High-dimensional state and action spaces com- bined with sparse reward structures in reinforcement learning (RL) environments typically require advanced control architec- tures. Hierarchical Reinforcement Learning (HRL) demonstrates superior performance compared to atomic RL approaches in these challenging scenarios. HRL can manage the complexity of commands to achieve task objectives through its hierarchical structure. One of the key challenges in HRL is efficiently training each level’s policy with optimal data collection from its experience. Off-policy correction is a critical technique for facilitating sample-efficient off-policy training in HRL, as it addresses the non-stationary issue of higher-level policy training. However, existing methods typically employ indirect probabilistic approaches that fail to accurately capture the current capability of the lower-level policy. This mismatch ultimately constrains the effectiveness of higher-level policy training. In this paper, we propose a novel HRL model that supports direct off-policy correction based on a Flow-based Deep Generative Model (FDGM). This approach leverages the inverse operation of FDGM to achieve goals aligned with the current knowledge of the lower-level policy. Additionally, our model addresses the limitations of FDGM to enable its effective use in HRL. Through comparative experiments on benchmark environments, our model demonstrates superior performance over existing models

Data videos are widely used in media and education, but can overwhelm viewers if poorly organized. We assess whether a hierarchical drill-down structure improves comprehension and reduces extraneous cognitive load in linear, non-interactive data videos. Building on cognitive load theory and narrative visualization research, we propose a conceptual model that divides a narrative into successive layers of detail. We conducted an online between-subjects experiment (N = 100) comparing a drill-down video with an equivalent flat baseline. To isolate visual-structuring effects and reflect common sound-off contexts (e.g., autoplay feeds, public displays), we used short, caption-free videos without audio. Independent-samples t-tests showed slightly better recall with drill-down but no statistically significant differences in recall, cognitive load, or self-reported comprehension. Qualitative feedback highlighted that fast pacing and high visual density in both videos imposed substantial cognitive demands, likely overshadowing any structural benefits. Our findings encourage designs that combine drill-down structuring with adaptive pacing, persistent visual anchors, and multimedia cues.

Unsupervised pre-training has been on the lookout for the virtue of a value function representation referred to as successor features (SFs), which decouples the dynamics of the environment from the rewards. It has a significant impact on the process of task-specific fine-tuning due to the decomposition. However, existing approaches struggle with local optima due to the unified intrinsic reward of exploration and exploitation without considering the linear regression problem and the discriminator supporting a small skill sapce. We propose a novel unsupervised pre-training model with SFs based on a non-monolithic exploration methodology. Our approach pursues the decomposition of exploitation and exploration of an agent built on SFs, which requires separate agents for the respective purpose. The idea will leverage not only the inherent characteristics of SFs such as a quick adaptation to new tasks but also the exploratory and task-agnostic capabilities. Our suggested model is termed Non-Monolithic unsupervised Pretraining with Successor features (NMPS), which improves the performance of the original monolithic exploration method of pre-training with SFs. NMPS outperforms Active Pre-training with Successor Features (APS) in a comparative experiment.

Factuality issues remain challenging to abstractive text summarization despite considerable progress in recent years. This is partly because abstractive text summarization models and methods have a limited capacity to capture complex syntactic structures. Hence, researchers have explored syntactic structures in modeling abstractive text summarization, such as learning structure-aware representations and formulating structure- derived learning objectives. However, their efforts are confined as their optimization only uses cross-entropy-based maximum likelihood estimation, which may underlie some factual problems as we reason later. This paper proposes a syntactic structure- aware encoder-decoder model that incorporates novel learning tasks on syntactic structures. By doing so, we aim at generalizing a latent semantic space that encodes both lexical semantics and complex syntactic structures holistically to tackle the syntactic structure-related factual issues. Our experiments show that our approach improves overall summaries on auto-metric evaluations over the adopted baseline. The human evaluation also indicates that our approach betters summaries on factuality and fluency overall. Further qualitative assessment sheds light on the plausible reasons underlying the quantitative evaluation results.

