Trust-Aware Social-System-Inspired Clustering for Large-Scale Knowledge Discovery in Wireless Sensor Networks

Wireless sensor networks (WSNs) can be conceptualized as large-scale, dynamic social systems where nodes interact to achieve collective objectives. These networks generate extensive data through interactions, offering opportunities for large-scale knowledge discovery to optimize operations and enhance resilience. However, challenges such as limited resources and susceptibility to distributed denial-of-service (DDoS) attacks necessitate efficient and secure mechanisms for managing these “social” interactions. This article proposes a lightweight trusted framework that applies computational modeling principles to clustering in WSNs. The framework employs bi-directional long short-term memory (Bi-LSTM) networks for malicious node detection, mirroring the role of anomaly detection in social systems, and uses the walrus optimization algorithm (WOA) for optimized cluster head (CH) selection. By considering parameters such as residual energy, proximity to base stations, node density, and trust value, WOA ensures effective “role assignment” within the network, similar to optimizing functional roles in human social systems. The Bi-LSTM model analyzes node behavior to exclude malicious actors, fostering trusted, and efficient clustering. Evaluated in simulated DDoS attack scenarios, the framework significantly reduces the impact of attacks by isolating malicious nodes while improving network performance and resilience. Metrics such as stability period, throughput, network lifetime, energy efficiency, and attack mitigation are analyzed, demonstrating the framework’s effectiveness. This research bridges the domains of social system modeling and WSN operations, providing an energy-efficient and secure solution for managing large-scale dynamic networks.