Search

Relation Extraction (RE) is to obtain a predefined relation type of two entities mentioned in a piece of text, e.g., a sentence-level or a document-level text. Most existing studies suffer from the noise in the text, and necessary pruning is of great importance. The conventional sentence-level RE task addresses this issue by a denoising method using the shortest dependency path to build a long-range semantic dependency between entity pairs. However, this kind of denoising method is scarce in document-level RE. In this work, we explicitly model a denoised document-level graph based on linguistic knowledge to capture various long-range semantic dependencies among entities. We first formalize a Syntactic Dependency Tree forest (SDT-forest) by introducing the syntax and discourse dependency relation. Then, the Steiner tree algorithm extracts a mention-level denoised graph, Steiner Graph (SG), removing linguistically irrelevant words from the SDT-forest. We then devise a slide residual attention to highlight word-level evidence on text and SG. Finally, the classification is established on the SG to infer the relations of entity pairs. We conduct extensive experiments on three public datasets. The results evidence that our method is beneficial to establish long-range semantic dependency and can improve the classification performance with longer texts.

With the rapid development of the Internet of Things(IoT), in the last decades, law enforcement agencies have deployed extensive sensor networks for public safety purposes. Diverse kinds of trajectories from the sensor networks provide an unprecedented opportunity for intelligence analysis. The geographic co-movement pattern has rarely been used by the police force to infer social connections, although it has been prevalent in other fields. The previous studies have mainly focused on a singular form of trajectories with exact co-locations, and the spread of the co-locations is over-looked. In this paper, we propose a novel framework for detecting co-occurrence patterns over multiplex trajectories. Firstly, We constructed the foundation for the discovery of co-occurrence events, namely space-time resonance honeycomb. It consists of multiple polygonal zones over sensor networks. Secondly, we transform all trajectories into a series of space-time prisms, and co-location activities are recorded using a sliding window approach. Thirdly, we propose a novel feature: Geo-Spread, which captures the extent of the co-location spread. In the end, we combine multiple features and employ Random Forest to predict social connections. We conduct extensive experiments on both the public dataset and the real-world surveillance dataset. Experiment results on all datasets prove the effectiveness of the proposed framework by outperforming the state-of-the-art methods.

Graph neural networks have been proven to be very effective for representation learning of knowledge graphs. Recent methods such as SACN and CompGCN, have achieved the most advanced results in knowledge graph completion. However, previous efforts mostly rely on localized first-order approximations of spectral graph convolutions or first-order neighborhoods, ignoring the abundant local structures like cycles and stars. Therefore, the diverse semantic information beneath these structures is not well-captured, leaving opportunities for better knowledge representation which will finally help KGC. In this work, we propose LSA-GAT, a graph attention network with a novel neighborhood aggregation strategy for knowledge graph completion. The model can take special local structures into account, and derive a sophisticated representation covering both the semantic and structural information. Moreover, the LSA-GAT model is combined with a CNN-based decoder to form an encoder-decoder framework with a carefully designed training process. The experimental results show significant improvement of the proposed LSA-GAT compared to current state-of-the-art methods on FB15k-237 and WN18RR datasets.

Authorized content is a type of content that can be generated only by a certain Content Provider (CP). The content copies delivered to a user may bring rewards to the CP if the content is adopted by the user. The overall reward obtained by the CP depends on the user’s degree of interest in the content and the user’s role in disseminating the content copies. Thus, to maximize the reward, the content provider is motivated to disseminate the authorized content to the most interested users. In this paper, we study how to effectively disseminate the authorized content in Interest-centric Opportunistic Social Networks (IOSNs) such that the reward is maximized. We first derive Social Connection Pattern (SCP) data to handle the challenging opportunistic connections in IOSNs and statistically analyze the interest distribution of the users contacted or connected. The SCP is used to predict the interests of possible contactors and connectors. Then, we propose our SCP-based Dissemination (SCPD) algorithm to calculate the optimum number of content copies to disseminate when two users meet. Our dataset based simulation shows that our SCPD algorithm is effective and efficient to disseminate the authorized content in IOSNs.

