Improved Dragonfly Optimization and Adaptive Long Short‐Term Memory Model for Wireless Sensor Network Security Enhancement

Recently, wireless sensor networks have emerged as a significant technology in industrial monitoring, smart farming, and healthcare systems. Due to their open and distributed nature, wireless sensor networks are highly vulnerable to various attacks and energy constraints. Various artificial intelligence‐based techniques have been developed for energy‐efficient routing and intrusion detection, but their performances are limited by poor cluster head selection and slower convergence. To address these limitations, this study proposes an attention‐based improved dragonfly algorithm with adaptive long short‐term memory for effective energy‐efficient routing and intrusion detection in wireless sensor networks. The density peak K‐means clustering is designed to form stable clusters of sensor nodes according to the density. The improved dragonfly optimization algorithm is established to select the cluster head optimally, which utilizes the global search capability of the multi‐verse optimization algorithm with the exploration–exploitation balancing ability of the Lévy flight strategy, thereby reducing the ineffective solutions. Additionally, the multi‐level self‐attention mechanism is designed to extract the features from the traffic data that capture the short‐term, long‐term, and local variations. Further, an adaptive long short‐term memory model is designed to detect the attacks that dynamically adjust its weights based on network traffic changes, thereby improving the detection performance. Extensive experimental evaluations demonstrate that the proposed framework outperforms existing models with a higher accuracy of 98.4%, a higher packet delivery ratio of 98.2%, highlighting its suitability in secure and energy‐efficient wireless sensor network applications.