Abstract:Information retrieval-based software bug localization is an active research topic in the domain of software fault localization. It first analyzes the contents of the bug reports and program modules. Then it calculates the similarity between the bug reports and program modules. Finally, it recommends the most similar program modules to developers when given a bug report. This paper presents a systematic survey of existing research achievements of the domestic and international researchers in recent years. First, a research framework is proposed and three key factors (i.e., data sources, retrieval model, and application scenario), which may influence the performance of bug localization methods are identified. Next, existing research achievements in these three key factors are discussed in sequence. Then, the performance evaluation measures and datasets commonly used in information retrieval-based bug localization are summarized. Finally, conclusions of this study are drawn and a perspective of the future work in this research area is discussed.

Abstract:As the core technology of blockchain, consensus protocol has received great attention from academy and industry in recent years, and made a series of research achievements. Existing surveys on consensus protocols generally regard the consensus protocol as a whole, and do not decouple and compare its main components. In this survey, the consensus protocol is divided into two main components, i.e., blocker election and main chain consensus, and their analysis and comparison are conducted correspondingly. In the blocker election component, two mechanisms are mainly discussed, i.e., proof of work and proof of stake. For each mechanism, the encountered problems are analyzed and their corresponding solutions are compared with each other. In the main chain consensus component, its security goal is summarized and security comparison is conducted for probabilistic consensus and deterministic consensus. Through the comprehensive review of the state-of-the-art consensus protocol in blockchain, the developing status, developing trend and some important research directions are finally summarized for the consensus protocol.

Abstract:With the development of cloud computing, recently data center network has been a hot research topic in both academia and industry. Modern data center network is commonly organized in multi-rooted tree topology, such as fat-tree, with multiple available paths to provide high bisection bandwidth. Since the traditional load balancing scheme such as equal-cost multipath routing is not suitable for highly dynamic and bursty traffic in data center network, many load balancing mechanisms have been proposed. In this study, based on the fundamental research problems of load balancing in data center network, the international and domestic research progress of this area is introduced, including central controller-based, switch-based, and host-based load balancing schemes, and then the research trend of load balancing is prospected in data center network.

Abstract:A multimedia world in which human beings live is built from a large number of different modal contents. The information between different modalities is highly correlated and complementary. The main purpose of multi-modal representation learning is to mine the different modalities. Commonness and characteristics produce implicit vectors that can represent multimodal information. This article mainly introduces the corresponding research work of the currently widely used visual language representation, including traditional research methods based on similarity models and current mainstream pre-training methods based on language models. The current better ideas and solutions are to semanticize visual features and then generate representations with textual features through a powerful feature extractor. Transformer is currently used in various tasks of representation learning as the mainstream network architecture. This article elaborates from several different angles of research background, division of different studies, evaluation methods, future development trends, etc.

Abstract:Few-shot learning is defined as learning models to solve problems from small samples. In recent years, under the trend of training model with big data, machine learning and deep learning have achieved success in many fields. However, in many application scenarios in the real world, there is not a large amount of data or labeled data for model training, and labeling a large number of unlabeled samples will cost a lot of manpower. Therefore, how to use a small number of samples for learning has become a problem that needs to be paid attention to at present. This paper systematically combs the current approaches of few-shot learning. It introduces each kind of corresponding model from the three categories: fine-tune based, data augmentation based, and transfer learning based. Then, the data augmentation based approaches are subdivided into unlabeled data based, data generation based, and feature augmentation based approaches. The transfer learning based approaches are subdivided into metric learning based, meta-learning based, and graph neural network based methods. In the following, the paper summarizes the few-shot datasets and the results in the experiments of the aforementioned models. Next, the paper summarizes the current situation and challenges in few-shot learning. Finally, the future technological development of few-shot learning is prospected.

Abstract:Zero-shot learning aims to recognize the unseen classes by using the knowledge of the seen classes that has been learned. In recent years, ‘knowledge+data driven’ has become a new trend but lacking of unified definition of ‘knowledge’ in the current zero-shot tasks of computer vision. This study tries to define the ‘knowledge’ in this field and divided it into three categories, which are primary knowledge, abstract knowledge, and external knowledge. In addition, based on the definition and classification of knowledge, the current works on zero-shot learning (mainly in image classification task) are sorted out, they are divided into zero-shot models based on primary knowledge, zero-shot models based on abstract knowledge, and zero-shot models based on external knowledge. This study also introduces the problems which are domain shift and hubness in this field, and further summarizes existing works based on the problems. Finally, the paper summarizes the datasets and knowledge bases that commonly used in image classification tasks, the evaluation criteria of image classification experiment and the experimental results of representative models. The future works are also summarized and prospected.

Abstract:Machine learning applications span all areas of artificial intelligence, but due to storage and transmission security issues and the flaws of machine learning algorithms themselves, machine learning faces a variety of security- and privacy-oriented attacks. This survey classifies the security and privacy attacks based on the location and timing of attacks in machine learning, and analyzes the causes and attack methods of data poisoning attacks, adversary attacks, data stealing attacks, and querying attacks. Furthermore, the existing security defense mechanisms are summarized. Finally, a perspective of future work and challenges in this research area are discussed.