Person re-identification (re-id) is a cross-camera retrieval task which establishes a correspondence between images of a person from multiple cameras. Deep learning methods have been successfully applied to this problem and have achieved impressive results. However, these methods require a large amount of labeled training data. Currently, the labeled datasets in person re-id are limited in their scale and manual acquisition of such large-scale datasets from surveillance cameras is a tedious and labor-intensive task. In this paper, we propose a framework that performs intelligent data augmentation and assigns the partial smoothing label to generated data. Our approach first exploits the clustering property of existing person re-id datasets to create groups of similar objects that model cross-view variations. Each group is then used to generate realistic images through adversarial training. Our aim is to emphasize the feature similarity between generated samples and the original samples. Finally, we assign a non-uniform label distribution to the generated samples and define a regularized loss function for training. The proposed approach tackles two problems 1) how to efficiently use the generated data and 2) how to address the over-smoothness problem found in current regularization methods. The extensive experiments on four large-scale datasets show that our regularization method significantly improves the re-id accuracy compared to existing methods.

Mobile Crowd-Sensing (MCS) is a dominant sensing paradigm for Internet of Things (IoT), with a lot of potentials as it allows data collection through the user's sensor embedded mobile devices. The participation of people in IoT not only brings greater flexibility and sensing ability but also increases the risk of privacy breaches to the participants. Primarily, a worker's location data is vulnerable to information leaks as the task assignment in MCS is location-based. Most existing mechanisms that preserve worker's location in MCS are designed under the assumption that the platform is trusted, which may be not valid in real-world applications. Besides, the existing studies focus either on task selection problem for workers or task assignment problem for the platform. Therefore, this paper investigates both task bidding and assignment while preserving location privacy. We propose two task selection strategies: Minimize Total Cost (MTC) and Minimize Average Cost (MAC). Each worker submits a cost that is obfuscated using differential privacy to the platform. We propose probability cost-efficient worker selection mechanism (PCE-WSM) to determine winners and probability individual-rationality critical payment mechanism (PIR-CPM) to determine payments to winners. We prove that PIR-CPM is truthful and can achieve probability-individual rationality by theoretical analysis. To evaluate our proposed strategies, we conduct extensive experiments on both synthetic and real-world datasets, and the experimental results validate that PCE-WSM can achieve enough privacy preservation without incurring a high payment.

With the boom in the production of smartphones and easy access to networks, mobile crowd-sensing (MCS) is becoming a promising and rapidly growing sensing paradigm for the Internet of Things (IoT). In MCS, workers can get rewarded by participating in data collection while traveling to their destination from some starting point. However, the privacy leakage becomes a serious and unavoidable problem that hinders users' engagement. Taking task at a specific location or even going through some seemingly innocent locations may both disclose user's sensitive location information. Simply concealing or obfuscating the sensitive locations could preserve the privacy to a certain extent, at the cost of reward reduction, which will significantly frustrate the workers. To address these issues, this paper provides a novel framework for task selection in MCS jointly covering comprehensive location privacy preservation, efficient resource consumption, and high task profits. In our framework, workers can select their paths freely, and customize their privacy requirements for protection against both direct location disclosure and destination inference attack. Two corresponding algorithms are proposed for path selection under various stages of privacy preservation. The theoretical analysis demonstrates the performance of two algorithms on both privacy and utility aspects. We also evaluate our proposed mechanisms through extensive experiments on a real-world data set. The results validate that the proposed mechanisms can achieve higher task earnings and better privacy preservation.

Person re-identification aims to identify pedestrians across non-overlapping camera views. Deep learning methods have been successfully applied to solving the problem and have achieved impressive results. However, these methods rely either on feature extraction or metric learning alone ignoring the joint benefit and mutual complementary effects of the person view-specific representation. In this paper, we propose a multi-view deep network architecture coupled with n-pair loss (JNPL) to eliminate the complex view discrepancy and learn nonlinear mapping functions that are view-invariant. We show that the problem of the large variation in viewpoints of a pedestrian can be well solved using a multi-view network. We simultaneously exploit the complementary representation shared between views and propose an adaptive similarity loss function to better learn a similarity metric. In detail, we first extract view-invariant feature representation from n-pair of images using multi-stream CNN and then aggregate these features for predictions. Given n-positive pairs and a negative example, the network aggregate the feature map of the n-positive pairs and predicts the identity of the person and at the same time learns features that discriminate positive pairs against the negative sample. Extensive evaluations on three large scale datasets demonstrate the substantial advantages of our method over existing state-of-art methods.

