1iVLjkQ0G8Mo8Azcc3czeZLcmEWMDDvnb/view?usp= sharing: Access data: 2023-04-18. The dataset consists of the simulation outcome of the cross-layer intrusion detection model with an IoT network. The dataset holds two validation outcome file types. pcap and .log files. Moreover, two networks are utilized for the observation. The dataset is commonly employed to detect the Hello Flood Attack, Version Number Attack, no attack cases, and Worst Parent Attack. For every attack model, five various examinations are performed according to the density. The dataset employed for the observation attack detection in the IoT-based routing model is offered as $Da_{q}^{\textit{inp}}$ , where, $q=1,2,\ldots,Q$ and the acquired attack detection data are given as $Q$ .

3.3 IoT routing model

IoT is termed the interconnection among various validating devices for supporting different applications related to monitoring and controlling. To assist multiple applications and devices from various vendors, the current IoT system accepted open standards with different protocols designed according to the wired global internet. Mostly, IoT networks are differentiated from the classical wired computing network based on their basic methodologies. These kinds of differences create essential complications in the TCP topologies presented in the IoT environment, and resolving these limitations effectively creates great in the structure of a network. Moreover, the IoT network holds more minimal-end resourced devices. The construction of these devices is obtained from minimal operational and manufacturing costs. Furthermore, IoT devices are designed with minimal validation power and must operate for huge periods with a battery. Based on power constraints, the IoT network utilized minimal energy-aided techniques with reduced power transmission rates. The common limitation presented in the IoT network is to accept the packet size with constrained links. IoT nodes save energy with wired networks and utilize the wireless mesh topology for efficient communication. The architectural view of the IoT-based routing model is given in Fig. 2.

Figure 2.

Structural presentation of IoT-based routing model.

4. Different attack detection model in secured IoT routing network using multi-scale depth-wise separable 1DCNN

4.1 Data aggregation

In this phase, essential data like the IoT-based intrusion detection dataset for the experiments are taken from online sources, and the data presented in the dataset are acquired from IoT devices and provided as the input. Here, the collected data utilized for the analysis are termed as $Da_{q}^{\textit{inp}}$ , and they are termed as the input to the data cleaning and transformation phase to ignore the unwanted details given in the data.

4.2 Cleaning of data and transformation

The aggregated data $Da_{q}^{\textit{inp}}$ for the analysis are taken as the input to the cleaning of data and transformation phase. Data cleaning is the procedure of fixing or eliminating the corrupted, duplicate, incorrectly formatted, or incomplete data presented in the dataset. When fusing more data resources, more mislabeled and duplicate data is generated. In case the data presented in the dataset are incorrect, results and approaches are termed as unreliable, and they are looked at as accurately correct. No accurate solutions are presented to acquire the cleaned data from the dataset, and also it may change according to dataset type. Moreover, the data cleaning procedure efficiently eliminates the data not associated with the dataset. Furthermore, data transmission is converting the data from one format to another. However, the data transformation procedures are termed data munging or data wrangling, efficiently transforming the raw data into another format. The attained cleaned and transformed data $Da_{q}^{\textit{dft}}$ are further utilized as the input in the weighted feature selection phase.

4.3 Weighted feature selection with heuristic tool

The cleaned and transformed data $Da_{q}^{\textit{dft}}$ are utilized as the input for efficient feature selection with a heuristic approach. From the cleaned and transformed data, five features are obtained in the range of $[{1,5}]$ , and also the five weights are selected in the bound of $[{0.01,0.90}]$ using FG-LOA. The equation for weighted feature selection is offered in Eq. (1).

$\displaystyle FE_{g}^{\textit{wfs}}=SF_{t}^{op}*Wi_{b}^{op}$ (1)

Here, the term $SF_{t}^{op}$ indicates the five different selected features, and the weights are presented as $Wi_{b}^{op}$ . The weighted selected features are termed as $FE_{g}^{\textit{wfs}}$ , and they are used as the input to the MDS-1DCNN-based attack detection phase. The architectural representation of the selection of weighted features is offered in Fig. 3.

Figure 3.

Pictorial presentation of a weighted selection of features using developed FG-LOA.

4.4 Developed MDS-1DCNN for attack detection

The weighted features $FE_{g}^{\textit{wfs}}$ are input to the MDS-1DCNN-based attack detection framework. Its structural representation is simple, which performs efficient monitoring and fault recognition in the real-world system. Its training procedure is simple, and also the computational cost is minimal. But, it rapidly minimizes the operation rate and also processes large datasets only. Thus, different complexities available in the 1DCNN model are overcome by modifying the learning rate, epochs, and the suitable number of hidden neuron counts in different ranges by FG-LOA to enlarge the accuracy of attack detection.

