Abstract:Volume data classification is a core issue of transfer function in volume rendering. Scalar-gradient magnitude histogram of volume is a classic feature space, and has been applied in volume classification for its nice result in visual extraction of boundaries between different materials. However, the design of transfer function based on scalar-gradient histogram has proven as a time-consuming and complex task which is hard for users to conduct interactions. In this paper, scalar-gradient histogram is treated as a density distribution of all voxels. This approach assumes that the density of a material center is higher than their neighbors and the distance between two material centers is far enough. By computing the minimum distance between each points and all other points with higher density in scalar-gradient histogram, a density-distance graph is constructed based on densities and minimum distances of all points. The density peaks are easily observed in the graph and can guide the users to select centers of each material as a progressive volume classification process through a set of specified interactions. Experimental results demonstrate that the presented approach does not require the prior knowledge of categories, and the volume classification is accurate with high performance.

Abstract:Hierarchical data is very common in daily life, and tree visualization, which is used to represent hierarchical data, is an important part of visual analysis. While comparative analysis is widely used in problem solving, tree comparison plays an important role in visual analysis. From the perspective of comparing tree numbers, there are inter-tree comparison, pair-wise comparison and multi-tree comparison. Taking data characteristics into consideration, static tree comparison and dynamic tree comparison can be applied. Tree comparison can also be classified into structure comparison and attribute comparison according to comparison tasks. This paper provides an overview of the tree comparison methods relating to visualization area. These methods are divided into three categories according to their representation:juxtaposition, superposition and animation. The paper analyzes both advantages and disadvantages of different approaches, and then gives suggestions in different situations. It concludes with a discussion about the interactive explorations and the challenges of tree comparison.

Abstract:Nowadays, there is increasing need to analyze the complex data with both hierarchical and multi-attributes in many fields such as food safety, stock market, and network security. The visual analytics appeared in recent years provides a good solution to analyze this kind of data. So far, many visualization methods for multi-dimensional data and hierarchical data, the typical data objects in the field of information visualization, have been presented to solve data analyzing problems effectively. However, the existing solutions can't meet requirements of visual analysis for the complex data with both multi-dimensional and hierarchical attributes. This paper presents a technology named Multi-Coordinate in Treemap (MCT), which combines rectangle treemap and multi-dimensional coordinates techniques. MCT uses treemap created with Squarified and Strip layout algorithm to represent hierarchical structure, uses four edges of treemap's rectangular node as the attribute axis, and through mapping property values to attribute axis, connecting attribute points and fitting curve, to achieve visualization of multi-attribute in hierarchical structure. This work applies MCT technology to visualize pesticide residue detection data and implements the visualization for detecting excessive pesticide residue in fruits and vegetables distributed in each provinces of China. This technology provides an efficient analysis tool for field experts. MCT can also be applied in other fields which require visual analysis of complex data with both hierarchical and multi-attribute.

Abstract:Circular treemaps provide an efficient approach for visualization of hierarchical data. This article presents variational circular treemaps with a layout algorithm by solving disk packing as a continuous optimization problem. Compared with the traditional circular treemaps, variational circular treemaps can achieve a higher space utilization ratio, and support natural interactions for data navigation, including focus+context distortions and drill-down and roll-up operations. Experimental results show the effectiveness of the presented method for visualization and interaction.

Abstract:With the rapid development of information technology, large amounts of text data have been produced, collected and stored. Topic modeling is one of the important tools in text analysis, and is widely used for large text collection analysis. However, the topic model usually cannot be combined with users' domain knowledge intuitively and effectively during the topic modeling process. In order to solve this problem, this paper proposes an interactive visual analysis system to help users refine generated topic models. First, the hierarchical Dirichlet process is modified to support the word constraints. Then, the generated topic models is displayed via a matrix view to visually reveal the underlying relationship between words and topics, and semantic-preserving word clouds is used to help users find word constraints effectively. User can interactively refine the topic models by adding word constraints. Finally, the applicability of this new system is demonstrated via case studies and user studies.

Abstract:Radviz is a radial visualization technique that maps data from multi-dimensional space onto a planar picture. The dimensions placed on the circumference of a circle, called dimension anchors, can be reordered to reveal different patterns in the dataset. Extending the number of dimensions can enhance the flexibility in the placement of dimension anchors to explore meaningful visualizations. This paper describes a method that rationally extends a dimension to multiple new dimensions in Radviz. This method first calculates the probability distribution histogram of a dimension. The mean shift algorithm is applied to get centers of probability density to segment the histogram, and then the dimension can be extended according to the number of segments of the histogram. The paper also suggests using Dunn's index and accuracy rate to find the optimal placement of DAs, so the better effect of visual clustering can be achieved and evaluated after the dimension expansion in Radviz. Finally, it demonstrates the effectiveness of the new approach on synthetic and real world datasets.