Following the application of Deep Learning to graphic data, Graph Neural Networks (GNNs) have become the dominant method for Node Classification on graphs in recent years. To assign nodes with preset labels, most GNNs inherit the end-to-end way of Deep Learning in which node features are input to models while labels of pre-classified nodes are used for supervised learning. However, while these methods can make full use of node features and their associations, they treat labels separately and ignore the structural information of those labels. To utilize information on label structures, this paper proposes a method called 3ference that infers from references with differences. Specifically, 3ference predicts what label a node has according to the features of that node in concatenation with both features and labels of its relevant nodes. With the additional information on labels of relevant nodes, 3ference captures the transition pattern of labels between nodes, as subsequent analysis and visualization revealed. Experiments on a synthetic graph and seven real-world graphs proved that this knowledge about label associations helps 3ference to predict accurately with fewer parameters, fewer pre-classified nodes, and varying label patterns compared with GNNs.

Location prediction is the key technique in many location based services including route navigation, dining location recommendations, and traffic planning and control, to mention a few. This survey provides a comprehensive overview of location prediction, including basic definitions and concepts, algorithms, and applications. First, we introduce the types of trajectory data and related basic concepts. Then, we review existing location-prediction methods, ranging from temporal-pattern-based prediction to spatiotemporal-pattern-based prediction. We also discuss and analyze the advantages and disadvantages of these algorithms and briefly summarize current applications of location prediction in diverse fields. Finally, we identify the potential challenges and future research directions in location prediction.

Abstract Cyber-physical systems (CPSs) connect physical equipment and the information system to realize real-time perception and intelligent control of the equipment. With the rapid development of wireless sensors, wireless sensor networks (WSNs) now play a major role in CPS and have become a focus of research. However, WSNs in CPS face great challenges in ensuring the high reliability and time critical performance of data transmission because of the characteristics of the complex structure and dynamic connectivity. In order to meet the performance requirement of CPS data transmission, a Reliability and Timeliness Guaranteed Opportunistic Routing (RTGOR) protocol is proposed that is based on opportunistic routing and combined with quantified transmission reliability and time guarantees. Our simulation results demonstrate that RTGOR performs better than the state-of-the-art protocol QMOR (QoS-assured Opportunistic Routing Mechanism), in terms of the packet delivery rate and end-to-end delay metrics.

Location prediction has attracted increasing attention in diverse fields due to its wide applications, such as traffic planning and control, weather forecasting, homeland security, and travel recommendation. Many existing algorithms forecast a user's next location by learning that user's past moving patterns. However, the individual moving patterns in many practical applications (e.g., the moving trajectory of a taxi driver) tend to be random, which poses a big challenge for location prediction. In this paper, we propose a new robust location prediction model that considers both individual preferences and social interactions (PSI) at a group level to alleviate the effect of randomness and improve the location prediction performance. Specifically, we first extract hot places of interesting (POIs) and normal POIs, respectively, via a two-stage clustering approach. To characterize exterior social interactions, an associated group is identified, and an outline of group moving patterns is then extracted based on association rule mining. Finally, the next location is predicted by learning the individual's regular patterns and group moving patterns via a pair-wise ridge regression. In contrast to the traditional approaches, our proposed algorithm has several desirable characteristics: 1) PSI provides an intuitive and quantitative way to model human movement from two aspects: the individual's internal moving preferences and group-level exterior social interactions; 2) Building upon group-level pattern mining, PSI provides a more robust prediction model by learning both individual and group trend information simultaneously, alleviating the randomness of location prediction from individual historical trajectory data only; and 3) The experimental results demonstrate that PSI achieves a better prediction performance compared to the state-of-the-art methods.

Abstract Wireless sensor and actor networks are becoming more and more popular in the recent years. Each WSAN consists of numerous sensors and a few actors working collaboratively to carry out specific tasks. Unfortunately, actors are prone to failure due to harsh deployment environments and constrained power, which may break network connectivity resulting in disjoint components. Thus, maintaining the connectivity among actors is especially important. This paper proposes hybrid connectivity restoration (HCR), which integrates proactive selection and reactive motion. An actor protectively selects a backup node through its one-hop neighbor table and informs the backup node to supervise its stage. Once it fails, the backup node moves to the best position to restore the connectivity of the failed node’s neighbors reactively. This triggers a local recovery process at the backup node, which is repeated until network connectivity is restored. In order to minimize travel distance, HCR selects the backup node which moves the shortest distance to restore connectivity. Furthermore, HCR opts to reduce the number of messages by just informing the failure to its backup node. The correctness and effectiveness of HCR are validated through both theoretical analysis and simulations.

