Search

Because the blockchain is secure and untamperable, it has been widely used in many industries, such as the financial industry, digital tokens, and e-commerce logistics. The remarkable security feature of the blockchain is that the blockchain verifies the transaction initiated on each block through the node, and its process is broadcast throughout the whole network to let everyone know. On the one hand, this ensures the security of every transaction, but on the other hand, it is easy to cause privacy disclosure problems for transaction users. Therefore, under the premise of ensuring the security of the blockchain, it has become a hot issue to protect the sensitive information of transaction users. A check-in privacy protection (CPP) algorithm based on check-in location generalization is proposed in this paper, which can be applied to blockchain transactions to solve the privacy leakage problem of transaction users’ sensitive information. CPP algorithm not only protects the privacy of check-in data but also keeps the high utility of trajectory pattern data. Firstly, location types are recommended in the sensitive check-in location generalization based on the user’s trajectory pattern by using Markov chain technology. Secondly, to make sure that the generalized locations can be scattered as much as possible to prevent the attacker from deducing back, a heuristic rule is designed to select the generalized location based on the recommended location types, and at the same time, the similarity between the anonymous trajectory and the original trajectory is maintained. In addition, a generalized location search strategy is designed to improve the efficiency of the algorithm. Based on the real spatial-temporal check-in data, the results of the experiment indicate that our algorithm can effectively protect the privacy of sensitive check-in while ensuring the high utility of trajectory pattern data.

Text classification is one of the most important and typical tasks in Natural Language Processing (NLP) which can be applied for many applications. Recently, deep learning approaches has shown their advantages in solving text classification problem, in which Convolutional Neural Network (CNN) is one of the most successful model in the field. In this paper, we propose a novel deep learning approach for categorizing text documents by using scope-based convolutional neural network. Different from window-based CNN, scope does not require the words that construct a local feature have to be contiguous. It can represent deeper local information of text data. We propose a large-scale scope-based convolutional neural network (LSS-CNN), which is based on scope convolution, aggregation optimization, and max pooling operation. Based on these techniques, we can gradually extract the most valuable local information of the text document. This paper also discusses how to effectively calculate the scope-based information and parallel training for large-scale datasets. Extensive experiments have been conducted on real datasets to compare our model with several state-of-the-art approaches. The experimental results show that LSS-CNN can achieve both effectiveness and good scalability on big text data.

Search engines use significant hardware and energy resources to process billions of user queries per day, where Boolean query processing for document retrieval is an essential ingredient. Considering the huge number of users and large scale of the network, traditional query processing mechanisms may not be applicable since they mostly depend on a centralized retrieval method. To remedy this issue, this paper proposes a processing technique for aggregated Boolean queries in the context of edge computing, where each sub-region of the network corresponds to an edge network regulated by an edge server, and the Boolean queries are evaluated in a distributed fashion on the edge servers. This decentralized query processing technique has demonstrated its efficiency and applicability for the document retrieval problem. Experimental results on two real-world datasets show that this technique achieves high query performance and outperforms the traditional centralized methods by 2–3 times.

To effectively manage and analyze graph data in many real-life graph applications, SCAN algorithm can efficiently help users to detect useful clusters, such as social networks, communication networks, gene networks, and so on. However, dirty data exist in graph data, for example, some edges are missing in graph data. In this case, the clustering results of SCAN over the dirty graph data cannot meet the users' requirements, such as, the missing nodes are not included in the desired clusters. To address this kind of problem, in this paper, we explore an effective explanation model to make the missing nodes be included in the desired clusters. The problem of explaining missing nodes in light of data modification is to explain why the missing nodes are not included in the desired clusters and how to make the missing nodes appear in the corresponding desired clusters by modifying the original graph dataset. To achieve this purpose, first, the clustering rational of SCAN algorithm is analyzed, and an unified explanation framework is proposed based on the analysis above. Moreover, based on the common explanation principle, the original clustering results should be maintained as much as possible in the new clustering results of SCAN, we design a penalty function to achieve this purpose. Then, we propose two explanation algorithms for making the missing nodes appear in the desired clusters by modifying the original graph dataset with minimum penalty value. Finally, comprehensive experiments are conducted, which demonstrate that the explored explanation model can efficiently explain the missing nodes on structural graph clustering.

In the applications of analyzing graph data, SCAN algorithm is an effective clustering algorithm for detecting meaningful clusters, which is widely used in many different graph applications. The problem of refining structural graph clustering parameters for unexpected node is to explain why an unexpected node is included in the specified cluster of the SCAN results and how to make the unexpected node disappear in the specified cluster. It is obvious that the SCAN results are very sensitive to the two clustering parameters, one is the similarity threshold $\epsilon$ , the other one is the density constraint $\mu$ , when they are input unreasonable, some unexpected nodes would be included in the specified clusters. To address this problem, how the parameters affect the scan results is analyzed firstly, then two effective refining algorithms for making the unexpected vertices disappear in the specified cluster are proposed, which optimize the initial SCAN parameters with minimum penalty from two aspects: one is to refine the parameter $\epsilon$ ; and the other is to refine the parameter $\mu$ . Moreover, to retain the original SCAN results as much as possible in the refined SCAN results, one penalty function is proposed. Finally, comprehensive experiments on real datasets show that our refining model can efficiently refine clustering parameters for the unexpected nodes of the clustering results of SCAN.