In the input layer of 1-DCNN [26], weighted features are used as the input. Essential features for the analysis are effectively acquired from the convolution layer and pooling layer. Moreover, efficient processing is performed in the form of layer by layer. Here, the convolutional layer consists of convolution kernels of equal sizes. Average pooling is achieved by the pooling layer, and the fully linked layer is employed to classify the outcome efficiently.

Convolution layer: In the local input region, a convolutional operation generates the appropriate one-dimensional characteristics maps. Similarly, multiple kernels are used to extract several characteristics from the input. In this phase, entire convolution kernels recognize the appropriate characteristics in the whole location from the input feature map. Furthermore, weight sharing and local connectivity are employed to reduce the limitation obtained in the network and also in counts of training parameters. The convolution layer of 1D-CNN is offered in Eq. (2).

$\displaystyle p_{h}^{g}=I\left({\sum\limits_{k=1}^{u}{t_{k}^{g-1}\ast r_{kh}^{% g}+e_{h}^{g}}}\right)$ (2)

Figure 4.

Implemented MDS-1DCNN-based attack detection framework.

The activation function of the convolution layer is offered as $I(\,)$ , the operator of a convolution is presented as $\ast$ , bias achieved in the kernel is termed as $e$ , number kernels are provided as $h$ , channel count available in the input is given as $u$ , and the convolutional kernels are represented as $r$ .

Pooling layer: The feature map counts as well as the dimensionality rate is presented in the data are improved efficiency in the down-sampling procedure, and also it led to complicated evaluations. However, the maximal pooling window process selects the increased parameter in a particular range.

In the developed MDS-1DCNN, the parameters of 1-DCNN are tuned in various ranges, where the epochs are modified in the limit of $[{50,100}]$ , hidden neurons are tuned in the range of $[{5,255}]$ , and the learning rates are tuned in the bound of $[{0.01,0.99}]$ by FG-LOA. The objective function of the MDS-1DCNN-based attack detection framework in a secured IoT system is offered in Eq. (3).

$\displaystyle ft_{1}=\mathop{\arg\min}\limits_{\left\{{Ep_{b}^{1\textit{dccn}}% ,Hn_{m}^{1\textit{dcnn}},Lr_{j}^{1\textit{dcnn}}}\right\}}\left({\frac{1}{AY}}\right)$ (3)

In the above equation, the learning rate of 1-DCNN is presented as $Lr_{j}^{1\textit{dcnn}}$ , epochs are referred to $Ep_{b}^{1\textit{dccn}}$ , and the hidden neurons are termed as $Hn_{m}^{1\textit{dcnn}}$ . Accuracy $A Y$ is the examination of nearness in the isolated gain, offered in Eq. (4).

$\displaystyle Ac=\frac{({gq+gw})}{({gq+gw+ge+gr})}$ (4)

Here, the value of false negative is indicated as $g r$ , the value of true positive is offered as $g w$ , the value of false positive is given as $g e$ and the value of true negative is termed as $g q$ . The structural view of the developed MDS-1DCNN-based detection of attack model is presented in Fig. 4.

5. Attack detection-based secured IoT routing model against different attacks with efficient objective function

5.1 Proposed FG-LOA

A novel optimization framework named FG-LOA is designed to improve the attack detection rate in a secured IoT-based routing framework by tuning the weights in features and epochs, the number of suitably hidden neuron counts, and the learning rate in MSD-1DCNN. Here, the GSO technique offers better accuracy and deployment rate, and it effectively resolves unconstrained global issues. But, they easily fall in local optimum issued and have minimal convergence speed. To resolve this complication in GSO, a new optimization model named LOA is utilized for the analysis, and this fused combination is termed FG-LOA. The LOA model easily detects the accurate outcome by performing deep searching, and also it has a better exploration rate. Here, if the choice $Y$ is equal to one, then the final solution update is executed based on the GOA technique else, the final updating is performed according to LOA in the developed framework.

GSO [27]: It is a nature-enthused heuristic model, which showcases a huge boundary and also unpredictable character changes based on the elder grasshopper to realize the exploration procedure. In the exploration region, slow and minimal-step motions are presented in the larva stage to eat the vegetation and are repeated. The characteristic nature of grasshopper is offered in Eq. (5).

$\displaystyle Z_{b}=j_{b}+X_{b}+C_{b}$ (5)

Here, the term $Z_{b}$ indicates the position of $b^{\text{th}}$ the grasshopper, different social attractions such as attraction and repulsion are offered as $j_{b}$ , wind advection of grasshopper is given as $C_{b}$ and the $b^{\text{th}}$ grasshopper gravity force is indicated as $X_{b}$ . Next, the grasshopper’s social interaction is depicted in Eq. (6).