Location-based services (LBS) leveraged by ubiquitous mobile devices have brought great convenience to mobile users in various aspects, including communication, information exchange, social activities, and so on. However, privacy concerns arise at the same time, since the users need to submit their locations and query contents to the LBS servers. To this end, location privacy and query privacy have been recognized. In particular, this paper focuses on query privacy for preventing the leakage of users' query contents. Cloaking region-based techniques using untrusted servers and client-based k-anonymity approaches have been devised to preserve query privacy in LBS. However, these works suffer from single point of failure or insufficiency of query privacy. To address this issue, we investigate effective k-anonymity-based solutions for query privacy in LBS. We formulate a probabilistic framework PkA, under which k-anonymity-based mechanisms can be initiated, and analyze a recent proposed algorithm DLS as an instance of PkA. An algorithm circle segment is presented to provide effective query privacy when query interests have similar prior probability. To obtain more effective query privacy in general cases, we propose two algorithms MEE and MER, which optimize two individual privacy metrics, denoted expected entropy and expected max-min ratio, adopted in this paper. We recognize two practical properties - No More Leakage and k-Effectiveness for effective query privacy, and our proposed algorithms satisfy both of No More Leakage and k-Effectiveness. We conduct evaluation based on real-life data sets and synthetic distributions of query interests, and the evaluation results demonstrate that our proposed algorithms produce significantly improved query privacy.

Sensor drift is a common issue in E-Nose systems and various drift compensation methods have received fruitful results in recent years. Although the accuracy for recognizing diverse gases under drift conditions has been largely enhanced, few of these methods considered online processing scenarios. In this paper, we focus on building online drift compensation model by transforming two domain adaptation based methods into their online learning versions, which allow the recognition models to adapt to the changes of sensor responses in a time-efficient manner without losing the high accuracy. Experimental results using three different settings confirm that the proposed methods save large processing time when compared with their offline versions, and outperform other drift compensation methods in recognition accuracy.

This paper mainly deals with how kernel method can be used for software defect prediction, since the class imbalance can greatly reduce the performance of defect prediction. In this paper, two classifiers, namely, the asymmetric kernel partial least squares classifier (AKPLSC) and asymmetric kernel principal component analysis classifier (AKPCAC), are proposed for solving the class imbalance problem. This is achieved by applying kernel function to the asymmetric partial least squares classifier and asymmetric principal component analysis classifier, respectively. The kernel function used for the two classifiers is Gaussian function. Experiments conducted on NASA and SOFTLAB data sets using F-measure, Friedman’s test, and Tukey’s test confirm the validity of our methods.

Searchable encryption technique enables the users to securely store and search their documents over the remote semitrusted server, which is especially suitable for protecting sensitive data in the cloud. However, various settings (based on symmetric or asymmetric encryption) and functionalities (ranked keyword query, range query, phrase query, etc.) are often realized by different methods with different searchable structures that are generally not compatible with each other, which limits the scope of application and hinders the functional extensions. We prove that asymmetric searchable structure could be converted to symmetric structure, and functions could be modeled separately apart from the core searchable structure. Based on this observation, we propose a layered searchable encryption (LSE) scheme, which provides compatibility, flexibility, and security for various settings and functionalities. In this scheme, the outputs of the core searchable component based on either symmetric or asymmetric setting are converted to some uniform mappings, which are then transmitted to loosely coupled functional components to further filter the results. In such a way, all functional components could directly support both symmetric and asymmetric settings. Based on LSE, we propose two representative and novel constructions for ranked keyword query (previously only available in symmetric scheme) and range query (previously only available in asymmetric scheme).

For both convenience and security, more and more users encrypt their sensitive data before outsourcing it to a third party such as cloud storage service. However, searching for the desired documents becomes problematic since it is costly to download and decrypt each possibly needed document to check if it contains the desired content. An informative query-biased preview feature, as applied in modern search engine, could help the users to learn about the content without downloading the entire document. However, when the data are encrypted, securely extracting a keyword-in-context snippet from the data as a preview becomes a challenge. Based on private information retrieval protocol and the core concept of searchable encryption, we propose a single-server and two-round solution to securely obtain a query-biased snippet over the encrypted data from the server. We achieve this novel result by making a document (plaintext) previewable under any cryptosystem and constructing a secure index to support dynamic computation for a best matched snippet when queried by some keywords. For each document, the scheme has O(d) storage complexity and O(log(d/s)+s+d/s) communication complexity, where d is the document size and s is the snippet length.