Radio Frequency Identification (RFID) is a crucial technology in the Internet of Things (IoT), enabling seamless wireless communication and data exchange. However, these technologies can pose significant security chal-lenges if not implemented with proper attention to security protocols-especially in communication, where pre-shared keys are not used between active tags and readers for device authentication. Some recent authentication protocols rely solely on a hash function, nonce, and single public kay agreement, which can lead to failure to implement robust security and proper authentication or ineffective for high security application environments. To effectively address these challenges this paper proposes a secure Elliptic Curve Cryptography (ECC) based lightweight mutual authentication protocol utilizing a hybrid key agreement protocol between active tag and reader for secure communication in RFID-enabled devices in the IoT environments. The informal analysis demonstrates a secure communication environment for data privacy and flexibility through effective key management. This protocol is adaptable to various applications by addressing specific requirements and limitations.

In the evolving landscape of Web 3.0, 5G/6G, and real-world applications, federated learning faces unique challenges. Traditional incentive mechanisms struggle to address the need to motivate both data owners to provide high-quality data and model experts to optimize model performance. To navigate this complex scenario, we introduce the Reciprocal Federated Learning Framework (RFLF). This innovative approach fosters a fair and dynamic reward structure that incentivizes both high-quality data contributions and optimal model development. Extensive experiments on benchmark datasets demonstrate that the RFLF significantly enhances fairness and efficiency within federated learning. These results showcase the RFLF’s potential to transform data-driven technologies, promoting both efficiency and equitable outcomes.

Effective narrative visualization communicates information by integrating story-telling and data visualization in a comprehensible, compelling manner. The compelling aspect of effective narrative visualization consequentially results in the potential to shift the attitude of an audience. However, there is much to understand about how narrative visualization can best be designed to influence target audiences. This paper focuses on an empirical experiment where we examined the effects of two communication strategies - anthropomorphism and personal identification - on a young adult audience. In particular, we wanted to understand which strategy, when integrated into narrative visualization, can nudge a specific audience's attitude towards greater consideration in the context of the COVID-19 pandemic. Our results indicated that the personal identification communication strategy was the most successful in nudging participants. This study contributes a better grasp of how technologies such as narrative visualization, using different communication strategies, can deliver more targeted messaging.

The Internet of Things (IoT) has significantly enhanced various aspects of our daily lives, including security, health, education, and energy efficiency, among others. Within the realm of IoT, image classification stands as a pivotal technique that has achieved notable success in domains such as facial recognition within security and scene recognition in transportation for traffic analysis. Nonetheless, the challenge emerges when tackling classification tasks with only limited labeled samples available for each category. Conventional machine learning techniques often struggle to attain satisfactory classification results under such circumstances. To address this issue, the concept of few-shot learning has emerged, aiming to achieve effective classification using only a small number of labeled samples. State-of-the-art few-shot learning models have introduced novel frameworks to tackle this problem. However, the inherent ambiguity and uncertainty within data often hinder the performance of classification methods. To overcome this limitation, this paper proposes the integration of fuzzy learning with few-shot learning in the context of feature extraction. The objective is to mitigate data fuzziness and enhance model performance. Leveraging a fuzzy extraction algorithm, we introduce fuzzy prototype networks and a fuzzy graph neural network with fuzzy reasoning. These frameworks are designed to analyze noisy and uncertain data, utilizing convolutional neural networks for feature extraction and applying fuzzy reasoning to capture ambiguity representations for features within each fuzzy set. The SoftMax function is then normalized to serve as a feature weight, effectively constraining the original feature vector. The effectiveness and efficiency of our proposed model are demonstrated through experimental evaluations conducted on various public datasets. The results showcase the model’s capability in addressing the challenges posed by limited labeled data and data uncertainty, thus reaffirming its potential in enhancing the performance of image classification tasks within the IoT context.

Most exploration research on reinforcement learning (RL) has paid attention to ‘the way of exploration’, which is ‘how to explore’. The other exploration research, ‘when to explore’, has not been the main focus of RL exploration research. The issue of ‘when’ of a monolithic exploration in the usual RL exploration behaviour binds an exploratory action to an exploitational action of an agent. Recently, a non-monolithic exploration research has emerged to examine the mode-switching exploration behaviour of humans and animals. The ultimate purpose of our research is to enable an agent to decide when to explore or exploit autonomously. We describe the initial research of an autonomous multi-mode exploration of non-monolithic behaviour in an options framework. The higher performance of our method is shown against the existing non-monolithic exploration method through comparative experimental results.