$\displaystyle j_{b}=\sum\limits_{\begin{subarray}{c}w=1\atop w\neq 1% \end{subarray}}^{Q}{J({s_{bw}})}\hat{s}_{bw}$ (6)

Next, the gravity force is presented in Eq. (7).

$\displaystyle X_{b}=-\partial\hat{v}_{\partial}$ (7)

Here, the term $\partial$ denotes the gravitational constant to the center of earth is provided as $\hat{v}_{\partial}$ . Finally, the advection of wind is presented in Eq. (8).

$\displaystyle C_{b}=S\hat{v}_{\theta}$ (8)

Here, the unified vector over the earth center is termed as $\hat{v}_{\theta}$ , and the term $S$ refers to the constant drift. Next, the function of social interaction is referred to $J$ in Eq. (9).

$\displaystyle J(s)=mi^{\frac{-q}{\tau}}-i^{-q}$ (9)

The span between the grasshoppers is referred to as $i$ , the scale of attractive length is presented as $\tau$ and the strength of attraction is offered as $m$ .

The arithmetical presentation to replicate the nature of the grasshopper is offered in Eq. (10).

$\displaystyle Z_{b}=g_{x}\left({\sum\limits_{\begin{subarray}{c}t=1\atop t\neq 1% \end{subarray}}^{N_{\textit{pop}}}{g_{n}\frac{f-c}{2}J({s_{bw}})\hat{s}_{bw}}}% \right)+\hat{G}_{u}$ (10)

Here, the inertia weight is offered as $g_{x}$ which is effectively utilized to stability of the exploration and exploitation stage, $N_{\textit{pop}}$ denote the grasshopper population size, the minimal range is presented as $c$ , and the maximal range is indicated as $f$ in the search space. The term $\hat{G}_{u}$ represents the recent optimal target value. Next, the unit vector values examined among the grasshopper are offered in Eq. (11).

$\displaystyle\hat{s}_{bw}=\frac{h_{w}-h_{b}}{s_{bw}}$ (11)

Here, the $w^{\text{th}}$ grasshopper position is given as $h_{w}$ and the location of $b^{\text{th}}$ the grasshopper is provided as $h_{b}$ . Finally, the internal parameter $g_{x}$ is depicted in Eq. (12).

$\displaystyle g_{x}=g_{n}=g_{mx}-M\frac{g_{mx}-g_{mn}}{G}$ (12)

The internal parameters $g_{x}$ are utilized to reduce the shrink coefficient available in the comfort zone, attraction zone, and repulsion zone. The minimal value is referred to as $g_{mn}$ , the maximal value is presented as $g_{mx}$ , the recent iteration number is provided as $M$ , and the increased iteration count is presented as $G$ .

LOA [28]: Lemurs are categorized as prosimian primates, which are neither apes nor monkeys. Lemurs are commonly available in the Madagascar region, and they are commonly seen in wetlands, spiny forests, mountains, and rainforests. Lemurs are social animals that live as a group referred to as troops, and they utilize scent glands to convey their location.

The lemurs presented in the population matrix are termed as $A$ with candidate solutions $C S$ and decision variables $D V$ . Moreover, the decision variable presented in the $r^{\text{th}}$ solution is offered in Eq. (13).

$\displaystyle H_{r}^{y}=\textit{Rand}(\,)\times({({Np-Mp})+Mp})$ (13) $\displaystyle\forall r\in({1,2,\ldots,j})\Lambda\forall y\in({1,2,\ldots k})$ (14)

The randomly distributed numbers in the different ranges are termed as $\textit{Rand}(\,)$ , and the upper and lower bound is termed as $N p$ and $M p$ . In the direction phase, the decision variable and solution are allocated using different options, and they are presented in Eq. (15).

$\displaystyle H_{r}^{y}=\left\{{{\begin{array}[]{ll}{h({r,y})+ab({h({r,y})-h({% bn,y})})*({\textit{rand}-0.5})*2}&{\textit{Rand}>\textit{frr}}\\ {h({r,y})+ab({h({r,y})-h({gb,y})})*({\textit{rand}-0.5})*2}&{\textit{Rand}<% \textit{frr}}\\ \end{array}}}\right.$ (15)

Here, the present lemur is noted as $h({r,y})$ , the best nearest lemur is offered as $h({bn,y})$ , the global best lemur is termed as $h({gb,y})$ , the free risk rate is the risk rate of entire troops, and it is given as frr and the arbitrary number presented in the range of $[{0,1}]$ is referred as Rand. Finally, the frr probability is termed as the essential coefficient of the LOA approach, and it is offered in Eq. (16).

