With the rapid deployment of electronic imaging devices, plenty of high-quality images are in the general public’s hands. These images can be profitable, such as providing retrieval services; however, it is difficult for the individual to profit without the support of the cloud platform. The straightforward idea is that image owners upload their images to the cloud; yet, it is infeasible as the cloud platforms are not fully-trusted. In previous works, in order to protect the privacy of image owners, many researchers consider the Secure content-based Image Retrieval (SIR) task, which enables cloud servers to provide retrieval services while not exposing the images from the owners. However, the existing schemes are often not friendly to users as it’s assumed that the owners have no profit demand and are unwilling to provide extra computation resources. This work introduces federated learning into SIR, which ensures better retrieval accuracy and efficiency; the additive secret sharing technology is utilized to protect the image information, and a better secure comparison protocol is proposed for better efficiency. We believe that the users can enjoy a better secure retrieval service with our proposed scheme. The experiment results and security analysis demonstrate that our scheme provides a significant accuracy advantage while ensuring efficiency and security.

Neural network pruning is essential for deploying deep learning models on resource-constrained devices by reducing computational and memory demands. In this paper, we propose a novel pruning framework, Entropy-Guided Search Space Optimization for Efficient Neural Network Pruning, which uses information entropy to assess the importance of convolutional layers. Specifically, we calculate the layer-wise entropy of pretrained weights, apply outlier detection to remove extreme values, and normalize the entropy values. These normalized values guide the selection of retention ratios, ensuring that layers with higher entropy retain more filters. By refining the subnetwork search space, our approach enhances the efficiency of the search process and improves overall subnetwork performance. The refined search space targets more promising regions, reducing computational overhead and resulting in higher-quality pruned networks. Through iterative optimization, the optimal subnetwork is identified and fine-tuned to produce the final pruned model. Experimental results on benchmark datasets show that our method significantly outperforms existing pruning methods, achieving substantial improvements in both accuracy and computational efficiency. This entropy-guided pruning strategy provides a robust and effective solution for neural network compression, suitable for a wide range of deep learning applications.

Cross-efficiency in data envelopment analysis is widely used for performance evaluation of decision-making units (DMUs), but it neglects the reference relationship between each pair of DMUs. In order to address the problem, this paper introduces the concept of social network to clearly describe the reference relationship between DMUs and then a fully ranking can be made. A novel cross-efficiency approach is developed by integrating social network analysis in this study. Firstly, we propose a pairwise comparison model based on cross-efficiency evaluation to identify the superiority and inferiority between any pair of DMUs. Secondly, based on pairwise comparison, we build a non-weighted directed social network where a direction generates from one DMU to another DMU if the former one references the latter one. Each edge in the unweighted social network captures the learning procedure from worse-performance DMU to better-performance DMU. Comparing with traditional cross-efficiency approaches, our proposed approach considers the reference relationship among DMUs rather than only cross-efficiency scores. Thus, the importance of each DMU in the network can be measured by its centrality. Finally, the proposed approach is employed to environmental efficiency evaluation of the cities along the Xiangjiang River Basin, then, some beneficial universal policies have been summarized.