In recent years abstractive text summarization (ATS) research has made considerable progress attributed to two key improvements, deep neural modeling and likelihood estimation based sampling, in the end-to-end optimization training. While modeling has grounded on a few de facto highly capable base models within encoder-decoder architecture, novel sampling ideas, such as random masking classification and generative prediction by unsupervised learning, have also been explored. They aim at improving prior knowledge, particularly of language modeling for downstream tasks. It has led to the notable performance gain of ATS. But several challenges remain, for example, undesirable word repeats. In this paper, we propose a determinantal point process (DPP) based novel sampling method to address the issue. It can be easily integrated with the existing ATS models. Our experiments and subsequent analysis have revealed that the adopted models trained by our sampling method reduce undesirable word repeats and improve word coverage while achieving competitive ROUGE scores.

The visual modeling method enables flexible interactions with rich graphical depictions of data and supports the exploration of the complexities of epidemiological analysis. However, most epidemiology visualizations do not support the combined analysis of objective factors that might influence the transmission situation, resulting in a lack of quantitative and qualitative evidence. To address this issue, we developed a portrait-based visual modeling method called +msRNAer. This method considers the spatiotemporal features of virus transmission patterns and multidimensional features of objective risk factors in communities, enabling portrait-based exploration and comparison in epidemiological analysis. We applied +msRNAer to aggregate COVID-19-related datasets in New South Wales, Australia, combining COVID-19 case number trends, geo-information, intervention events, and expert-supervised risk factors extracted from local government area-based censuses. We perfected the +msRNAer workflow with collaborative views and evaluated its feasibility, effectiveness, and usefulness through one user study and three subject-driven case studies. Positive feedback from experts indicates that +msRNAer provides a general understanding for analyzing comprehension that not only compares relationships between cases in time-varying and risk factors through portraits but also supports navigation in fundamental geographical, timeline, and other factor comparisons. By adopting interactions, experts discovered functional and practical implications for potential patterns of long-standing community factors regarding the vulnerability faced by the pandemic. Experts confirmed that +msRNAer is expected to deliver visual modeling benefits with spatiotemporal and multidimensional features in other epidemiological analysis scenarios.

Federated learning (FL) is an emerging solution for machine learning model training in edge/fog computing systems. Unlike traditional systems that collect and train models on clouds, FL allows multiple edge/fog nodes to train a global model collaboratively without revealing their local data to clouds. Compared with traditional systems, it is inherited with better privacy protection ability. Although the basic privacy protection is inherited in FL, the privacy leakage from shard models is still unsolved. Existing solutions attempt to enhance the privacy of shared model parameters by adding differential privacy (DP) noise. However, these solutions all suffer from accuracy loss and convergence problems owing to the injected noise. In this paper, we propose a novel federated learning protocol to solve the above problem. The model trained on a carefully selected sampling subset can achieve the same level privacy protection as DP while preserving the model accuracy. Experimentally, we proved that our protocol achieves better model accuracy in the same privacy guarantee compared with noise injecting DP methods.

In domain adaptation, where the feature distributions of the source and target domains are different, various distance-based methods have been proposed to handle the domain shift by minimizing the discrepancy between the source and target domains. These methods use hand-crafted bottleneck networks, which might hinder the alignment of hidden feature representations extracted from both domains. In this paper, we propose a new method called Alignment Architecture Search with Population Correlation (AASPC) to automatically learn the architecture of the bottleneck network that can align the source and target domains. The proposed AASPC method introduces a new similarity function called Population Correlation (PC) to measure the domain discrepancy. The proposed AASPC method leverages PC to learn the alignment architecture and domaininvariant feature representation. Experiments on several benchmark datasets, including Office-31, Office-Home, and VisDA-2017, show the effectiveness of the proposed AASPC method.

Designing neural network architecture for embedded devices is practical but challenging because the models are expected to be not only accurate but also enough lightweight and robust. However, it is challenging to balance those trade-offs manually because of the large search space. To solve this problem, we propose an Effective, Efficient, and Robust Neural Architecture Search (E2RNAS) method to automatically search a neural network architecture that balances the performance, robustness, and resource consumption. Unlike previous studies, the objective function of the proposed E2RNAS method is formulated as a multi-objective bi-level optimization problem with the upper-level subproblem as a multi-objective optimization problem that considers the performance, robustness, and resource consumption. To solve the proposed objective function, we integrate the multiple-gradient descent algorithm, a widely studied gradient-based multi-objective optimization algorithm, with the bi-level optimization. Experiments on benchmark datasets show that the proposed E2RNAS method can find robust architecture with low resource consumption and comparable classification accuracy.