$\displaystyle\textit{frr}=\textit{frr}(\textit{Hrr})-\textit{Citer}\times\left% ({\frac{({\textit{Hrr}-\textit{Lrr}})}{\textit{MXiter}}}\right)$ (16)

Here, the term Citer indicates the current iteration, Hrr represents the high-risk rate, Lrr indicates the low-risk rate and MXiter presents the maximum iteration rate. The computational complexity of the recommended technique is offered in Eq. (17).

$\displaystyle CC({lo})=CC({\textit{MXiter}\times Y\times U})$ (17)

Here, the variable $Y$ denotes the choice, and the variable $U$ refers to the decision variable and the final updating is performed according to the choice value. The pseudocode of the recommended framework is presented in Algorithm 1.

Algorithm 1: Implemented FG-LOA
Allot the population of GOA and LOA
Allot the constraints of GOA and LOA
For the whole solution
Accomplish the fitness for the entire solution
Renewal of random variable
Recognize the good optimal solution
End for
	If $Y=1$
		improvise the final update by GOA
	Else
		improvise the final update by LOA
	End If

5.2 Malicious node mitigation using FG-LOA

The basic sensor node includes more tiny nodes with it. Here, the nodes are executed unsupervised, which leads to vulnerability. In most of the WSN systems, data collection is executed in a distributed manner or central location, so it is essential to attain control among the node identities presented in the network. Moreover, generating new nodes effectively limits the attackers. Most IoT networks are presented with minimal security. Here, the constrained resource and form factor in the IoT device generates to perform node capture electronically or physically. Later, malicious node improvement is employed by vanishing the data and information of the security key from memory. In the developed model, an essential node for the analysis is selected and finds the shortest distance, consumption of energy, delay, and path loss among the selected nodes. In the selected node, malicious node detection is performed. In case, a malicious node is presented in the selected node, then set the penalty $P E$ as 0.9, or else set the penalty as 0. The fitness of this offered model is to minimize the path loss ratio, shortest distance, energy consumption, penalty, and delay with the initiated FG-LOA technique. The objective function of the suggested technique is offered in Eq. (18).

$\displaystyle FT=\mathop{\arg\min}\limits_{\{{sn}\}}\left({\frac{1}{SC+PL+EG+% DE+PE}}\right)$ (18)

Here, the term $s n$ indicates the selected node, $S C$ denotes the shortest distance, $P L$ indicates the path loss, $E G$ represents the energy consumption, $D E$ presents the delay, and $P E$ refers to the penalty. A systematic representation of the malicious node mitigation model using FG-LOA is offered in Fig. 5.

Figure 5.

Pictorial representation of malicious node mitigation model with FG-LOA.

5.3 Objective constraints description

A description of the different objective constraints is given as follows.

The shortest distance $S C$ is the span between the source and the destination node, offered in Eq. (19).

$\displaystyle SC=\sqrt{\sum\limits_{r=1}^{R}{({Ku_{w}-Kt_{w}})^{2}}}$ (19)

Here, the term $Nd_{1}=({Ku_{1},Ku_{2},\ldots Ku_{w}})$ represents the source nodes, and the term $Nd_{2}=(Kt_{1},\linebreak Kt_{2},\ldots Kt_{w})$ indicates the destination node.

Energy consumption $E G$ is obtained over the average energy available in the live node, termed in Eq. (20).

$\displaystyle EG=EG_{h}-({ge_{h}^{yt}+ge_{h}^{wr}})$ (20)

Here, the energy utilized to obtain the data is given as $ge_{h}^{yt}$ , energy presented in a node $h$ is offered as $EG_{h}$ , and the energy used while transmitting the data is presented as $ge_{h}^{wr}$ .

Delay $D E$ is the examination of transmission as well as propagation delay obtained in the packet, offered in Eq. (21).

$\displaystyle DE=\frac{MX\sum\nolimits_{a=1}^{A}{SP_{a}}}{z}$ (21)

Here, the term $MX\sum\nolimits_{a=1}^{A}{SP_{a}}$ indicates the data transmission between the sensor node and the base station, and the network count is presented as $z$ .

Path loss $P L$ is considered over the energy loss validation when wave propagation is obtained over the transmitter and the receiver. Penalty $P E$ is applied to the approach when any order is used to perform the routing in some cases.

6. Calculation of results

6.1 Simulation setup

A novel detection of attack framework for a secured IoT-based routing model with a deep structure framework was executed in Python. Here, better attack detection rates were obtained by using the population size as 10, maximal iteration rate as 25, and chromosome length as 13. In the recommended attack detection framework, multiple observations were employed over the conventional techniques, and the implemented framework secured a better efficacy rate than the traditional models. Next, the implemented framework was contrasted over various attack detection methods like Convolutional Neural Networks (CNN) [29], Deep Neural Networks (DNN) [30], Recurrent Neural Networks (RNN) [31] and 1-DCNN [26] and also the different heuristic techniques like Deer Hunting Optimization Algorithm (DHOA) [32], Honey Badger Algorithm (HBA) [33], GOA [27] and LOA [28].

