Location Protection Technology for Wireless Sensor Networks Based on Differential Privacy using DSRNN-MOA

Wireless Sensor Networks (WSNs) play a pivotal role in modern data-driven applications, yet concerns persist regarding the privacy and security of sensitive location information. The attributes of Wireless Sensor Networks (WSNs) make them susceptible to eavesdropping, enabling attackers to intercept data packets across single or multiple communication links. This interception allows for the extraction of sensitive data from various sensor information, presenting a significant challenge to location privacy. Consequently, it becomes crucial to implement effective measures for safeguarding the privacy of training sample data when utilizing WSNs. In this manuscript, Dynamically Stabilized Recurrent Neural Network (DSRNN) optimized with Mother Optimization Algorithm (DSRNN-MOA) is proposed. Initially data is taken from WSN dataset. Afterward the data is fed to Variational Bayesian-based Maximum Correntropy Cubature Kalman Filtering (VBMCCKF) based pre-processing process. The preprocessing output is provided to the Dynamically Stabilized Recurrent Neural Network to enhance source location protection while addressing challenges related to recurrent network stability and gradient issues. The learnable parameters of the DSRNN is optimized using MOA. The proposed strategy, LPWSN-DSRNN-MOA, is implemented in MATLAB, and its effectiveness is assessed using a number of performance evaluation measures, including ROC analysis, accuracy, precision, recall, f1-score, mean squad error, and recall. The proposed LPWSN-DSRNN-MOA method shows the highest accuracy of 98%, precision of 99%, specificity of 98% and F1-score of 99% while comparing other existing methods such as Location Protection for Wireless Sensor Networks based on Artificial Neural Network(LPWSN-ANN), Location Protection for Wireless Sensor Networks based on Deep Neural Network (LPWSN-DNN), and Location Protection for Wireless Sensor Networks based on Machine Learning (LPWSN-ML) respectively.