Search

Membership inference attack opens up a newly emerging and rapidly growing research to steal user privacy from text classification models, a core problem of which is shadow model construction and members distribution optimization in inadequate members. The textual semantic is likely disrupted by simple text augmentation techniques, which weakens the correlation between labels and texts and reduces the precision of member classification. Shadow models trained exclusively with cross-entropy loss have little differentiation in embeddings among various classes, which deviates from the distribution of target models, then impacts the embeddings of members and reduces the F1 score. A competitive and High-Availability Membership Inference Attack against Text Classification Model (HAMIATCM) is proposed. At the data level, by selecting highly significant words and applying text augmentation techniques such as replacement or deletion, we expand knowledge of attackers, preserving vulnerable members to enhance the sensitive member distribution. At the model level, constructing contrastive loss and adaptive boundary loss to amplify the distribution differences among various classes, dynamically optimize the boundaries of members, enhancing the text representation capability of the shadow model and the classification performance of the attack classifier. Experimental results demonstrate that HAMIATCM achieves new state-of-the-art, significantly reduces the false positive rate, and strengthens the capability of fitting the output distribution of the target model with less knowledge of members. [ABSTRACT FROM AUTHOR]

The alkylation of toluene catalyzed by modified HZSM-5 was an economical and efficient way to produce para-xylene (PX) due to the full utilization of toluene and fewer by-products. It has attracted significant attention in optimizing the modification method for both high activity and PX selectivity. In this study, we employed SiO2 and MgO to co-modify HZSM-5 for the alkylation of toluene with dimethyl carbonate. The catalysts in which SiO2 was deposited first with trimethoxysilane (TMOS) as the precursor, showed significantly higher activity and PX selectivity than that of depositing MgO first. When the content of SiO2 and MgO was both 5%, the conversion of toluene was 12%, with the selectivity of PX up to 85% (400 °C) for the 5Si-5Mg/ZSM-5 catalyst (depositing SiO2 first). However, the conversion and selectivity were only 5% and 56% for the 5Mg-5Si/ZSM-5 catalyst (depositing MgO first). The analysis of structure and acid properties indicated that the pre-deposition of SiO2 can regulate the location of MgO, leading to the suppression of the external acid sites and the inhibition of the pore blockage by MgO. Owing to the pore size of ZSM-5 being slightly smaller than the molecular size of TMOS, TMOS molecules would be embedded in the pore entrance, making SiO2 tend to prevent MgO from entering the micropore of ZSM-5. [ABSTRACT FROM AUTHOR]

Dissimilatory nitrate reduction processes, including denitrification, anaerobic ammonium oxidation (anammox), and dissimilatory nitrate reduction to ammonium (DNRA), are important pathways of nitrate transformation in the aquatic environments. In this study, we investigated potential rates of denitrification, anammox, and DNRA in the sediments of two subtropical rivers, Jinshui River and Qi River, with different intensities of human activities in their respective catchment, China. Our objectives were to assess the seasonality of dissimilatory nitrate reduction rates, quantify their respective contributions to nitrate reduction, and reveal the relationship between dissimilatory nitrate reduction rates, functional gene abundances, and physicochemicals in the river ecosystems. Our results showed higher rates of denitrification and anammox in the intensively disturbed areas in autumn and spring, and higher potential DNRA in the slightly disturbed areas in summer. Generally, denitrification, anammox, and DNRA were higher in summer, autumn, and spring, respectively. Relative contributions of nitrate reduction from denitrification, anammox, and DNRA were quite different in different seasons. Dissimilatory nitrate reduction rates and gene abundances correlated significantly with water temperature, dissolved organic carbon (DOC), sediment total organic carbon (SOC), NO3-, NH4+, DOC/NO3-, iron ions, and sulfide. Understanding dissimilatory nitrate reduction is essential for restoring nitrate reduction capacity and improving and sustaining ecohealth of the river ecosystems. [ABSTRACT FROM AUTHOR]

Background and aims: Agricultural abandonment has been taken as one of the most widely used strategies for ecological restoration. Abandoned croplands expanded worldwide since the middle of the twentieth century and will increase considerably in the future. Whether agricultural abandonment alters the abundances of soil carbon (C) and nitrogen (N) mineralization genes, and whether these functional genes are linked to alterations in C and N mineralization rates remain unknown. Methods: We took advantage of a well-established chronosequence of nature restoration sites on abandoned croplands which represent a century of vegetation succession in China's Qinling Mountains, to determine soil C and N mineralization rates, and their relationships with vegetation succession, edaphic characteristics, soil microbial communities, and functional genes based on GeoChip 5.0. Results: We found that soil C and N mineralization rates, microbial biomass and composition, and the number and abundances of functional genes were significantly affected by vegetation succession following agricultural abandonment. Soil C and N mineralization rates increased by sixfold and threefold, respectively, along the cropland to forest successional gradient. The NO3-N, NH4-N and soil available phosphorus (P) concentrations were important factors associated with the C and N mineralization genes. The shifts in edaphic environments (i.e., soil moisture, cation exchange capacity, and C/P ratio), vegetation-derived substrates (i.e., soil organic C and N), and soil microbial composition (i.e., Fungi/Bacteria ratio, amounts of Gram-negative bacteria, Actinomycetes, and arbuscular mycorrhizal fungi), rather than the abundances of functional genes involved in mineralization processes, controlled soil C and N mineralization rates along the long-term vegetation successional gradient. Conclusions: Abandoning cropland for natural succession could facilitate the recuperation of soil C and N processes and alter soil microbial composition and functional genes. The abundances of microbial functional genes are less important than previously expected in regulating soil C and N mineralization rates. [ABSTRACT FROM AUTHOR]