6.2 Efficacy metrics

Multiple quantitative metrics performed to examine the recommended attack detection method are elaborated as follows.

(a) The Union of highly precise characteristics as well as the detected features is referred to as precision, and it is presented in Eq. (22).

$\displaystyle PL=\frac{gq}{gq+gr}$ (22)

(b) The comparison of negative essential items that are classified positive in the whole surroundings when an examination is executed is said to be FPR $Y N$ , and it is provided in Eq. (23).

$\displaystyle YN=\frac{gv}{gv+gw}$ (23)

(c) The Examination of double monitoring speed in the validation is termed as MCC $E C$ , and it is presented in Eq. (24).

$\displaystyle EC=\frac{gq\times gw-ge\times gr}{\sqrt{(gq+ge)({gq+gr})(gw+ge)(% {gw+gr})}}$ (24)

(d) The ratio of positive integer examined accurately is referred to as Sensitivity $Q Z$ and offered in Eq. (25).

$\displaystyle QZ=\frac{gq}{gq+gr}$ (25)

(e) The advancements in entire terms utilized for the observation are said to be NPV $R V$ , provided in Eq. (26).

$\displaystyle RV=\frac{gw}{gw+ge}$ (26)

(f) The preposition of negative items obtained correctly is referred to as Specificity $U M$ , given in Eq. (27).

$\displaystyle UM=\frac{gw}{gw+ge}$ (27)

Figure 6.

Confusion matrix observation over the developed attack detection framework.

Figure 7.

State-of-art-analysis on the recommended attack detection framework over (a) Delay, (b) Energy, (c) Path loss and (d) Shortest distance.

Figure 8.

Efficacy validation on the offered attack detection framework over classical detection methods with (a) Accuracy, (b) F1-Score and (c) Precision.

(g) The allocation of positive charge from the negative exploration outcome is given as FNR $W X$ , offered in Eq. (28).

$\displaystyle WX=\frac{gr}{gr+gu}$ (28)

(h) Summing the accurate values in the observation is termed as F1-score $T B$ , which is presented in Eq. (29).

$\displaystyle TB=2\times\frac{2gq}{2gq+ge+gr}$ (29)

6.3 Validation of confusion matrix on the recommended approach

Validation of confusion matrix analysis performed on the recommended attack detection model is offered in Fig. 6. Here, the developed scheme is contrasted with the actual value and the predicted value. In this case, the developed model secured an accuracy rate of 96.87, which is highly better than the classical techniques.

6.4 Validation of the suggested framework using baseline approaches

Different state-of-the-art done on the recommended attack detection framework are represented in Fig. 7. The major goal of the recommended attack detection framework is to minimize the delay, energy, path loss, and shortest distance in the secured IoT routing system. Delay analysis executed in the suggested attack detection model obtained a better delay analysis rate of 17.3%, 12.24%, 10.41%, and 15.68% efficient than the classical techniques like DHOA-MDS-1DCNN, HBA-MDS-1DCNN, GOA-MDS-1DCNN, and LO-MDS-1DCNN, respectively. Thus, the suggested attack detection framework secured a better performance rate than the classical techniques.

6.5 Efficacy validation of the offered approach

Performance analysis employed on the offered attack detection framework over the classical attack detection and heuristic model is shown in Figs 8 and 9. Accuracy validation done in the recommended attack detection model attained an elevated efficacy rate as 21.3% superior to DNN, 15.7% efficient than CNN, 18.8% improved than RNN and 17.11% advanced than 1-DCNN. Hence, the offered approach attained a better efficacy rate than the classical attack detection framework, and the FG-LOA-MDS-1DCNN-based attack detection model attained an elevated performance rate.

6.6 Performance analysis of the suggested technique

Efficacy observation performed in the implemented attack detection framework over the classical baseline techniques and the attack detection methods are tabulated in Tables 2 and 3. Accuracy analysis executed on the implemented attack detection framework secured better efficacy rate as 1.97%, 1.973%, 1.63%, and 0.97% efficient than the conventional techniques like DHOA-MDS-1DCNN, HBA-MDS-1DCNN, GOA-MDS-1DCNN, and LO-MDS-1DCNN, correspondingly. Similarly, different attack detection analyses performed on the suggested framework over the classical models secured a better efficacy rate than the classical techniques. Thus, the suggested FG-LOA-MDS-1DCNN-based attack detection framework achieved a better performance rate than the conventional models.

