An Autonomous Underwater Vehicle-Assisted Adaptive Trust Prediction Model for UASNs

Underwateracoustic sensor networks (UASNs) provide significant support for marine intelligent transportation system. As an effective security mechanism, the trust prediction model has been gradually deployed in UASNs to detect anomalous attacks and guarantee the security of marine transportation. However, in existing research on trust models for UASNs, the impact of multidimensional trust features on the trust prediction process is usually treated as the same, even in dynamic current environments and dynamic network topologies. In addition, underwater sensor nodes undertake excessive trust computation in the traditional UASN architecture. In this paper, we introduce the Autonomous Underwater Vehicle (AUV) as the edge device to construct the edge-enabled UASN architecture, and propose an AUV-assisted adaptive Trust Prediction model, which we name ATP. Furthermore, by analyzing the characteristics of the edge-enabled UASN architecture, we extract trust features from AUV and underwater sensor nodes respectively, and AUV is utilized for trust computation and prediction. In order to improve the performance of anomalous node detection, we adaptively adjust the contribution of different trust features to trust prediction based on the attention mechanism. Evaluation results indicate that our proposed ATP model is able to detect anomalous nodes effectively, and it achieves higher accuracy than that of the two established trust models.