Abstract:Meteorological data are multisource, multidimensional, large-scaled and multi-scaled. As a result, it is hard to display a complicated 3d scene in meteorology using traditional visualization methods. This paper presents a visualization system, called AVIS for global scale meteorological data. AVIS employs several standard methods for visualizing density fields, vector fields and tensor fields, as well as non-spatial data. AVIS implements spherical volume rendering and hybrid rendering to support the study of internal structures. AVIS also utilizes a cross-platform and parallel framework to support the visualization both in browsers and on other platforms. The framework benefits from a back-end computing cluster and can accelerate the efficiency of computation and rendering. Use cases verify that the presented system can show meteorological data from many aspects and help users analyze multiple types of data comprehensively.

Abstract:With the popularity of security surveillance systems, more and more surveillance cameras have been installed at various traffic roads and public places. A lot of surveillance videos are produced every day. Currently, surveillance video analysis is being done by manual monitoring, which is very inefficient. Most researches on video analysis focus on abnormal behavior detection and tracking of the target individual and lack of analysis of associations between objects/scenes, and there have no effectively representation and analytical methods for the association between objects and scenes. This paper presents a three-dimensional trajectory of moving objects associated video visual analysis to assist manual analysis of videos. First, video data preprocessed to get trajectory information of target object. Because two-dimensional trajectory is unable to deal with the recurring motion, self-intersecting motion and pause, and also lacks of time information, it is difficult be used to analyze correlation between multi-object trajectories in same space. Therefore, time dimension is added to generate three-dimensional trajectory to unite scenes and objects to calculate correlation between trajectories and build the association model. This approach supports sketch interaction so that users can add sketch annotation to assistant analyzing. By utilizing the absence information of objects in certain scene and some predefined rules, the presented approach can be used to alert the abnormal action and assistant user to make corresponding decisions.

Abstract:With growing volume of publications in recent years, researchers have to read much more literatures. Therefore, how to read a scientific article in an efficient way becomes an importance issue. When reading an article, it's necessary to read its references in order to get a better understanding. However, how to differentiate between the relevant and non-relevant references, and how to stay in topic in a large document collection are still challenging tasks. This paper presents GUDOR (GUidedDOcument Reader), a visualization guided reader based on citation and summarization. It (1) extracts the important sentences from a scientific article with an objective-based summarization technique, and visualizes the extraction results by a multi-resolution method; (2) identifies the main topics of the references with a LDA (Latent Dirichlet Allocation) model; (3) tracks user's reading behavior to keep him or her focusing on the reading objective. In addition, the paper describes the functions and operations of the system in a usage scenario and validates its applicability by a user study.

Abstract:With the dramatic countrywide development of mobile internet, it becomes very important to extract valuable information from mobile device log data and report the analysis result through visualization method to help application developers and distributors maximize monetization opportunity. Currently, most of mobile log data analysis work is based on single dimension statistics, e.g., app download rank, and user retention rates. In order to mine deep information hiding behind mobile device log data and summarizes user characteristics. A method is proposed for analyzing users' characteristics and computing users' profile. An app topic model is constructed based on mobile log data, user clusters are build according to app topics, and two visualization methods are designed to show user characteristics clusters. Furthermore, user clusters are combined with time information and geographical information to show user characteristics from additional dimensions. Finally, a mobile log data visualization analysis B/S system is implemented to demonstrate the validity of the method by a case study.

Abstract:Cyber security visualization is a multi-discipline research field. Visualization techniques have injected new vitality into traditional analysis methods for cyber security. However, most existing studies focus on the visual expression and overlook the visual support for the data analysis process. This paper presents a top-down model for anomaly detection on network traffic time-series data drawing from the experience of cyber security analysts. A prototype system is designed based on this model, and it includes four collaborative views with direct and rich interactions. A number of experiments, including port scanning and DDoS attacking, are carried out to demonstrate that this system can support network traffic time-series analysis on overview to detail, point to area and past to future process flows.

Abstract:While model-driven engineering (MDE) methodology has made significant improvements in terms of efficiency and effectiveness in many areas of software development, the same cannot be said of the development of data visualization systems. With this challenge in mind, this paper introduces DVDL (data visualization description language), a modular and hierarchical visualization description language that take advantage of the model-based design of MDE to describe visualization development at an abstract level. The paper also presents DVIZ (data visualization), a visualization system based on DVDL. With a growing popularity and demand for data visualization technology, a number of visualization tools have emerged in recent years, though few of them can be considered as adaptable and scalable as DVIZ. Key features in DVIZ include data source selection by user, property configuration of visual elements, and result publishing and sharing. The system also supports real-time result generation and multi-visual interaction. Lastly, since DVIZ is web-based, it supports result distribution across various social media.

Abstract:Feature model is an essential concept and artifact in feature oriented software development (FOSD). It depicts commonality and variability (C&V) of products in terms of features. With increasingly frequent software evolution, keeping the feature model in consistent with the evolution is very important. Most of the related researches usually analyze the C&V on the requirement level, and modeling the analyzed C&V by the feature model. However, since the feature changes may cause the ripple effect during the modeling process, some new commonalities and variability may be derived. The current researches are still not able to resolve this problem, which leads to some potential overlooking commonalities and inefficiency in reuse. This paper proposes an approach to extend the feature model and analyze the software evolution based on the feature model. The extensions of feature dependency and evolution meta-operators can support the ripple effect analysis of the feature changes, as well as the exploration of the potential commonalities. The new approach also develops some refactoring strategies and a semi-automated tool to support commonality extraction and feature refactoring. In addition, rules and strategies are designed to resolve typical configuration conflicts. Finally, the paper employs a case study to validate the applicability and effectiveness of the presented method.