Hierarchical structures are very common in the real world for recording all kinds of relational data generated in our daily life and business procedures. A very popular visualization method for displaying such data structures is called "Tree". So far, there are a variety of Tree visualization methods that have been proposed and most of them can only visualize one hierarchical dataset at a time. Hence, it raises the difficulty of comparison between two or more hierarchical datasets.In this paper, we proposed Pansy Tree which used a tree metaphor to visualize merged hierarchies. We design a unique icon named pansy to represent each merged node in the structure. Each Pansy is encoded by three colors mapping data items from three different datasets in the same hierarchical position (or tree node). The petals and sepal on Pansy are designed for showing each attribute’s values and hierarchical information. We also redefine the links in force layout encoded by width and animation to better convey hierarchical information. We further apply Pansy Tree into CNCEE datasets and demonstrate two use cases to verify its effectiveness.The main contribution of this work is to merge three datasets into one tree that makes it much easier to explore and compare the structures, data items and data attributes with visual tools.

We propose BarcodeTree (BCT), a novel visualization technique for comparing topological structures and node attribute values of multiple trees. BCT can provide an overview of one hundred shallow and stable trees simultaneously, without aggregating individual nodes. Each BCT is shown within a single row using a style similar to a barcode, allowing trees to be stacked vertically with matching nodes aligned horizontally to ease comparison and maintain space efficiency. We design several visual cues and interactive techniques to help users understand the topological structure and compare trees. In an experiment comparing two variants of BCT with icicle plots, the results suggest that BCTs make it easier to visually compare trees by reducing the vertical distance between different trees. We also present two case studies involving a dataset of hundreds of trees to demonstrate BCT's utility.

Today there are abounding collected data in cases of various diseases in medical sciences. Physicians can access new findings about diseases and procedures in dealing with them by probing these data. Clinical data is a collection of large and complex datasets that commonly appear in multidimensional data formats. It has been recognized as a big challenge in modern data analysis tasks. Therefore, there is an urgent need to find new and effective techniques to deal with such huge datasets. This paper presents an application of a new visual data mining platform for visual analysis of the stroke data for predicting the levels of risk to those people who have the similar characteristics of the stroke patients. The visualization platform uses a hierarchical clustering algorithm to aggregate the data and map coherent groups of data-points to the same visual elements - curved 'super-polylines' that significantly reduces the visual complexity of the visualization. On the other hand, to enable users to interactively manipulate data items (super-polylines) in the parallel coordinates geometry through the mouse rollover and clicking, we created many 'virtual nodes' along the multi-axis of the visualization based on the hierarchical structure of the value range of selected data attributes. The experimental result shows that we can easily verify research hypothesis and reach to the conclusion of research questions through human-data & human-algorithm interactions by using this visual platform with a fully transparency manner of data processing.

Information diffusion analysis is important in social media. In this work, we present a coherent ego-centric and event-centric model to investigate diffusion patterns and user behaviors. Applying the model, we propose Diffusion Map+ (D-Maps+), a novel visualization method to support exploration and analysis of user behaviors and diffusion patterns through a map metaphor. For ego-centric analysis, users who participated in reposting (i.e., resending a message initially posted by others) one central user’s posts (i.e., a series of original tweets) are collected. Event-centric analysis focuses on multiple central users discussing a specific event, with all the people participating and reposting messages about it. Social media users are mapped to a hexagonal grid based on their behavior similarities and in the chronological order of repostings. With the additional interactions and linkings, D-Map+ is capable of providing visual profiling of influential users, describing their social behaviors and analyzing the evolution of significant events in social media. A comprehensive visual analysis system is developed to support interactive exploration with D-Map+. We evaluate our work with real-world social media data and find interesting patterns among users and events. We also perform evaluations including user studies and expert feedback to certify the capabilities of our method.

In this work, we present Interaction+, a tool that enhances the interactive capability of existing web-based visualizations. Different from the toolkits for authoring interactions during the visualization construction, Interaction+ takes existing visualizations as input, analyzes the visual objects, and provides users with a suite of interactions to facilitate the visual exploration, including selection, aggregation, arrangement, comparison, filtering, and annotation. Without accessing the underlying data or process how the visualization is constructed, Interaction+ is application-independent and can be employed in various visualizations on the web. We demonstrate its usage in two scenarios and evaluate its effectiveness with a qualitative user study.