Table 2
Analysis of the suggested attack detection framework over the classical baseline techniques

TERMS	DHOA-MDS- 1DCNN [32]		HBA-MDS- 1DCNN [33]		GOA-MDS- 1DCNN [27]		LO-MDS- 1DCNN [28]		FG-LOA- MDS-1DCNN
Accuracy	95		95		95	.3125	95	.9375	96	.875
Sensitivity	95		95		95		96	.25	96	.25
Specificity	95		95		95	.41667	95	.83333	97	.08333
Precision	86	.36364	86	.36364	87	.35632	88	.50575	91	.66667
FPR	5		5		4	.583333	4	.166667	2	.916667
FNR	5		5		5		3	.75	3	.75
NPV	98	.27586	98	.27586	98	.28326	98	.71245	98	.72881
FDR	13	.63636	13	.63636	12	.64368	11	.49425	8	.333333
F1-Score	90	.47619	90	.47619	91	.01796	92	.21557	93	.90244
MCC	0	.872786	0	.872786	0	.879957	0	.896178	0	.918527

Figure 9.

Efficacy validation on the recommended model over a classical heuristic model with (a) Accuracy, (b) F1-Score and (c) Precision.

Table 3

Examination of the recommended attack detection technique over the conventional attack detection model

TERMS	DNN [30]		CNN [29]		RNN [31]		1-DCNN [26]		FG-LOA- MDS-1DCNN
Accuracy	80	.9375	85	.9375	85	.625	85	.3125	96	.875
Sensitivity	82	.5	87	.5	86	.25	85		96	.25
Specificity	80	.41667	85	.41667	85	.41667	85	.41667	97	.08333
Precision	58	.40708	66	.66667	66	.34615	66	.01942	91	.66667
FPR	19	.58333	14	.58333	14	.58333	14	.58333	2	.916667
FNR	17	.5	12	.5	13	.75	15		3	.75
NPV	93	.23671	95	.34884	94	.90741	94	.47005	98	.72881
FDR	41	.59292	33	.33333	33	.65385	33	.98058	8	.333333
F1-Score	68	.39378	75	.67568	75		74	.31694	93	.90244
MCC	0	.570022	0	.672455	0	.662559	0	.652646	0	.918527

Figure 10.

Convergence validation on the recommended model over (a) Node 50, (b) Node 100, (c) Node 150 and (d) Node 200.

6.7 Convergence analysis on the developed routing model

Convergence analyses performed on the initiated FG-LOA-MDS-1DCNN-based attack detection framework for secured IoT routing are contrasted over different classical approaches are presented in Fig. 10. Here, the developed model secured a better efficacy rate as 0.72% better than DHOA-MDS-1DCNN, 0.71% improved than HBA-MDS-1DCNN, 0.36% efficient than GOA-MDS-1DCNN and 0.35% efficient than LO-MDS-1DCNN in node 50. Thus, the suggested LOA-MDS-1DCNN-based attack detection framework secured a better efficacy rate than the classical approaches.

6.8 Validation of statistical analysis of the implemented routing technique

Statistical analysis performed in the recommended attack detection framework over different nodes is given in Table 4. The recommended FG-LOA-MDS-1DCNN-based attack detection model secured better performance rates as 0.86%, 2.02%, 0.98%, and 1.43% efficient than the classical heuristic techniques like DHOA-MDS-1DCNN, HBA-MDS-1DCNN, GOA-MDS-1DCNN and LO-MDS-1DCNN regarding best metrics in node variation 50. Hence the implemented attack detection framework attained better performance.

Table 4
Statistical validation on the recommended a model with different node variation

Performance measures	HBA-MDS- 1DCNN [33]	GOA-MDS- 1DCNN [27]	LO-MDS- 1DCNN [28]	FG-LOA- MDS-1DCNN	DHOA-MDS- 1DCNN [32]
Node variation-50
Worst	2.900396	2.911727	3.013867	2.968319	2.864291
Standard deviation	0.024281	0.038049	0.071585	0.044085	0.035336
Median	2.748808	2.800921	2.769276	2.795535	2.758099
Mean	2.755945	2.817364	2.78732	2.805559	2.774732
Best	2.748808	2.781462	2.752136	2.76467	2.725107
Node variation-100
Standard deviation	0.071585	0.090515	0.033668	0.040726	0.072627
Worst	2.766599	3.220752	2.973575	2.974591	3.033905
Median	2.766599	2.776339	2.746395	2.751884	2.782307
Mean	2.766599	2.814375	2.763424	2.768168	2.786335
Best	2.766599	2.758853	2.746395	2.736944	2.719808
Node variation-150
Standard deviation	0.020852	0.087112	0.073135	0.029057	0.046873
Worst	2.84728	3.049754	2.945345	2.848315	2.847107
Mean	2.758835	2.799323	2.781895	2.782037	2.770167
Best	2.737595	2.752441	2.741501	2.768941	2.723223
Median	2.761727	2.75695	2.741501	2.768941	2.73901
Node variation-200
Worst	2.940256	2.864268	2.800077	2.769679	3.021948
Standard deviation	0.049627	0.026906	0.019946	0.009619	0.053712
Mean	2.815314	2.824597	2.764582	2.761715	2.803813
Best	2.741527	2.777096	2.753374	2.741932	2.741214
Median	2.795885	2.810284	2.753374	2.762827	2.794318

