Jiang, BinFeng, JiachengCui, XuerongLuo, FeiWang, Huihui HelenSong, Houbing2025-06-172025-06-172025Bin Jiang et al., “GCN-LG-Trust Model for Attack Detection and Cluster Optimization in Underwater Wireless Sensor Networks,” IEEE Sensors Journal, 2025, 1–1, https://doi.org/10.1109/JSEN.2025.3570812.https://doi.org/10.1109/JSEN.2025.3570812http://hdl.handle.net/11603/38927The researchers focus on the security issues of underwater wireless sensor networks (UWSNs) due to the specificity of network environment. In recent years, researchers have proposed trust models to assess the reliability of each node in the network. However, nodes in UWSNs are usually managed in clusters. Existing trust model evaluation algorithms and model training methods are not well adapted to the network clustering structure. For this reason, we propose a trust model optimization based on graph convolutional network (GCN) and LightGBM for cluster management in this paper, which we named GCN-LG-Trust. This trust model designs a trust parameter estimation method based on the node activity characteristics of the cluster network. And the network trust evidence is used to train the trust model by GCN-LG deep learning method. The GCN-LG-Trust model is tested in single attack and hybrid attack scenarios. The experimental results show that the proposed method has better results in terms of trust evaluation and accuracy.12 pagesen-US© 2025 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.AccuracyMachine learning algorithmsUMBC Security and Optimization for Networked Globe Laboratory (SONG Lab)SensorsWireless sensor networksHidden Markov modelsTrust modelUnderwater wireless sensor networkClustering algorithmsMalicious attackClustering networkEnergy consumptionComputational modelingLightGBMGraph convolutional networkAttenuationAdaptation modelsText