Search

The E-CLSC (E-closed-loop supply chain) game model dominated by manufacturer is set, and information value about asymmetry fairness concern of E-platform (E-commerce platform) is calculated for manufacturer, recycler, and E-platform. By signaling model under various signal costs, the authors study the condition for E-platform to transmit real information about fairness concern so as to reduce profit loss for all parties in E-CLSC. The authors prove that E-platform has the motivation to disguise or exaggerate fairness concerns in order to obtain more profit, and manufacturers must try to identify the E-platform's real fairness concern to avoid profit loss. Besides, only when different types of E-platforms need significantly different signaling cost, both of them would like to send real fairness-concern signal, and thus manufacturer can effectively identify E-platform's real information about fairness concern so as to improve recycling rate and optimize the whole E-CLSC operation.

Electromechanical coupling plays a key role in determining the performance of stretchable strain sensor. Current regulation of the electromechanical coupling in stretchable strain sensor is largely restricted by the intrinsic mechanical properties of the device. In this study, a microfluidic strain sensor based on the core-shell package design with the auxetic metamaterial (AM) is presented. By overriding the mechanical properties of the device, the AM in the package effectively tunes the deformation of the microfluidic channel with the applied strain and configures the directional strain sensitivity with a large modulation range. The gauge factor (GF) of the strain sensor in the radial direction of the channel can be gradually shifted from the intrinsically negative value to a positive one by adopting the AMs with different designs. By simply replacing the AM in the package, the microfluidic strain sensor with the core-shell package can be configurated as an omnidirectional or directional stretchable strain sensor. With the directional sensitivity brought by the rational AM design, the application of the AM-integrated strain sensor in the skin-mounted tactile detection is demonstrated with high tolerance to unintended wrist movements.

The genus Hydnellum is a kind of ectomycorrhizal fungi that can play a role in the material cycle by connecting the plant roots to the soil, and some species of Hydnellum are medicinal fungi with vital research value. The species diversity of Hydnellum is unclear in China. In this study, five new species of Hydnellum are described from China based on morphological characters and phylogenetic analyses inferred from two datasets of ITS + LSU and ITS + LSU + SSU + RPB2 sequences. H. chocolatum is characterized by its chocolate basidiomata with the fibrillose, spongy to tomentose pileal surface, and subglobose to globose basidiospores measuring (4.5–)5–6 × 4–5(–5.8) μm. H. concentricum is characterized by its zonate pileal surface, thin context, short stipe, presence of both simple septa and clamp connections in generative hyphae of spines, and subglobose to ellipsoidal basidiospores measuring (3.5–)4–5(–5.2) × (3.2–)3.5–5 μm. H. crassipileatum is characterized by its thick pileus with the reddish brown to grayish brown pileal surface, and subglobose to ellipsoidal basidiospores measuring 4–6(–6.5) × 4–5.5 μm. H. melanocarpum is characterized by its vinaceous brown to black pileus with spongy pileal surface, presence of both simple septa and clamp connections in generative hyphae of spines, and subglobose basidiospores measuring 4.5–5.5(–6) × (3.5–)3.8–5.1 μm. H. radiatum is characterized by its radially aligned stripes on pileal surface, grayish brown context, short stipe, and subglobose to ellipsoidal basidiospores measuring (3.5–)4–5 × 3–4.5(–5) μm. Full descriptions, illustrations, and phylogenetic trees to show the placement of the new species are provided.

The genus

Isobaric heat capacity is one of the key thermophysical properties for working fluids in thermal systems and plays an important role in the development of equation of state. Due to the lack of experimental data for promising hydrofluoroolefins (HFOs), we carried out theoretical predictions for condensed liquid phase with three different methods: traditional cubic equation of state (CEOS), fundamental equations of state (FEOS) explicit in Helmholtz free energy and a corresponding state equation (CSE). Both CEOS and CSE are generalized models that only need several characteristic parameters to conduct calculation, while FEOS are specific models in which the structure and parameters are regressed from experimental data of one certain fluid. Liquid heat capacity data of 9 HFOs were calculated, including well-known R1234yf, R1234ze(E), and latest prospective R1123, R1216, R1243zf, R1234ze(Z), R1336mzz(Z), R1141, and R1125zc. A critical comparison was carried out between different calculation methods. The comparison showed that both CSE and FEOS predicted available experimental data well with AADs % less than the reported experimental uncertainties. For refrigerants without experimental heat capacity, CSE and FEOS showed good agreements with AADs % less than 3.6 %. The selected CEOS always provided relatively larger predictions than experiments or the other two methods. Despite that, calculated data of CEOS showed a strong linear relation with the other two methods, which suggested that there might be a potential linear modification or correlation for CEOS in improving its heat capacity calculations. Furthermore, when there were no experimental data available, CSE used in this study could be a useful preliminary evaluation tool for the liquid heat capacity of hydrofluoroolefins, considering its accuracy and easy application characteristics.

