GPrUO‐TD2NN: Optimized Drift‐Enabled Deep Learning Framework for Intrusion Detection in Wireless Sensor Networks

ABSTRACT Intrusion detection holds a significant application in identifying malicious packets and maintaining cybersecurity. The multiplex distribution of network traffic and the evolving nature of threats impact the effectiveness of the detection methods. Moreover, the prevailing approaches faced complexities due to the excessive false alarms and feature losses. The research proposes grouped probabilistic update optimized transfer learning with drift‐enabled distributed neural network (GPrUO‐TD2NN) to address the challenges and to exhibit effective outcomes for diverse network threats. The internal covariance of the data distributions is identified using the drift mechanism, and the generalization of targeted data is enhanced by including transfer learning. In addition, the grouped probabilistic update optimized synthetic minority oversampling (GPrUS2M) approach refines the data by minimizing the imbalances, and neighbor sets are chosen with a definiteness over the attributes of each instance. The grouped probabilistic update optimization (GPrUO) algorithm enhances the selection of better neighbor sets and also assists in parameter tuning of the detection model. Moreover, the effectiveness of the GPrUO‐TD2NN in intrusion detection is computed using various metrics that exhibit superior outcomes with higher 96.92% precision, 97.29% recall, 97.16% accuracy, and 97.10% F1 score using 90% training.