Linear Projective Statistical Regressive Mamba Deep Learning for Multifactor Node Authentication in WSN

Abstract

A Wireless Sensor Network (WSN) is a network consists of numerous small electronic devices called sensor nodes that are distributed within the environments to collect, monitor, and transmit data wirelessly. These sensors transmit the collected data to a central system or sink nodes for analysis. As the data is transmitted over an insecure channel, they are vulnerable to a variety of attacks, and hence the network security plays an important part in the WSN. Thus, a secure authentication protocol is necessary to avoid security issues within WSNs. Moreover, the authentication process validates and identifies devices before transmitting the data. In this paper, a novel Linear Projective Statistical Regressive Mamba Deep Learning (LiPSR-MDL) model is developed for efficient node authentication to perform secure data environmental data transmission in WSN. In Mamba Deep Learning architecture, number of sensor nodes is collected as an input. After that, multiple characteristics of sensor nodes such as residual energy, trust level, cooperative score is calculated in convolutional layer. Followed by, Mamba architecture utilizes the Structured State Space Sequence (SSM) model to effectively analyze the   multifactor characteristics of the sensor nodes by using statistical kitchen sink regressive analysis. Based on the regression analysis, normal and malicious sensor nodes are authenticated using sigmoid activation function. The output layers display the node authentication results in an accurate manner. Finally, secure environmental data transmission is carried out between the normal sensor nodes and sink nodes. Experimental analysis is carried out with metrics authentication accuracy, authentication time, energy efficiency, transmission success ratio, data drop rate and jitter with respect to different number of environmental data packets and sensor nodes. The overall simulation analyses illustrate that the proposed LiPSR-MDL model improves the authentication accuracy, transmission success ratio, energy efficiency, and minimizes the data drop rate and jitter when compared to conventional methods.