There are often multiple routes between regions. Drivers choose different routes with different considerations. Such considerations, have always been a point of interest in the transportation area. Studies of route choice behaviour are usually based on small range experiments with a group of volunteers. However, the experiment data is quite limited in its spatial and temporal scale as well as the practical reliability. In this work, we explore the possibility of studying route choice behaviour based on general trajectory dataset, which is more realistic in a wider scale. We develop a visual analytic system to help users handle the large-scale trajectory data, compare different route choices, and explore the underlying reasons. Specifically, the system consists of: 1. the interactive trajectory filtering which supports graphical trajectory query; 2. the spatial visualization which gives an overview of all feasible routes extracted from filtered trajectories; 3. the factor visual analytics which provides the exploration and hypothesis construction of different factors' impact on route choice behaviour, and the verification with an integrated route choice model. Applying to real taxi GPS dataset, we report the system's performance and demonstrate its effectiveness with three cases.

This paper presents a novel visual analysis tool, EnsembleGraph, which aims at helping scientists understand spatiotemporal similarities across runs in time-varying ensemble simulation data. We abstract the input data into a graph, where each node represents a region with similar behaviors across runs and nodes in adjacent time frames are linked if their regions overlap spatially. The visualization of this graph, combined with multiple-linked views showing details, enables users to explore, select, and compare the extracted regions that have similar behaviors. The driving application of this paper is the study of regional emission influences over tropospheric ozone, based on the ensemble simulations conducted with different anthropogenic emission absences using MOZART-4. We demonstrate the effectiveness of our method by visualizing the MOZART-4 ensemble simulation data and evaluating the relative regional emission influences on tropospheric ozone concentrations.

We propose a novel calculation method of personality based on the Chinese physiognomy. The proposed solution combines the ancient and the modem physiognomy to summarize the corresponding relation between the personality and facial feature and model the baseline to shape the face feature. We compute histogram of image by searching for the threshold values to create a binary image in an adaptive way. The two-pass connected component method indicates the feature region. We encode the binary image to remove the noise point, so that the new connected image can provide a better result. The method was tested on ORL face database.

Social media data with geotags can be used to track people's movements in their daily lives. By providing both rich text and movement information, visual analysis on social media data can be both interesting and challenging. In contrast to traditional movement data, the sparseness and irregularity of social media data increase the difficulty of extracting movement patterns. To facilitate the understanding of people's movements, we present an interactive visual analytics system to support the exploration of sparsely sampled trajectory data from social media. We propose a heuristic model to reduce the uncertainty caused by the nature of social media data. In the proposed system, users can filter and select reliable data from each derived movement category, based on the guidance of uncertainty model and interactive selection tools. By iteratively analyzing filtered movements, users can explore the semantics of movements, including the transportation methods, frequent visiting sequences and keyword descriptions. We provide two cases to demonstrate how our system can help users to explore the movement patterns.

Visualizing dynamic graphs are challenging due to the difficulty to preserving a coherent mental map of the changing graphs. In this paper, we propose a novel layout algorithm which is capable of maintaining the overall structure of a sequence graphs. Through Laplacian constrained distance embedding, our method works online and maintains the aesthetic of individual graphs and the shape similarity between adjacent graphs in the sequence. By preserving the shape of the same graph components across different time steps, our method can effectively help users track and gain insights into the graph changes. Two datasets are tested to demonstrate the effectiveness of our algorithm.

Understanding the Origin-Destination (OD) patterns between different regions of a city is important in urban planning. In this work, based on taxi GPS data, we propose OD-Wheel, a novel visual design and associated analysis tool, to explore OD patterns. Once users define a region, all taxi trips starting from or ending to that region are selected and grouped into OD clusters. With a hybrid circular-linear visual design, OD-Wheel allows users to explore the dynamic patterns of each OD cluster, including the variation of traffic flow volume and traveling time. The proposed tool supports convenient interactions and allows users to compare and correlate the patterns between different OD clusters. A use study with real data sets demonstrates the effectiveness of the proposed OD-Wheel.

While prior works on enclosure approach, guarantees the space utilization of a single geometrical area, mostly rectangle, this paper proposes a flexible enclosure tree layout method for partitioning various polygonal shapes that break through the limitation of rectangular constraint. Similar to Treemap techniques, it uses enclosure to divide display space into smaller areas for its sub-hierarchies. The algorithm can partition a polygonal shape or even an arbitrary shape into smaller polygons, rotated rectangles or vertical-horizontal rectangles. The proposed method and implementation algorithms provide an effective interactive visualization tool for partitioning large hierarchical structures within a confined display area with different shapes for real-time applications. We demonstrated the effective of the new method with a case study, an automated evaluation and a usability study.

