In recent times, WSN has extended the application range from early deployment for battlefield intelligence surveillance to fields like meteorological weather forecasting, emergency response support, factory automation security application, and so on. WSN consists of inexpensive and small sensors without a current architecture [1]. They are frequently utilized for sensing, processing, transmitting, and receiving data from the region they are positioned beforehand it is transported to a base station. A conventional WSN contains several sensors that are classified based on the environment and the structure (topology) where they are deployed. Basically, WSN is classified based on the deployment of the sensors in the environment [2]. This node is of equivalent capacity, while others have different capacities, based on the infrastructure. The three major kinds type WSN architecture are hierarchical, flat-based (tree), and cluster-based. Moreover, the environment where the sensors are positioned in a WSN is classified into five groups, such as mobile WSN, underground WSN, terrestrial WSN, underwater WSN, and multimedia WSN system [3]. The sensors in a WSN are frequently deployed in inaccessible, remote, and harsh regions and are frequently represented by resource constraints including limited storage, limited power, short communication range, and limited bandwidth [4]. They are integrated with the susceptibility of the wireless medium (that is open and shared) have generated sensors vulnerable to distinct security attacks namely the denial of service (DoS) [5]. Fig. 1 illustrates the WSN structure.

Even though conventional security approaches like authentication and cryptography could offer security at certain level, they alone could not handle compromised node attacks [6]. When the nodes are compromised, it launches attacks based on commands from the outside that could control or cripple the entire WSN system. For instance, malicious node could attract the information from another node via distinct method, and when it begins to receive the information, it could randomly receive or drop all information that considerably reduces the routing system efficiency [7]. To handle this type of node is to detect and monitor them. While there are no centralized authorities in WSN, nodes must detect and monitor malicious nodes in a distributed way. Several solutions are presented for securing WSN [8], together with routing. Since routing executes data distribution to BS, it is major protocol for WSN. Therefore, secured routing is strong against packet drop and disruption, and alteration act on routing process is vital [9,10]. To protect routing, particularly against compromised nodes, several solutions have been introduced. One of these solutions is trust establishment, utilized in several study areas. Trust establishment identifies untrustworthy and trustworthy nodes by estimating them according to past performance or behavior. It prevents untrustworthy nodes and only chooses trustworthy in routing process [11]. As trust method is efficient and simple in compromised node detection, a considerable study is performed for enhancing cooperation and improving security in the networks.

Rathee et al. [12], proposed an ACO based QoS aware energy balancing secured routing (QEBSR) approach for WSN. Improved heuristics to calculate the end-to-end delay of communication and the trust factor of the node on the routing path are presented. The presented method has been related to 2 current approaches: energy efficient routing with node compromised resistance and distributed energy balanced routing. Haseeb et al. [13] developed an energy-effective and secured routing method (ESR) for avoiding intrusion from IoT based WSN to improve the data trustworthiness and network period. First, the presented method makes distinct energy-effective clusters based on the intrinsic qualify of nodes. Next, according to the (k,n) threshold-based Shamir secret sharing system, the security and reliability of the sensory data amongst the BS and CH are accomplished.

Haseeb et al. [14] presented an intrusion prevention architecture for mobile IoT devices using the incorporation of WSN for providing data security with enhanced network delivery ratio. The presented method is comprised of two sub-elements. First, autonomously organized and non-overlapping clusters are created and sustained the clusters stability according to the uncertainty principle. Next, end-to-end secured and multihop routing path is designed according to the blockchain architecture. In Selvi et al. [15], a secured routing method named energy aware trust based secured routing approach is presented whereby the trust score assessment is utilized for identifying the malevolent user efficiently in WSN and spatio-temporal constraint is utilized with DT approach to choosing optimal path.

This study develops a novel statistical analysis with dingo optimizer enabled reliable routing scheme (SADO-RRS) for WSN. The proposed SADO-RRS technique aims to detect the existence of attacks and optimal routes in WSN. In addition, the presented SADO-RRS technique derives a new statistics based linear discriminant analysis (LDA) for attack detection, Moreover, a trust based dingo optimizer (TBDO) algorithm is applied for optimal route selection in the WSN and accomplishes secure data transmission in WSN. Besides, the TBDO algorithm involves the derivation of the fitness function involving different input variables of WSN. In order to demonstrate the enhanced outcomes of the SADO-RRS technique, a wide range of simulations was carried out.

In this study, a novel SADO-RRS technique has been developed for reliable routing in WSN. The proposed SADO-RRS technique determines the presence of attacks and optimal routes in WSN. In addition, the presented SADO-RRS technique derives a new statistics based LDA for attack detection. Furthermore, a TBDO algorithm is applied for optimal route selection in the WSN and accomplishes secure data transmission in WSN.

This section inspects the performance validation of the SADO-RRS technique with existing techniques. The results are inspected interms of distinct aspects.