7. Conclusion

An innovative energy-efficient attack detection strategy with malicious node mitigation was designed to identify various attacks in the secured IoT-based routing system. At first, the important data for the experiment were aggregated from the IoT network and subjected as input to the data cleaning and transformation phase. Once the data cleaning and transformation were performed, the data were offered to the weighted features selection phase. In this phase, the weights of the essential features were modified by the developed optimization model FG-LOA. Further, the weighted features were subjected to an MDS-1DCNN-based attack detection phase, and the parameters of 1DCNN were tuned by the suggested model FG-LOA to maximize the accuracy of the attack detection rate. Later, effective routing was achieved among the selected nodes, and they effectively examined and reduced several constraints like energy consumption, path loss, delay, and shortest distance. Accuracy analysis performed in the recommended attack detection model secured a better efficacy rate as 21.3% superior to DNN, 15.7% more efficient than CNN, 18.8% improved than RNN, and 17.11% more advanced than 1-DCNN. Thus, the suggested attack detection model secured a better attack recognition rate than the classical techniques.

References

Muzammal

Murugesan

Jhanjhi

. A comprehensive review on secure routing in internet of things: Mitigation methods and trust-based approaches. IEEE Internet of Things Journal. 2021; 8(6): 4186-210.

Raoof

Lung

Matrawy

. Securing RPL using network coding: The chained secure mode (CSM). IEEE Internet of Things Journal. 2022; 9(7): 4888-98.

Mick

Tourani

Misra

. LASeR: Lightweight authentication and secured routing for NDN IoT in smart cities. IEEE Internet of Things Journal. 2018; 5(2): 755-64.

Guo

Lin

Peng

. Deep-reinforcement-learning-based QoS-aware secure routing for SDN-IoT. IEEE Internet of Things Journal. 2020; 7(7): 6242-51.

Zhang

Ren

Song

Liu

Zhang

Qian

. An energy-efficient multilevel secure routing protocol in IoT networks. IEEE Internet of Things Journal. 2022; 9(13): 10539-53.

Hatzivasilis

Papaefstathiou

Manifavas

. SCOTRES: Secure routing for IoT and CPS. IEEE Internet of Things Journal. 2017; 4(6): 2129-41.

Raoof

Matrawy

Lung

. Enhancing routing security in IoT: Performance evaluation of RPL’s secure mode under attacks. IEEE Internet of Things Journal. 2020; 7(12): 11536-46.

Alotaibi

. Improved blowfish algorithm-based secure routing technique in IoT-based WSN. IEEE Access. 2021; 9: 159187-97.

Sugave

Jagdale

. Monarch-EWA: Monarch-earthworm-based secure routing protocol in IoT. The Computer Journal. 2020; 63(6): 817-31.

10.

Haseeb

Islam

Almogren

Din

Almajed

Guizani

. Secret sharing-based energy-aware and multi-hop routing protocol for IoT based WSNs. IEEE Access. 2019; 7: 79980-8.

11.

Liu

Shen

Liu

Jiang

Taleb

. Incentive jamming-based secure routing in decentralized internet of things. IEEE Internet of Things Journal. 2021; 8(4): 3000-13.

12.

Salim

Osamy

Khedr

Aziz

Abdel-Mageed

. A secure data gathering scheme based on properties of primes and compressive sensing for IoT-based WSNs. IEEE Sensors Journal. 2021; 21(4): 5553-5571.

13.

Sharma

Bhardwaj

Banyal

Gupta

Nkenyereye

. An opportunistic approach for cloud service-based IoT routing framework administering data, transaction, and identity security. IEEE Internet of Things Journal. 2022; 9(4): 2505-12.

14.

Sun

Tian

Wang

Fan

. Lightweight anonymous geometric routing for Internet of Things. IEEE Access. 2019; 7: 29754-62.

15.

Abbas

Javaid

Almogren

Gulfam

Ahmed

Radwan

. Securing genetic algorithm enabled SDN routing for blockchain based Internet of Things. IEEE Access. 2021; 9: 139739-54.

16.

