JoWUA

Abstract

Among IoT networks, Wireless Sensor Networks (WSNs) play a very important role. The wireless WSN has a crucial energy efficiency vs security (reliability) tradeoff when routing data; The protocols usually don't analyze data importance vs reliable behavior of a node and would cause early collapse of a WSN and security attacks. This article presents SAGT (Semantic-Aware Graph Transformer) in an end-to-end framework that aims to perform data routing and trust estimation at the same time. By injecting the data-oriented semantic contexts, e.g., data urgency and node role in the network, into the multi-head graph attention, SAGT could select better nodes and bypass the untrustworthy nodes/resource exhausted nodes. Experimental simulations confirmed the framework's higher performance; the introduced model attains a 94.8% PDR and a 15% reduction in the average energy consumption per packet against state-of-the-art ones. Furthermore, the framework reaches 91.4% accuracy on the detection of the adversarial nodes, effectively isolating the Byzantine attacks from the graph network. This demonstrates that semantic-aware attention significantly overcomes semantic-insensitive routing with respect to performance and efficiency, as it guarantees a flexible, adaptable, yet reliable method that fully supports the deployment of future intelligent and mission-critical networked applications by narrowing the communication-efficiency semantics gap. This approach to cross-layer optimization represents a landmark for future self-healing wireless networks and ultra-reliable networks.