Linear Regressive Momentum Optimized Dense Neural Network for Energy and Secure Data Transmission in WSN

Wireless Sensor Networks (WSNs) is a distributed network initially designed as simple monitoring systems comprising numerous sensor nodes focusing on data collection, processing, and transmission. In WSNs, a huge amount of vulnerabilities can arise, specifically those initiating from malicious nodes (MNs), which direct to cooperate data integrity, network stability, and reliability. Although security remains critical, current MN detection methods are time-consuming and increased latency for constrained WSNs. In order to overcome these issues, a novel Linear Regressive Momentum Optimized Dense neural Network (LiRMO-DenseNet) model is developed to enhance the data transmission security in WSN. The proposed LiRMO-DenseNet model utilizes the dense neural network concept to categorize the sensor nodes as legitimate sensor nodes or intruders with help of several layers such as input, numerous hidden layers, and output layer. First, the number of sensor nodes is given to the input layer. After that, the input layer transmits the collected sensor nodes to the first hidden layer. In that layer, the different characteristics of the sensor nodes like energy, cooperativeness and trust level are computed. Then, the computed values of the sensor nodes are given to third hidden layer. In that layer, Changepoint linear regression analysis is carried out for analyzing the sensor node with their characteristics by setting the threshold. Depending on the analysis, the nodes are classified as legitimate sensor nodes and intruders. A new part of this process is fine-tuning of dense neural network, where the Metaheuristic Walrus Optimization algorithm is employed to update the hyperparameter of dense neural network for minimizing the training and validation errors, thereby boosting the accuracy of node classification. Finally, the accurate node classification is carried out at output layer. With the selected legitimate sensor nodes, secure data transmission is achieved in WSN. The effectiveness of the proposed LiRMO-DenseNet model is assessed using a comprehensive set of performance measures, including accuracy, confidentiality rate, data integrity rate, packet delivery rate, throughput and delay. The simulation findings demonstrate that the proposed LiRMO-DenseNet model consistently achieves superior security performance, exhibiting higher confidentiality and reduced delay compared to existing deep learning based methods