Android Application Framework is an integral and foundational part of the Android system. Each of the two billion (as of 2017) Android devices relies on the system services of Android Framework to manage applications and system resources. Given its critical role, a vulnerability in the framework can be exploited to launch large-scale cyber attacks and cause severe harms to user security and privacy. Recently, many vulnerabilities in Android Framework were exposed, showing that it is indeed vulnerable and exploitable. While there is a large body of studies on Android application analysis, research on Android Framework analysis is very limited. In particular, to our knowledge, there is no prior work that investigates how to enable symbolic execution of the framework, an approach that has proven to be very powerful for vulnerability discovery and exploit generation. We design and build the first system, Centaur , that enables symbolic execution of Android Framework. Due to the middleware nature and technical peculiarities of the framework that impinge on the analysis, many unique challenges arise and are addressed in Centaur . The system has been applied to discovering new vulnerability instances, which can be exploited by recently uncovered attacks against the framework, and to generating PoC exploits.

Calcium chloride (CaCl2) aqueous solution as a liquid desiccant has the disadvantage of low vapor pressure and easy crystallization. Ionic liquids (ILs) such as [Emim]Cl, [Emim]Ac, and [Emim]Br hav...

The application of the excess entropy scaling (EES) method to predict the viscosity, thermal conductivity and thermal diffusivity of HFC/HFO refrigerants is evaluated in this paper. The universal coefficients of the EES model were firstly obtained from the properties of HFC refrigerants, and the accuracy of the model was further investigated with HFO properties. It was suggested that the EES model correlated the viscosity very well with the average absolute deviations (AADs) of most HFC refrigerants lower than 6.55% except R32. The correlations also provided very good prediction on the viscosity for R1234yf and R1234ze(E), but not for R1336mzz(Z). The prediction of thermal conductivity for both HFC and HFO refrigerants was generally well with the maximum AAD of 11.44%. However, the paper also indicated that there were no universal thermal diffusivity coefficients for even HFC refrigerants, and the linear function could not fit the thermal diffusivity curve very well. Therefore, the exclusively two-order polynomial correlations based on the EES model were presented for each HFC/HFO refrigerant.

Ejector refrigeration cycle (ERC) with advantages of simple structure and low cost holds great application potential in cascade/hybrid cycles to improve the overall system performance by removing or recovering the heat from the main cycle. In this paper, a theoretical and experimental investigation of the ERC as a part of a cascade system was carried out. The operating parameters were optimized. The experimental ERC test rig was designed, developed and investigated at high evaporating temperatures and wide ranges of operating conditions. The influence of operating conditions on the efficiency of the ejector and ERC was analyzed. Experimental results and analysis in this study can be helpful for the application and operating condition optimization of ERC in cascade/hybrid refrigeration systems.

Xinjiang Uyghur Autonomous Region in China embraces a unique geographical and ecological environment, and the macrofungi represent a rich resource. However, few studies on the genus

Perenniporia is an important genus of Polyporaceae. In its common acceptation, however, the genus is polyphyletic. In this study, phylogenetic analyses on a set of Perenniporia species and related genera were carried out using DNA sequences of multiple loci, including the internal transcribed spacer (ITS) regions, the large subunit nuclear ribosomal RNA gene (nLSU), the small subunit mitochondrial rRNA gene (mtSSU), the translation elongation factor 1-α gene (TEF1) and the b-tubulin gene (TBB1). Based on morphology and phylogeny, 15 new genera, viz., Aurantioporia, Citrinoporia, Cystidioporia, Dendroporia, Luteoperenniporia, Macroporia, Macrosporia, Minoporus, Neoporia, Niveoporia, Rhizoperenniporia, Tropicoporia, Truncatoporia, Vanderbyliella, and Xanthoperenniporia, are proposed; 2 new species, Luteoperenniporia australiensis and Niveoporia subrusseimarginata, are described; and 37 new combinations are proposed. Illustrated descriptions of the new species are provided. Identification keys to Perenniporia and its related genera and keys to the species of these genera are provided.

