Zero-watermarking is an effective solution for image copyright protection without altering the original content. However, current deep learning-based methods suffer from two key limitations. First, most feature extraction networks, originally designed for classification, lack robust feature learning essential for resisting attacks. Second, conventional methods seldom incorporate the generated watermark back into training, missing opportunities to further optimize the model. To address these issues, we propose ConZWNet, a two-stage framework that integrates contrastive learning with feedback-driven zero-watermark generation. In the first stage, we use ConvNeXt to learn invariant, attack-resistant features via contrastive learning on weak–strong augmentation. In the second stage, a residual network coupled with a Multi-Layer Perceptron (MLP) fuses features from host and copyright images to produce a latent zero-watermark, which is then verified by an MLP-based copyright identification network. This feedback loop optimizes feature fusion and transforms zero-watermark generation into a self-supervised process. Extensive experiments demonstrate that ConZWNet achieves state-of-the-art robustness against various attacks while ensuring high distinguishability among host images and copyrights. Ablation studies confirm the effectiveness of components, including two-stage architecture, contrastive learning, weak–strong augmentation, and copyright identification network. The source code is publicly available at https://github.com/hanhongxin1028/ConZWNet .

Time series forecasting is essential for our daily activities and precise modeling of the complex correlations and shared patterns among multiple time series is essential for improving forecasting performance. Spatial-Temporal Graph Neural Networks (STGNNs) are widely used in multivariate time series forecasting tasks and have achieved promising performance on multiple real-world datasets for their ability to model the underlying complex spatial and temporal dependencies. However, existing studies have mainly focused on datasets comprising only a few hundred sensors due to the heavy computational cost and memory cost of spatial-temporal GNNs. When applied to larger datasets, these methods fail to capture the underlying complex spatial dependencies and exhibit limited scalability and performance. To this end, we present a Scalable Adaptive Graph Diffusion Forecasting Network (SAGDFN) to capture complex spatial-temporal correlation for large-scale multivariate time series and thereby, leading to exceptional performance in multivariate time series forecasting tasks. The proposed SAGDFN is scalable to datasets of thousands of nodes without the need of prior knowledge of spatial correlation. Extensive experiments demonstrate that SAGDFN achieves comparable performance with state-of-the-art baselines on one real-world dataset of 207 nodes and outperforms all state-of-the-art baselines by a significant margin on three real-world datasets of 2000 nodes.