Brain network analysis plays an increasingly important role in studying brain function and the exploring of disease mechanisms. However, existing brain network construction tools have some limitations, including dependency on empirical users, weak consistency in repeated experiments and time-consuming processes. In this work, a diffusion-based brain network pipeline, DGCL is designed for end-to-end construction of brain networks. Initially, the brain region-aware module (BRAM) precisely determines the spatial locations of brain regions by the diffusion process, avoiding subjective parameter selection. Subsequently, DGCL employs graph contrastive learning to optimize brain connections by eliminating individual differences in redundant connections unrelated to diseases, thereby enhancing the consistency of brain networks within the same group. Finally, the node-graph contrastive loss and classification loss jointly constrain the learning process of the model to obtain the reconstructed brain network, which is then used to analyze important brain connections. Validation on two datasets, ADNI and ABIDE, demonstrates that DGCL surpasses traditional methods and other deep learning models in predicting disease development stages. Significantly, the proposed model improves the efficiency and generalization of brain network construction. In summary, the proposed DGCL can be served as a universal brain network construction scheme, which can effectively identify important brain connections through generative paradigms and has the potential to provide disease interpretability support for neuroscience research.

One of the main reasons for Alzheimer’s disease (AD) is the disorder of some neural circuits. Existing methods for AD prediction have achieved great success, however, detecting abnormal neural circuits from the perspective of brain networks is still a big challenge. In this work, a novel decoupling generative adversarial network (DecGAN) is proposed to detect abnormal neural circuits for AD. Concretely, a decoupling module is designed to decompose a brain network into two parts: one part is composed of a few sparse graphs which represent the neural circuits largely determining the development of AD; the other part is a supplement graph, whose influence on AD can be ignored. Furthermore, the adversarial strategy is utilized to guide the decoupling module to extract the feature more related to AD. Meanwhile, by encoding the detected neural circuits to hypergraph data, an analytic module associated with the hyperedge neurons algorithm is designed to identify the neural circuits. More importantly, a novel sparse capacity loss based on the spatial-spectral hypergraph similarity is developed to minimize the intrinsic topological distribution of neural circuits, which can significantly improve the accuracy and robustness of the proposed model. Experimental results demonstrate that the proposed model can effectively detect the abnormal neural circuits at different stages of AD, which is helpful for pathological study and early treatment.

In this digital era, social images are the most vital information carrier on multimedia social platforms. More and more users are interested in sharing social images with mobile terminals on multimedia social platforms. Social image sharing also faces potential risks from malicious use, such as illegal sharing, piracy, and misappropriation. This paper mainly concentrates on secure social image sharing. To address how to share social images in a safe way, a social image security scheme is proposed. The technology addresses the social image security problem and the active tracing problem. First, discrete wavelet transform (DWT) is performed directly from the JPEG image. Then, the high-bit planes of the LL, LH, and HL are permuted with cellular automation (CA), bit-XOR, and singular value decomposition (SVD) computing, and their low-bit planes are chosen to embed a watermark. In the end, the encrypted and watermarked image is again permuted with cellular automation in the discrete cosine transform (DCT) domain. Experimental results and security analysis show that the social image security method not only has good performance in robustness, security, and time complexity but can also actively trace the illegal distribution of social images. The proposed social image security method can provide double-level security for multimedia social platforms.

The brain network is an effective tool and has been widely used in the field of brain neurodegenerative disease analysis. Due to the high cost of accessing medical image data, efforts have been devoted to investigating data augmentation. However, the brain network containing topological characteristics of non-European space is different from the traditional image data, which makes it challenging to synthesize brain structural connectivity and limits the application of brain network analysis. In this paper, a Hemisphere-separated Cross-connectome Aggregating Learning (HCAL) model is proposed to synthesize realistic and diverse brain structural connectivities. Specifically, the latent representation is transformed from structural connectivity by the graph variational autoencoder (GVAE). To generate more diverse and high-quality structural connectivities, the hemisphere-separated generator with a cross-connectome aggregating mechanism is developed to first learn local topological patterns by splitting the whole brain into inter- and intra-hemispheres, then capture global topological characteristics among all the neighbors for each brain region. Also, the connectivity-aware discriminator is devised to make the adversarial training stable and enhance the disease diagnosis. Evaluations of the public Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset show that the proposed model generates brain structural networks with higher quality and more diversity than related methods. In addition, the classification performance using the augmented data achieved by our approach achieves an accuracy improvement of over 3% compared with the competing method. The proposed model provides an alternative way to synthesize brain networks for connectivity-based disease analysis.

Federated Learning (FL) uses local data to perform distributed training on clients and combines resulting models on a public server to mitigate privacy exposure by avoiding data sharing. However, further research indicates that communication overheads continue to be the primary limitation for FL relative to alternative considerations. This is especially true when training models on non-independent and identically distributed data, such as financial default risk data, where FL’s computational costs increase significantly. This study aims to address financial credit risk data by establishing a dynamic receptive field adjustment mechanism for feature extraction, efficiently extracting default features with varying distributions and attributes. Additionally, by constructing a distributed feature fusion architecture, characteristics from both local and overarching models are aggregated to attain higher accuracy with lower transmission costs. Experimental results demonstrate that the proposed FL framework can utilize dynamic receptive fields to allocate convolutional kernel weights, thereby improving feature extraction. In the feature fusion stage, the proposed Multi-Fusion strategy efficiently customizes the aggregation of features from local and global models. The final solution reduces the communication rounds in federated learning by approximately 80%.

As big data technologies continue to evolve, recommendation systems have found broad application in domains such as online retail and social networking platforms. However, centralized recommendation systems raise numerous data privacy concerns. Federated learning addresses these concerns by allowing model training on client devices and aggregating model parameters without sharing raw data. Nevertheless, federated learning faces critical challenges related to feature extraction efficiency and noise sensitivity, limiting its application in e-commerce recommendation systems where data heterogeneity and high-dimensional features are prevalent. To address these gaps, this paper introduces a novel multi-view federated learning framework, Fed-FR-MVD, designed to enhance feature extraction efficiency and improve recommendation accuracy in e-commerce applications. Fed-FR-MVD integrates a FR mechanism within a multi-view structure, incorporating both item and user perspectives to improve feature representation and robustness. This approach yields a 12%–18% increase in recommendation accuracy across various performance metrics compared to single-view and other multi-view methods. By addressing data heterogeneity and optimizing feature utilization through dynamic rescaling, Fed-FR-MVD effectively mitigates the impact of noisy data, with performance maintained across noise levels of 5%–15%. Experimental results demonstrate that Fed-FR-MVD fills a key research gap by providing a more resilient and efficient framework for federated recommendation systems in privacy-sensitive and data-diverse e-commerce environments.