Xinjiang Uyghur Autonomous Region in China embraces a unique geographical and ecological environment, and the macrofungi represent a rich resource. However, few studies on the genus Pluteus have been reported from Xinjiang. In 2021, the macrofungal resources in Xinjiang were surveyed, and 10 specimens belonging to the genus Pluteus were collected. Based on the morphological study and molecular analysis, three species were recognized, P. aletaiensis, P. brunneidiscus, and P. hongoi. Pluteus aletaiensis is proposed as a new species. It is characterized by its bright yellow lamellae and stipe, brittle texture, subfusiform to vesicular pleurocystidia, with short pedicels to broadly lageniform to obtuse at apices, a hymeniderm pileipellis, containing dark brown intracellular pigment, and it grows on the ground. Pluteus brunneidiscus, a new record to China, is characterized by uneven, smooth, grayish brown to brown pileus, with an entire margin, and pointed or flatter apices intermediate cystidia, without apical hooks. Pluteus hongoi, a new record to Xinjiang Uyghur Autonomous Region, China, is characterized by the apical hook’s structure (commonly bifid) of pleurocystidia. The nuclear internal transcribed spacer (nrITS) and translation elongation factor 1-alpha (TEF1-a) region were used for the molecular analysis. Phylogenetic trees were constructed using both the maximum likelihood analysis (ML) and Bayesian inference (BI). Detailed descriptions of the three species are presented herein. Finally, a key to the list of eight species of the genus Pluteus knew from Xinjiang is provided.

Graph Neural Networks (GNNs) has been widely used to address the sparsity and cold start problems in recommendation system. By propagating embeddings from multi-hop neighbor nodes among the interaction graph and update target user and item embeddings, GNNs-based methods can achieve better recommendation performance. But those methods directly concatenate the output of each layer and ignore the different influences between different layers, and they simply use the inner product of the user and item’s embeddings to calculate the similarity and make recommendation based on it, which is insufficient to reveal the complex and nonlinear interactions.In this work, we propose to learn multi-order interactions between users and items and capture correlations between different-order information. We design a new recommendation framework MCCR, which treats each layer’s output as differentorder feature, and propose a multi-order interaction module to represent feature interactions. We adopt a multi-layer 3D CNN module to learn high-order interaction signals between users and items in an explicit approach. Through extensive experiments on three real-world datasets, which shows that MCCR evidently outperforms the state-of-the-art methods consistently.

Ionic liquids (ILs), 1-ethyl-3-methylimidazolium iodine [Emim]I and 1,3-dimethylimidazolium iodine [Dmim]I, were proposed as a novel potential solvent in traditional bromide aqueous solution (H2O + LiBr) in absorption refrigeration cycles. The vapor pressure data of two ternary systems H2O + LiBr + [Emim]I (mass ratio LiBr/[Emim]I = 4) at temperatures from 350.78 to 414.19 K and H2O + LiBr + [Dmim]I (mass ratio LiBr/[Dmim]I = 2) at temperatures from 350.69 to 409.63 K were reported by the boiling point method. The absorbent mass fractions are from 0.30 to 0.57 and 0.30 to 0.60, respectively. The measured data were correlated by using the Antoine-type equation, and the average absolute relative deviation (AARD %) between measured data and regression data was 0.54 and 0.39%. Compared with the vapor pressures of IL mixtures with the same organic cation, the vapor pressure order is LiBr + [Emim]I + H2O < LiBr + [Emim]Cl + H2O < LiBr + [Emim]Ac + H2O and LiBr + [Dmim]I + H2O < LiBr + [Dmim]Cl + H2O < LiBr + [D...

Since the master key length becomes longer and longer in ciphers, an adversary often needs to preform the multi-round side channel analysis (SCA) in order to recover the master key by enough round keys. Traditional multi-round SCA is launched by adaptive manner in practice, which means that the input of each round is calculated in an on-the-fly way based on all round keys of anterior rounds. However, compared to the classical single-round SCA, the multi-round SCA in adaptive manner is severely limited in several practical scenarios, because all round keys of anterior rounds must be properly recovered before the attack against the next round. In this paper, we focus on the Feistel-SP networks, break the interdependency between the alternating measurement and analysis phases, propose a Multi-round non-Adaptive Chosen Message (MACM) approach, which can reduce the multi-round SCA to the single-round attack. In MACM, the set of plaintexts applied to multiple rounds is calculated in an off-line way. We also prove that the revealed round keys by MACM are adequate to recover the master key. Furthermore, we carefully analyze the advantages of MACM regarding to robustness and compatibility. In order to further manifest the validity of MACM, we perform extensive experiments on three typical Feistel-SP ciphers, Camellia, CLEFIA and SM4, the master keys are recovered as expected, and the number of traces in MACM is at least 25% less than that in the adaptive manner.

