Entropy Dove Swarm Optimization (Edso) Based Cluster Head Selection and Stacked Ensemble Learning-Clone Attack Detection (Sel-Cnd) for Wireless Sensor Network (Wsn)

Over the previous several decades, academics have paid much more attention to the development of the wireless sensor networks (WSNs) technology. The advantages for technology is the sheer quantity of sensor nodes it has. Through communication, the sensors are able to establish a network. Typically, these sensor nodes are employed in a variety of apps, including target tracking, fire detection, pressure monitoring, health monitoring. The drawback is that WSNs frequently operate in dangerous, crucial situations without limiting physical access. They are extremely susceptible to node replication or clone node assaults because of this reality. After seizing and extracting the authentic sensor nodes, the adversary can launch a replication attack and gather sensitive data including node IDs and keys. This potential will therefore make it easier for the attacker to take over the entire network and perform all of the authorized nodes’ activities. It is crucial for researchers to develop a mitigation strategy and a detection methodology for clone attacks in light of this vulnerability. In this paper, clone attack detection protocol termed as Stacked Ensemble Learning-Clone Attack Detection (SEL-CND) is proposed.This identifies the Mobile WSN’s clone nodes. The sensor network is segmented into clusters in this standard. A cluster head and an arbitrary number of sensor nodes are present in each cluster. Cluster Head is selected by Entropy Dove Swarm Optimization (EDSO) algorithm has been introduced for improving the performance in the network. EDSO algorithm has been introduced based on the foraging behaviors of doves. Then SEL-CND mechanism has been introduced to detect the clone node present in the WSN. SEL-CND system is performed depending on the Bi-Directional Long Short Term Memory (Bi-LSTM), Gated Recurrent Unit (GRU), and Autoencoder (AE). A meta-learner is trained to produce predictions utilizing the many predictions made by diverse weak learners, and stacking frequently takes these predictions into account, learns them concurrently, and integrates together. to identify clones both locally and globally via an affordable identification verification process. The simulation’s output indicates that a number of factors, including false negative and ratio analyses, precision, recall, and detection analyses, have performance analyses.