Space-filling visualization techniques have proved their capability in visualizing large hierarchical structured data. However, most existing techniques restrict their partitioning process in vertical and horizontal direction only, which cause problem with identifying hierarchical structures. According to Gestalt research, limiting tree map visualisation to rectangles blocks the utilisation of human capability on object recognition, due to the same fixed size (90 degrees) of all the angles of the shapes in the tree visualisation. However, this assertion was only supported by theory and not rooted in empirical perception data. We conducted a series of controlled experiments to investigate the effect of shape variation of data elements and container in visual data analysis process. We first studied how shape variation affects user's perception in the visual data analysis process. We compared combined treemap with traditional rectangular treemaps, slice & dice treemaps and squarifed treemaps. Finally, we demonstrated the effect of the new approach which combines rectangular and non-rectangular treemaps and validate the method based on the empirical results.

Space-filling visualization techniques have proved their capability in visualizing large hierarchical structured data. However, most existing techniques restrict their partitioning process in vertical and horizontal direction only, which cause problem with identifying hierarchical structures. This paper presents a new space-filling method named Angular Treemaps that relax the constraint of the rectangular subdivision. The approach of Angular Treemaps utilizes divide and conquer paradigm to visualize and emphasize large hierarchical structures within a compact and limited display area with better interpretability. Angular Treemaps generate various layouts to highlight hierarchical sub-structure based on user's preferences or system recommendations. It offers flexibility to be adopted into a wider range of applications, regarding different enclosing shapes. Preliminary usability results suggest user's performance by using this technique is improved in locating and identifying categorized analysis tasks.

Visualization of uncertainty in datasets is a new field of research, which aims to represent incomplete data for analysis in real scenarios. In many cases, datasets, especially multi-dimensional datasets, often contain either errors or uncertain values. To address this challenge, we may treat these uncertainties as scalar values like probability. For visual representation in parallel coordinates, we draw a small "circle" to temporarily define a dummy vertex for an uncertain value of a data item, at the crossing point between polylines and the axis of certain dimension. Furthermore, these temporary positions of uncertainty could be permuted to achieve visual effectiveness. This feature provides a great opportunity by optimizing the order of uncertain values to tackle another important challenge in information visualization: clutter reduction. Visual clutter always obscures the visualizing structure even in small datasets. In this paper, we apply Sugiyama's layered directed graph drawing algorithm into parallel coordinates visualization to minimize the number of edge crossing among polylines, which has significantly improved the readability of visual structure. Experiments in case studies have shown the effectiveness of our new methods for clutter reduction in parallel coordinates visualization. These experiments also imply that besides visual clutter, the number of uncertain values and the type of multi-dimensional data are important attributes that affect visualization performance in this field.

Highlighting was the basic viewing control mechanism in computer graphics and visualization to guide users' attention in reading diagrams, images, graphs and digital texts. As the rapid growth of theory and practice in information visualization, highlighting has extended its role that acts as not only a viewing control, but also an interaction control and a graphic recommendation mechanism in knowledge visualization and visual analytics. In this work, we attempt to give a formal summarization and classification of the existing highlighting methods and techniques that can be applied in Information Visualization, Visual Analytics and Knowledge Visualization. We propose a new three-layer model of highlighting. We discuss the responsibilities of each layer in the different stage of the visual information processing.

The traditional solutions to the stock market security are not sufficient in identifying attackers and further attack plans from the analysis of existing events.Therefore, it is difficult for analysts to prevent future unexpected events or frauds by only monitoring the realtime trading information. The event-driven fraud detection in financial market could not help analysts to find attack plans and the further intention of attackers. This paper proposed a new framework of visual analytics for stock market security. The proposed solution consists of two stages: 1) visual surveillance of market performance, and 2) behavior-driven visual analysis of trading networks.In the first stage, we use a 3D treemaps to monitor the real-time stock market performance and to identify a particular stock that produced an unusual trading pattern. We then move to the next stage: social network visualization to conduct behavior-driven visual analysis of suspected pattern. Through the visual analysis of social (or trading)network, analysts may finally identify the attackers (the sources of the fraud), and further attack plans.