This paper presents a novel triple heat sources-driven absorption heat pump system. The exhaust gas is used as the heat source for a high-pressure generator. The superheated refrigerant vapor generated by the high-pressure generator is used for heating the solution in the low-pressure generator 1 by releasing the latent and sensible heat. The low-temperature solar thermal energy serves as the heat source of the low-pressure generator 2. The geothermal energy supplies heat to the evaporator. Heat discharged from the condenser and the absorber is supplied to the air conditioner heat medium water. The high- and low-pressure generators are coupled to each other through an ejector. The introduction of the ejector reduces the pressure of the low pressure generator 2 below the condensing pressure so that the lower temperature solar thermal energy can be utilized. An experimental setup for this heat pump cycle is built, and an experimental investigation is carried out. The influences of the independent variables on the key internal operating parameters and heating performance have been investigated. Experimental studies show that when the hot air temperature is 213 °C and the temperature of the hot water is 43.5 °C, a heating capacity of 26.6 kW can be achieved with a heating performance coefficient greater than 1.7. In this way, an efficient and full use of waste heat is realized, and the results can provide an important basis for the design and wide application of low-temperature solar thermal and geothermal energy.

Embedding knowledge graphs (KGs) into continuous vector space is an essential problem in knowledge extraction. Current models continue to improve embedding by focusing on discriminating relation-specific information from entities with increasingly complex feature engineering. We noted that they ignored the inherent relevance between relations and tried to learn unique discriminate parameter set for each relation. Thus, these models potentially suffer from high time complexity and large parameters, preventing them from efficiently applying on real-world KGs. In this paper, we follow the thought of parameter sharing to simultaneously learn more expressive features, reduce parameters and avoid complex feature engineering. Based on gate structure from LSTM, we propose a novel model TransGate and develop shared discriminate mechanism, resulting in almost same space complexity as indiscriminate models. Furthermore, to develop a more effective and scalable model, we reconstruct the gate with weight vectors making our method has comparative time complexity against indiscriminate model. We conduct extensive experiments on link prediction and triplets classification. Experiments show that TransGate not only outperforms state-of-art baselines, but also reduces parameters greatly. For example, TransGate outperforms ConvE and RGCN with 6x and 17x fewer parameters, respectively. These results indicate that parameter sharing is a superior way to further optimize embedding and TransGate finds a better trade-off between complexity and expressivity.

The nanofluid has been widely used in many heat transfer areas due to its significant enhancement effect on the thermal conductivity. Therefore, the methods that can accurately predict their thermal conductivities are very important to evaluate and analyze the heat transfer process. In this paper, a novel artificial neural network (ANN) model was proposed to predict the thermal conductivity of nanofluids with ethylene glycol and could be used in a wide range with excellent accuracy. A total of 391 experimental data with a wide range of temperatures (4°C~90°C), nanoparticles (metal, metal oxide, etc.), volume concentrations (0.05%~10%), and particle sizes (2 nm ~ 282 nm) were collected. To build the ANN model, the temperature, thermal conductivities of the base fluid and nanoparticles, the size and volume concentration of the nanoparticles were selected and used as the input parameters. There were 5 nodes, 10 nodes and 1 node in input layer, hidden layer and output layer, respectively. The predicted results of the ANN model coincided with the experimental data very well with the correlation coefficient and mean square error (MSE) were 0.9863 and 3.01×10-5, respectively. The relative deviations of 99.74% data were within ±5%. The model was expected to be a good practical method to predict the thermal conductivity of nanofluids with ethylene glycol.

1-Ethyl-3-methylimidazolium chlorine ([Emim]Cl) and 1-ethyl-3-methylimidazolium bromide ([Emim]Br) as an additive to working fluid [lithium bromide aqueous solution (H2O/LiBr) and lithium chloride aqueous solution (H2O/LiCl)] of absorption cycles were proposed. The vapor pressure of H2O + LiCl + [Emim]Cl (mass ratio LiCl/[Emim]Cl = 2) in the temperature from 314.79 to 435.48 K, H2O + LiCl + [Emim]Br (mass ratio LiCl/[Emim]Br = 2) in the temperature from 317.62 to 431.27 K, and H2O + LiBr + [Emim]Cl (mass ratio LiBr/[Emim]Cl = 2) in the temperature from 303.10 to 409.91 K was measured by the boiling point method. The mass fractions of the absorbent were from 0.30 to 0.60. The experimental data were regressed using the Antoine-type equation, and the average absolute relative deviation (AARD %) between the experimental data and calculated values of three systems was 0.44, 0.51, and 0.54%. Compared with that of the previously reported ionic liquids, ([Emim]Ac) as an absorbent for LiBr + H2O, vapor pressure of...

