Energy-efficient secure routing in wireless sensor networks using fuzzy inference clustering and attention-based multi-scale deep learning

Wireless Sensor Networks (WSNs) are widely used in applications such as environmental monitoring, smart agriculture, healthcare, industrial automation, and military surveillance. However, their performance and lifetime are often limited by energy constraints, inefficient clustering, insecure routing, and vulnerability to network attacks. Existing approaches frequently suffer from high energy consumption, increased routing overhead, poor scalability, and limited accuracy in detecting intrusions. To address these issues, the proposed approach employs a Mamdani-type Fuzzy Inference System (FIS) for adaptive cluster formation, ensuring balanced and stable clustering. Cluster Head (CH) selection is carried out using Dynamic Adaptive Fig Tree–Wasp Symbiotic Coevolutionary Optimization (DA-FTWSCO) to improve energy efficiency and extend network lifetime. For secure communication, the Adaptive Trust-Synchronized Packet Control Protocol (ATSPCP) is used to compute trust values, followed by optimal path selection using the Improved Grizzly Bear Fat Increase Optimizer (IGBFIO). Additionally, an Enhanced Multi-scale Dilated MobileNet with Attention Mechanism (EMSD-MobileNet-AM) is utilized for effective intrusion detection, enabling accurate identification of denial-of-service (DoS) and zero-day attacks. Simulation outcomes demonstrate that the proposed method achieves low energy consumption (1.02 J), high throughput (0.93 Mbps), low end-to-end delay (4.0 ms), and a high attack detection rate (98%). The results demonstrate that the proposed framework outperforms several existing methods in terms of efficiency, security, and scalability.