As an useful and important graph clustering algorithm for discovering meaningful clusters, SCAN has been used in a lot of different graph analysis applications, such as mining communities in social networks and detecting functional clusters of genes in computational biology. SCAN generates clusters in light of two parameters ∊ and µ. Due to the users lack the necessary professional knowledge, however, the parameter ∊ needs to be changed multiple times to obtain the desired clustering results. Every time the parameter ∊ changes, the new clustering result R’ can be obtained by executing the SCAN algorithm once, which takes a lot of time. To address this problem, in this paper, based on the previously clustering result R, we explore an effective incremental clustering strategy when the parameter ∊ changes dynamically to ∊‘. Although the SCAN results are affected by the parameters ∊ and µ, some vertices in $R$ are unaffected when the clustering parameter ∊ increases or decreases. Moreover, some useful information for clustering is constant, such as the structural neighborhood of any vertex, and the degree of any vertex. In this case, these information can be stored for use in the process of obtaining the new clustering result R’ with regard to the changed parameter ∊’ and the original parameter µ. Therefore, we explore two effective incremental clustering algorithms for the dynamically changing parameter ∊’, which avoids re-executing the SCAN algorithm based on the new parameter ∊’ and µ. Finally, we conduct comprehensive experimental studies, which illustrates that the incremental clustering model can effectively obtain the new clustering results when the clustering parameter ∊ changes dynamically.

Acquisitional issues widely exist in GPS and sensor networks. They pertain to where, when, and how often data is physically acquired (sampled) and delivered to some query processing systems. Due to the dynamic environment that the data is acquired with the change of monitoring time, acquisitional data are typically a time-stamped stream where a time-stamped value could either contain noises or be missing. Aiming to improve the quality of data acquisition, we focus on the explanations of missing values in this paper. Several techniques have been developed to provide the explanation on relational data, however, they cannot be directly applied in acquisitional stream data due to its dynamic feature, such as the “change” of acquisitional stream data between two adjacent monitoring time is often constrained by some rules. We show that an explanation could be incorrect or unreasonable if those constraints are not taken into account. We propose one novel chasing technique by considering both spatial and temporal correlations to explain missing values and guarantee a minimal explanation. Experimental results show that our approach can efficiently return high-quality and minimal explanations of missing values.

Ridesharing is becoming a popular mode of transportation with profound effects on the industry. We propose and investigate a novel carpooling with the common departure (CCD) problem that is to find optimal matching planning with minimal global travel cost. This problem applies to carpool among employees in large companies. This type of functionality holds the potential to bring significant benefits to society and the environment, such as reducing energy consumption and reducing traffic congestions. Although there are many ridesharing services for passengers and drivers, there is less research work on carpooling with common departure, especially when considering the global optimal aspects. To this end, a Three-level Matching Strategy for solving the CCD problem is established, including destination grid-based allocation, driving route and grid-based allocation, and passenger supplement strategy. The purpose is to reduce the detour distance traveled by the driver when serving passengers. Meanwhile, the probability of passengers traveling individual is reduced, so that the global travel cost is minimized. It can be seen that the Three-level Matching Strategy is an effective algorithm for the CCD problem. Finally, we conduct extensive experiments on real road networks and show that the Three-level Matching Strategy can save 66.7% of global travel costs compared with the non-matching method. Simultaneously, the average system response time was 2.65 min when the experimental data nodes density was 16% and show that our methods achieve satisfactory performance.

With the rapid development of mobile Internet and sharing economy, carpooling over real-time taxi-calling service has attracted more and more attention. Many popular taxi-calling service platforms, e.g. didi, Uber, have developed carpooling service to the passengers. Their goal is maximizing overall profit while meeting passengers’ convenience and economic benefits. This paper studies the problem of distributed tasks matching over road network. We first propose a random server selection algorithm to send tasks to different servers. Next, we propose a novel index to maintain tasks in each server. Last of all, we discuss how to execute tasks matching over road network. The results demonstrate the superior performance of our proposed algorithms.

Continuous query processing over sliding window is an important problem in stream data management. Given the sliding window W and the continuous query q, q monitors to the multidimensional data objects in the window. When the window slides, q returns all the skyline objects in the window. Many scholars have carried out researches on such problems. The core idea is to delete objects that cannot be query results by using the temporal sequence relationship between objects in the window, and when the window slides, the algorithm can find the query results from the rest. However, the algorithm is sensitive to data timing relationships such as problems. In the worst case, the size of the candidate object equals to the size of the data in the window. In this paper, we propose a partition-based framework to support continuous skyline query over sliding window. It partitions the window into a group of sub-window, and maintain the skyline objects in each sub-window. In this way, it could effectively overcome the impact the object arrived order to the algorithm performance. In addition, we propose a self-adaptive algorithm to partition the window according to the distribution of streaming data. A large number of experiments prove the effectiveness and high efficiency of the proposed algorithm.

After users publish the social data, the privacy information was available to an attacker through node re-identification. In real life, users have different privacy protection requirements. Due to the personalization of users' privacy needs, the protection technology of attacks with the same background knowledge is insufficient to meets users' privacy protection requirements. To solve the problem, this paper proposes spatio-temporal classification levels of social network data background knowledge, and provides corresponding protection methods for different background knowledge intensification of each level. Specifically, three levels of protection requirements are defined on the basis of the gradually increasing background knowledge of the spatio-temporal features of the attacker. The effectiveness of the framework is verified by a large number of experiments.