In this work, a feed-forward artificial neural network (ANN) was developed for the calculation of isobaric heat capacity of pure HFC and HFO refrigerants in liquid phase. First of all, a total of 1142 available experimental data points for different pure HFC or HFO refrigerants were collected and evaluated before being used to train and test the network. By a trial-and-error method, optimum structural of the network was found out to be an input layer using four dimensionless input parameters, one hidden layer with 34 neurons, and an output layer with reduced residual heat capacity as output. The ANN was applied for 12 different refrigerants, and the predicted isobaric heat capacity showed satisfactory agreement with experimental data. The overall average absolute deviation (AAD %) and maximum absolute deviation (MAD %) were 0.383% and 5.92% respectively.

It is very important to develop a reliable method for the application of the pulsating heat pipe (PHP), however, the currently proposed heat transfer correlations or theoretical models still have some apparent shortcomings, such as great deviation and poor flexibility. Considering the advantages of artificial neural network (ANN) in analyzing complex systems, a fully connected feed forward ANN model was used to predict the thermal resistance of a closed vertical meandering PHP with water. The Number of turns, filling ratio, heat flux, inner diameter and the length ratio of evaporation section were selected as the input parameters. By applying a trial-and-error method, the neuron number in the hidden layer was optimized to be 10. A total of 221 points of experimental data under different working conditions collected from the published literature were applied to build the ANN model. The results showed a good agreement between the experimental data and the ANN model with the MSE and correlation coefficient of 0.0025 and 0.9962, respectively. Furthermore, the influence of the heat flux on the relative deviation was also investigated. It suggested that the ANN model could have better prediction results when the heat flux was within the range of 6500–14,500 W/m2.

The blooming of live comment videos leads to the need of automatic live video comment generating task. Previous works focus on autoregressive live video comments generation and can only generate comments by giving the first word of the target comment. However, in some scenes, users need to generate comments by their given prompt keywords, which can't be solved by the traditional live video comment generation methods. In this paper, we propose a Transformer based non-autoregressive conditional masked live video comments generation model called CMVCG model. Our model considers not only the visual and textual context of the comments, but also time and color information. To predict the position of the given prompt keywords, we also introduce a keywords position predicting module. By leveraging the conditional masked language model, our model achieves non-autoregressive live video comment generation. Furthermore, we collect and introduce a large-scale real-world live video comment dataset called Bili-22 dataset. We evaluate our model in two live comment datasets and the experiment results present that our model outperforms the state-of-the-art models in most of the metrics.

Obtaining demographic attributes of online users is of great significance for retail marketing, targeted advertisement and many other scenarios. Users' wanderings on various websites and applications contains user preference on different items, and can be leveraged to infer one's private attributes. Existing studies usually focus on manually defined features, relationships in online social networks, or modeling global user preferences. However, attribute inference from the most common behavioral data (e.g., browsing history, shopping cart) is recently overlooked, and still requires further research. In this work, we propose a Topic-Enhanced Attentive Recurrent Neural Network (TEA-RNN) model to capture both local neighborhood-based features (with attentive RNN) and global patterns (with topic model) within user behaviors, and apply multi-task learning mechanism with weighted losses to further leverage the latent relationships within demographics. Experimental results on real-world datasets demonstrates the effectiveness of TEA-RNN by comparing with several commonly used baselines.

More and more work has focused on incorporating different kinds of literals into Knowledge Graph to promote the performance of knowledge embedding. These literals contain numeric literals, text literals, image literals and so on. These additional descriptions are connected to the entities through certain attributes. To incorporate numeric literals, some methods combine the embeddings of literals part with the traditional part - embeddings of entities. However, in the construction of literals embeddings, these existing methods consider the differences of these attributes: one dimension represents one attribute. But they ignore semantic meanings of attributes themselves. In this paper, we propose two methods to incorporate attributes semantics into knowledge graph embeddings from two perspectives: LiteralEAN and literalE-AT. They concatenate with the embeddings of numeric literals by different ways. Furthermore, their extension model LiteralE-C is also proposed as having a more comprehensive representation of attributes semantics. In an empirical study over two standard datasets FB15k and FB15k-237, we evaluate our models for link prediction. We demonstrate that they show an effective way to improve LiteralE and achieve state-of-the-art results. In ablation experiments, we find combined models do better than their singular counterparts in most cases.