Abstract:Addressing the requirement for defect detection, this paper proposes critical operation oriented dynamic symbolic execution. First, based on the defined critical operations and the relevant critical paths, a set of initial inputs are evaluated by computing the ability of covering critical operations under different contexts, and efficient initial inputs can be selected for the following dynamic symbolic execution. Second, leveraging the critical operations, dynamic symbolic execution is guided to explore paths which are more prone to defects. In this way, defect detection becomes a process of locating critical operations and exploring critical paths. A prototype system called CrashFinder is implemented and tested on a number of Linux x86 executables. The experimental results show that this approach is effective in initial input evaluation and efficient in defect detection.

Abstract:The general purpose graphic computing units (GPGPUs) have become the new platform for high performance computing due to their massive parallel computing power, and in recent years more and more high performance database research has placed focus on GPU database development. However, today's GPU database researches commonly inherit ROLAP (relational OLAP) model, and mainly address how to realize relational operators in GPU platform and performance tuning, especially on GPU oriented parallel hash join algorithm. GPUs have higher parallel computing power than CPUs but less logical control and management capacity for complex data structure, therefore they are not adaptive for directly migrating the in-memory database query processing algorithms based on complex data structure and memory management. This paper proposes a GPU vectorized processing oriented hybrid OLAP model, semi-MOLAP, which combines direct array access and array computing of MOLAP with storage efficiency of ROLAP. The pure array oriented GPU semi-MOLAP model simplifies GPU data management, reduces complexity of GPU semi-MOLAP algorithms and improves their code efficiency. Meanwhile, the semi-MOLAP operators are divided into co-computing operators on CPU and GPU platforms to improve utilization of both CPUs and GPUs for higher query processing performance.

Abstract:Community recommendation has become increasingly important in sifting valuable communities from massive amounts of communities on the Internet. In recent years novel recommendation is attracting attention, because of the limitation of accurate recommendation which purely pursues accuracy. Existing novel recommendation methods are not suitable for Web community as they fail to utilize unique features of Web community, including the social network established by interactions between users, and the topics of Web community. In this paper, a novel recommendation method, NovelRec, is proposed to suggest communities that users have not seen but are potentially interested in, in order to better broaden users' horizons and improve the development of communities. Specifically, the method explores neighborhood relationships and topical associations from the aforementioned features. First, NovelRec identifies candidate communities for users based on neighborhood relationships between users, and computes accuracy of the candidates using topical associations between users. Next, NovelRec computes novelty of the candidates based on a new metric of user-community distance, and the distance metric is defined by associations between users and communities on both user neighborhood and topic taxonomy. Finally, NovelRec balances novelty with accuracy for the candidates to improve the overall recommendation quality. Experimental results on a real data set of Douban communities show that the proposed method outperforms competitors on the recommendation novelty, and guarantees the recommendation accuracy.

Abstract:Coverage rate is not only an important criteria to assess wireless sensor network but also a key research subject. An optimized coverage algorithm driven by probability model is proposed in this paper. The solution of expectation value of sensor nodes coverage and tolerance as well as the verification process of expectation value of first coverage of concerned target nodes are obtained by calculation of probability coverage model. Regarding network energy, communication path is optimized by means of scheduling policy of node state. As for decrement in node energy, the significance of existence for fitting functional limit is proven. Thus, the energy of sensor nodes matches effectively and the consumption of node energy is restrained. The relationship among coverage functions of sensor nodes in the optimized monitoring area is proven. The simulation experiment shows the proposed algorithm can improve the quality of coverage and the service of network, restrain the consumption of network energy, and prolong the network lifetime.

Abstract:A verifiable dictionary-based searchable encryption scheme is proposed for verifying the completeness of search results. The security of the proposed scheme is analyzed under the security model of adaptive indistinguishability. Compared with the existing schemes, the proposed scheme has advantages in the following aspects:the size of trapdoor is constant, the updating doesn't require recalculation, and especially the search result is verifiable.

Abstract:In commodity OS, the OS kernel runs in the highest privilege layer to manage hardware resources and provides system services. Thus, security-sensitive applications are vulnerable to compromises the underlying untrusted kernel. In this paper, an approach named AppFort is proposed to protect applications from an untrusted OS kernel. To address the high overheads of existing solutions, AppFort makes use of the unique combination of an x86 hardware feature (operand address size), kernel code integrity protection and kernel control flow integrity protection, to intercept and verify both hardware and software operations of the untrusted kernel. As a result, AppFort efficiently protects application's memory, control flows and file I/O, even if the kernel is fully compromised. Experimental results demonstrate that AppFort only incurs very small overhead, which is much better than previous work.