Tab. 1 provides detailed intrusion detection results of the SADO-RRS technique with recent methods. Fig. 3 inspects the precn analysis of the SADO-RRS technique with compared methods. The figure reported that the SVM model has resulted in least precn of 0.6758. Followed by, the SOM and ANN-IDS models have obtained slightly enhanced precn values of 0.7432 and 0.7948 respectively. In line with, the DMN-NB, GAN, and DELM techniques have accomplished moderately closer precn values of 0.8021, 0.8531, and 0.8964 respectively. However, the SADO-RRS model has attained maximum precn of 0.9313.

Fig. 4 examines the recal analysis of the SADO-RRS technique with compared methods. The figure exposed that the SVM model has resulted to least recal of 0.7111. Then, the SOM and ANN-IDS systems have obtained slightly enhanced recal values of 0.7732 and 0.8313 correspondingly. Likewise, the DMN-NB, GAN, and DELM techniques have accomplished moderately closer recal values of 0.8283, 0.8830, and 0.9310 correspondingly. But, the SADO-RRS method has attained higher recal of 0.9546.

Fig. 5 demonstrates the accy analysis of the SADO-RRS technique with compared methods. The figure outperformed that the SVM methodology has resulted in minimal accy of 0.6950. Afterward, the SOM and ANN-IDS techniques have obtained somewhat increased accy values of 0.7550 and 0.8120 correspondingly. Besides, the DMN-NB, GAN, and DELM techniques have accomplished moderately closer accy values of 0.8150, 0.8650, and 0.9123 correspondingly. At last, the SADO-RRS model has attained superior accy of 0.9437.

Tab. 2 and Fig. 6 illustrate the network lifetime (NLFT) examination of the SADO-RRS technique with existing ones under distinct number of compromised nodes. The results indicated that the SADO-RRS technique has resulted in increased NLFT over the other methods under all compromised nodes. For instance, with 5% of compromised nodes, the SADO-RRS technique has offered higher NLFT of 961 rounds whereas the DEBR, EENC, and QEBSR techniques have attained lower 793, 878, and 927 rounds respectively. Similarly, with 30% of compromised nodes, the SADO-RRS approach has obtainable maximal NLFT of 919 rounds whereas the DEBR, EENC, and QEBSR methodologies have attained lower 615, 766, and 868 rounds correspondingly.

Tab. 3 and Fig. 7 depict the NLFT examination of the SADO-RRS approach with existing ones under distinct number of nodes. The outcomes referred that the SADO-RRS methodology has resulted in increased NLFT over the other methods under all nodes. For instance, with 100 nodes, the SADO-RRS approach has accessible maximal NLFT of 1040 rounds whereas the DEBR, EENC, and QEBSR systems have obtained minimal 693, 827, and 888 rounds correspondingly. At the same time, with 400 nodes, the SADO-RRS approach has obtainable superior NLFT of 2439 rounds whereas the DEBR, EENC, and QEBSR methodologies have reached minimal 1976, 2098, and 2360 rounds correspondingly.

Tab. 4 and Fig. 8 depict the PDR examination of the SADO-RRS technique with existing ones under distinct count of compromised nodes. The results referred that the SADO-RRS technique has resulted in maximum PDR over the other methods under all compromised nodes. For instance, with 5% of compromised nodes, the SADO-RRS system has offered increased PDR of 0.9670 rounds whereas the DEBR, EENC, and QEBSR systems have reached lesser of 0.9363, 0.9617, and 0.9649 rounds correspondingly. Also, with 30% of compromised nodes, the SADO-RRS approach has obtainable increased PDR of 0.9572 rounds whereas the DEBR, EENC, and QEBSR approaches have achieved reduced 0.9369, 0.9418, and 0.9480 rounds correspondingly.

A detailed PLR examination of the SADO-RRS technique with recent models in Tab. 5 and Fig. 9. The outcomes indicated the betterment of the SADO-RRS technique over the other methods under distinct number of compromised nodes. For instance, with 5% of compromised nodes, the SADO-RRS technique has attained reduced PLR of 0.0330 whereas the DEBR, EENC, and QEBSR techniques have obtained least PLR of 0.0637, 0.0383, and 0.0351 respectively. Also, with 30% of compromised nodes, the SADO-RRS approach has reached lower PLR of 0.0428 whereas the DEBR, EENC, and QEBSR approaches have attained minimal PLR of 0.0631, 0.0582, and 0.0520 correspondingly.

In this study, a novel SADO-RRS technique has been developed for reliable routing in WSN. The proposed SADO-RRS technique determines the presence of attacks and optimal routes in WSN. In addition, the presented SADO-RRS technique derives a new statistics based LDA for attack detection. The presented approach develops a FF utilizing 3 input parameters such as trust level, distance to neighbors, and energy to optimum route selective. In the proposed technique, it could be vital to decrease the linear group of main functions. The experimental result analysis of the SADO-RRS technique highlighted the enhanced outcomes. In future, data aggregation approaches can be integrated into the SADO-RRS technique for enhanced performance in WSN.