Recent years have witnessed the fast development of the Financial Internet of Things (FIoT), which integrates the Internet of Things (IoT) into financial activities. At the same time, the FIoT is facing an increasing number of stealthy network attacks. Long short-term memory (LSTM) can be used as an anomaly-detecting method to perceive such attacks since it specializes in discovering anomaly behaviors through the time correlation in FIoT traffic. However, current LSTM-based anomaly detection schemes have not considered the specific correlations among the features of the whole traffic. In addition, current schemes are usually trained based on local traffic with rare cooperation among different detecting nodes, leading to the result that current schemes usually suffer from insufficient adaptability and low coordination. In this paper, we propose a feature-attended federated LSTM (FAF-LSTM) for FIoT to address the above issues. FAF-LSTM combines feature-attended LSTM and federated learning to make full use of the deep correlation in data and enhance the accuracy of the trained model via cooperation among different detecting nodes. In FAF-LSTM, the features are grouped so that the model can learn the time–spatial correlation inner the flows of each group as well as their impact on the output. Meanwhile, the parameter aggregation is optimized based on feature correlation analysis. Simulations are conducted to verify the effect of FAF-LSTM. The results show that FAF-LSTM has good performance in anomaly detection. Compared with independently trained LSTM and traditional federated learning-based LSTM, FAF-LSTM can improve the detection accuracy by up to 39.22% and 334.36%, respectively.

Resource-based regions support national economic development and are essential sources of basic energy and raw materials. In the post-pandemic era, however, there are practical situations to deal with, such as a fractured industrial chain, a weaker industrial structure, and a sharp reduction in economic benefits. Based on data collected from 68 cities in China, from 2010 to 2021, with 816 observations, this paper explores the industrial development process of resource-based regions in China and the change in the toughness of the industrial structure under the impact of COVID-19. The paper studies and analyzes industrial development trends, industrial structure toughness, and spatial spillover effects. The methods used are the Markov chain model and the Industrial Structure Advancement Index. By building the spatial Dubin model, the paper analyzes the spatial spillover effect of regional industrial development. It decomposes the spillover effect using the partial differential model based on regression. The results show that, during the study period, the comprehensive development level of industries in resource-based regions in China was slowly improving and tended to stabilize after entering the post-pandemic era. The evolution of an advanced industrial structure is significantly heterogeneous among regions, and each region has different toughness. The impact of COVID-19 has reduced the toughness of China’s resource-based regions’ industrial structure. The spatial spillover effect of regional industrial development is significant. Labor force, technology input, and industrial-structure optimization have different impacts on the industrial development of neighboring regions. In the post-pandemic era, China has used new management methods for more innovation. In order to achieve low-carbon, environmental protection, and sustainable development of resources, realize the rapid recovery of the toughness of industrial structure in China’s resource-based cities, and reduce the impact of the COVID-19 pandemic, China proposes to expand the supply of resources, improve the allocation of resources, optimize the direction, promote the rational flow and efficient aggregation of various factors, and enhance the impetus for innovation and development.

Identifying the spatial attributes of economic, social and environmental development is a prerequisite for China to raise the quality of development. Based on the parallel data of 30 Chinese provinces from 2010 to 2019, this study uses a spatial econometric model to explore the spatial impacts of corporate social responsibility (CSR) and industrial structure upgrading on carbon emissions. The regulating effect of CSR during industrial structure adjustment for carbon emission reduction was also analyzed. It was found that regional carbon emissions were reduced due to CSR and industrial structure adjustment, and the former was beneficial during industrial structure adjustment for carbon emission reduction. The carbon emissions in neighboring areas have also been suppressed to some extent as a result of industrial structure adjustment. However, CSR can encourage some “free riding” behaviors, due to the economic externalities, which emit more carbon into the surroundings. Additionally, carbon emissions show different responses to CSR in various regions. Therefore, strategies must enhance overall social responsibility and formulate different policies in various regions to promote CSR as an influential factor in curbing carbon emissions. Moreover, spatial governance should consider the comparative advantages of different regions, form complementary advantages and fully enhance the cooperation between CSR and industrial structure upgrading on carbon emission reduction.

Geological hazards cause changes in the quality of the ecological environment, affect the function and stability of ecosystems, and negatively impact the maintenance and restoration of ecological functions in ecological functional areas (EFAs). This study integrates machine learning, geographic information technology, and multivariate statistical analysis modeling to develop a technical framework for quantitative analysis of ecological risk assessment (ERA) based on the causal logic between geological hazards and ecosystems. The results of the geological disaster ERA are mapped to EFAs, effectively identifying and quantifying the risk characteristics of different EFAs. The results show that: (1) The hazard–vulnerability–exposure ERA framework effectively identifies the distribution characteristics of high ecological risk around the Qilian Mountains, with high risk in the east and low risk in the west. (2) In high ecological risk areas, high hazard–high vulnerability–low exposure is the main combination pattern, accounting for 83.3%. (3) Overall, hazard and vulnerability have a greater impact on geological disaster ecological risk than exposure, with path coefficients of 0.802 (significant at p = 0.01 level) and 0.438 (significant at p = 0.05 level), respectively, in SEM. The random forest model (R² = 0.748) shows that social factors such as human density and road density contribute significantly more to extreme high risk than other factors, with a contribution rate of up to 44%. (4) Thirty-five ecological functional units were systematically grouped into four clusters and used to formulate a “layered” spatial policy for EFAs. The results of the research are expected to provide support for maximizing the policy impact of EFAs and formulating management decisions that serve ecological protection.