In recent years, audiences can find their interested TV play or movie videos by labels easily. However, for finding shots with certain semantic content in these videos, it is still a problem to annotate video shots by labels. Some existing approaches train models with annotated shots which cost a lot in labeling manually. Some other methods in solving this kind of task assume that the content of a video is only limited in the labels of the video. They ignore that the labels of a video are too coarse-grained to cover all content of the video. In this paper, we propose a multi-label, multi-instance topic model to annotate video shots by video labels. In a multi-label, multi-instance framework, video shots can be regarded as instances and shot labels are learned from labels in video level which makes the cost of labeling cheaper. On the other hand, our model learns label semantics by controlling the relationship between video labels and shots to solve coarse-grained problem. Furthermore, we also learn keywords for every video. The experiments on a large-scale real-world dataset show that our model outperforms other baseline models substantially.

In recent years, attribute-based access control (ABAC) is more and more popular because of its flexibility and fine-grained data management. However, manually configuring authorization policies in ABAC system is a time-consuming, labor-intensive, and tedious work. Many researchers explore the ways of automatically configuring authorization policies by parsing requirement specifications that are expressed in natural language. Previous works only focus on English and ignore the semantics of comparative relationship. In this paper, we propose a method based on Chinese including procedures of key words extraction, tag alignment and expression transformation. It can parse Chinese sentence into constraint expressions and authorization sign, by which ABAC system is able to configure authorization policies automatically. Our evaluation results show that it has good performance not only in the independent tests on each procedure but also in the systematic tests on the whole method.

Live video comment generating task aims to automatically generate real-time viewer comments on videos like real viewers do. Like providing search suggestions by search engines, this task can help viewers find comments they want to post by providing generated comments. Previous works ignore the interactivity and diversity of comments and can only generate general and popular comments. In this paper, we incorporate post time of the comments to deal with the real-time related comment interactions. We also take the video type labels into consideration to handle the diversity of comments and generate more related and informative comments. To this end, we propose a pre-training based encoder-decoder joint model called PLVCG model. This model is composed of a bidirectional encoder to encode context comments and visual frames jointly as well as an auto-regressive decoder to generate real-time comments and classify the type of the video. We evaluate our model in a large-scale real-world live comment dataset. The experiment results present that our model outperforms the state-of-the-art on live video comment ranking and generating task significantly.

As a typical application of decentralization, federated learning prevents privacy leakage of crowdsourcing data for various training tasks. Instead of transmitting actual data, federated learning only updates model parameters of server by learning multiple sub-models from clients. However, these parameters may be leaked during transmission and further used by attackers to restore client data. Existing technologies used to protect parameters from privacy leakage do not achieve the sufficient protection of parameter information. In this paper, we propose a novel and efficient privacy protection method, which perturbs the privacy information contained in the parameters and completes its ciphertext representation in transmission. Regarding to the perturbation part, differential privacy is utilized to perturb the real parameters, which can minimize the privacy information contained in the parameters. To further camouflage the parameters, the weak encryption keeps the ciphertext form of the parameters as they are transmitted from the client to the server. As a result, neither the server nor any middle attacker can obtain the real information of the parameter directly. The experiments show that our method effectively resists attacks from both malicious clients and malicious server.

Detecting local topics from social media is an important task for many applications, ranging from event tracking to emergency warning. Recent years have witnessed growing interest in leveraging multi-modal social media information for local topic detection. However, existing methods suffer great limitation in capturing comprehensive semantics from social media and fall short in bridging semantic gaps among multi-modal contents, i.e., some of them overlook visual information which contains rich semantics, others neglect indirect semantic correlation among multi-modal information. To deal with above problems, we propose an effective local topic detection method with two major modules, called IEMM-LTD. The first module is an image-enhanced multi-modal embedding learner to generate embeddings for words and images, which can capture comprehensive semantics and preserve both direct and indirect semantic correlations. The second module is an embedding based topic model to detect local topics represented by both words and images, which adopts different prior distributions to model multi-modal information separately and can find the number of topics automatically. We evaluate the effectiveness of IEMM-LTD on two real-world tweet datasets, the experimental results show that IEMM-LTD has achieved the best performance compared to the existing state-of-the-art methods.