Raoof

Matrawy

Lung

. Routing attacks and mitigation methods for RPL-based Internet of Things. IEEE Communications Surveys & Tutorials. 2019; 21(2): 1582-606.

17.

Haseeb

Islam

Almogren

Din

. Intrusion prevention framework for secure routing in WSN-based mobile Internet of Things. Ieee Access. 2019; 7: 185496-505.

18.

Nandhini

Kuppuswami

Malliga

DeviPriya

. Enhanced Rank Attack Detection Algorithm (E-RAD) for securing RPL-based IoT networks by early detection and isolation of rank attackers. The Journal of Supercomputing. 2022; 79(6): 6825-48.

19.

Muzammal

Murugesan

Jhanjhi

Hossain

Yassine

. Trust and mobility-based protocol for secure routing in Internet of Things. Sensors. 2022; 22(16): 6215.

20.

Alsukayti

Singh

. A lightweight scheme for mitigating RPL version number attacks in IoT networks. IEEE Access. 2022; 10: 111115-33.

21.

Seyfollahi

Moodi

Ghaffari

. MFO-RPL: A secure RPL-based routing protocol utilizing moth-flame optimizer for the IoT applications. Computer Standards & Interfaces. 2022; 82: 103622.

22.

Sahay

Geethakumari

Mitra

. A novel blockchain based framework to secure IoT-LLNs against routing attacks. Computing. 2020; 102(11): 2445-70.

23.

Ragesh

Kumar

. Trust-based secure routing and message delivery protocol for signal processing attacks in IoT applications. The Journal of Supercomputing. 2022; 79(3): 2882-9091.

24.

Janani

Ramamoorthy

. Threat analysis model to control IoT network routing attacks through deep learning approach. Connection Science. 2022; 34(1): 2714-54.

25.

Kumar

. BITA-Based Secure and Energy-Efficient Multi-Hop Routing in IoT-WSN. Cybernetics and Systems. 2022; 54(6): 809-35.

26.

Huang

Tang

Dai

Wang

. Signal status recognition based on 1DCNN and its feature extraction mechanism analysis. Sensors. 2018; 19(9): 2018.

27.

Saremi

Mirjalili

Lewis

. Grasshopper optimisation algorithm: theory and application. Advances in Engineering Software. 2017; 105: 30-47.

28.

Abasi

Makhadmeh

Al-Betar

Alomari

Awadallah

Alyasseri

Doush

Elnagar

Alkhammash

Hadjouni

. Lemurs optimizer: A new metaheuristic algorithm for global optimization. Applied Sciences. 2022; 12(19): 10057.

29.

Huang

Zeng

Chen

Chang

Zhang

. Hierarchical digital modulation classification using cascaded convolutional neural network. Journal of Communications and Information Networks. 2021; 6(1): 72-81.

30.

Low

Park

Teoh

. Stacking-based deep neural network: deep analytic network for pattern classification. IEEE Transactions on Cybernetics. 2020; 50(12): 5021-34.

31.

Xia

Wang

. A recurrent neural network for solving nonlinear convex programs subject to linear constraints. IEEE Transactions on Neural Networks. 2005; 16(2): 379-86.

32.

Brammya

Praveena

Ninu Preetha

Ramya

Rajakumar

Binu

. Deer hunting optimization algorithm: A new nature-inspired meta-heuristic paradigm. The Computer Journal. 2019; 133.

33.

Hashim

Houssein

Hussain

Mabrouk

Al-Atabany

. Honey badger algorithm: New metaheuristic algorithm for solving optimization problems. Mathematics and Computers in Simulation. 2022; 192: 84-110.

An effective attack detection framework using multi-scale depth-wise separable 1DCNN via fused grasshopper-based lemur optimizer in IoT routing system

Abstract

Keywords

1. Introduction

2.1 Related works

2.2 Summary

Table 1 Uppercomings and Lowercomings of secured IoT routing against different attacks model

3.1 Proposed energy efficient attack detection-based secured IoT routing model

3.3 IoT routing model

4.1 Data aggregation

4.2 Cleaning of data and transformation

4.3 Weighted feature selection with heuristic tool

5.1 Proposed FG-LOA

6.1 Simulation setup

6.2 Efficacy metrics

6.4 Validation of the suggested framework using baseline approaches

6.5 Efficacy validation of the offered approach

6.6 Performance analysis of the suggested technique

Table 2 Analysis of the suggested attack detection framework over the classical baseline techniques

6.8 Validation of statistical analysis of the implemented routing technique

Table 4 Statistical validation on the recommended a model with different node variation

References

Table 1
Uppercomings and Lowercomings of secured IoT routing against different attacks model

Table 2
Analysis of the suggested attack detection framework over the classical baseline techniques

Table 4
Statistical validation on the recommended a model with different node variation