Abstract:Event-based social network (EBSN) is a new type of social network combining online network and offline network, which has received more and more attentions in recent years. There have been many researchers in important research institutions domestic and abroad to study it and they have achieved a lot of research results. In an EBSN recommendation system, one important task is to design better and more reasonable recommendation algorithms to improve recommendation accuracy and user satisfaction. The key is to fully combine various contextual information in EBSN to mine the hidden features of users, events, and groups. This study mainly reviews the latest research progress of the EBSN recommendation system. First, the definition, structure, attributes, and characteristics of EBSN are outlined, the basic framework of EBSN recommendation systems is introduced, and the differences between EBSN recommendation system and other recommendation systems are analyzed. Secondly, the main recommendation methods and recommended contents of the EBSN recommendation system are generalized, summarized, compared, and analyzed. Finally, the research difficulties and development future trends of the EBSN recommendation system are analyzed, and conclusions of the study are drawn.

Abstract:As the core computing chip of network equipment, network processor can complete the essential services such as routing lookup, high-speed packet processing, and QoS guarantee. Facing the transformation of network environment brought by ultra-high bandwidth and intelligent terminal, the design of the next generation network processor (NGNP) with high performance and evolution is a hot issue in the field of network communication, which is widely concerned by scholars. Combining the advantages of different chip architectures and high-speed services, NGNP has the characteristics of accelerating packet processing, dynamic configuration of hardware resources, and intelligent service application. In this study, the existing research is analyzed and compared from the design of NGNP which using new programmable technology, new network architecture oriented and for new high-performance service. The industrialization process of NGNP is summarized. Finally, the high performance evolvable network processor (HPENP) architecture is proposed. By introducing the hardware and software collaborative packet processing pipeline, multi-level cache and packet scheduling, resource management and programming interface, the details of HPENP design are given and a prototype system is developed and its performance is tested. In this study, the development direction and intelligent application scenario of autonomously controlled network processor architecture are confirmed, and the possible research direction in the future is discussed.

Abstract:Network measurement provides the network designers and managers with fine-grained information on the operational statuses of the network and is the basis for efficient network management and optimization. Network tomography is a hot topic in the field of network measurement and is an end-to-end approach for network measurement. Unlike the traditional internal approaches for network measurement, network tomography uses the end-to-end measurements to infer the internal network performance and network states, thereby incurring low overhead to achieve the network measurement that is independent of the network composition and the network protocols. This paper systematically summarizes the representative research works about network tomography in the past few years. First, the basic model of network tomography is given and three key factors that impact the performance of network tomography are identified: the monitoring node placement, the measurement path construction, and the measurement data analysis. Then, the related works are reviewed on these three factors separately. In particular, the major limitations of existing network tomography methods in practical applications are explored, and the efficient solutions proposed in recent years are introduced. Lastly, some challenges and future research directions are discussed in the field of network tomography based on existing research works.

Abstract:Deep learning has achieved great success in the field of computer vision, surpassing many traditional methods. However, in recent years, deep learning technology has been abused in the production of fake videos, making fake videos represented by Deepfakes flooding on the Internet. This technique produces pornographic movies, fake news, political rumors by tampering or replacing the face information of the original videos and synthesizes fake speech. In order to eliminate the negative effects brought by such forgery technologies, many researchers have conducted in-depth research on the identification of fake videos and proposed a series of detection methods to help institutions or communities to identify such fake videos. Nevertheless, the current detection technology still has many limitations such as specific distribution data, specific compression ratio, and so on, far behind the generation technology of fake video. In addition, different researchers handle the problem from different angles. The data sets and evaluation indicators used are not uniform. So far, the academic community still lacks a unified understanding of deep forgery and detection technology. The architecture of deep forgery and detection technology research is not clear. In this review, the development of deep forgery and detection technologies are reviewed. Besides, existing research works are systematically summarize and scientifically classified. Finally, the social risks posed by the spread of Deepfakes technology are discussed, the limitations of detection technology are analyzed, and the challenges and potential research directions of detection technology are discussed, aiming to provide guidance for follow-up researchers to further promote the development and deployment of Deepfakes detection technology.

Abstract:The application of deep learning in the field of medical image segmentation has attracted great attentions, among which the U-Net proposed in 2015 has been widely concerned because of its good segmentation effect and scalable structure. In recent years, with the improvement of the performance requirements of medical image segmentation, many scholars are improving and expanding the U-Net structure, such as the improvement of encoder-decoder, or the external feature pyramid, and so on. In this study, the medical image segmentation technology based on U-Net structure improvement is summarized from the aspects of performance-oriented optimization and structure-oriented improvement. Related methods are reviewed, classified and summarized. The paper evaluates the parameters and modules, and then summarizes the ideas and methods for improving the U-Net structure for different goals, which provides references for related research.

Abstract:Steganography and steganalysis are one of the research hotspots in the field of information security. The abuse of steganography has caused many potential safety hazards. For example, illegal elements use steganography for covert communications to carry out terrorist attacks. The design of traditional steganalysis methods requires a large amount of prior knowledge, and the steganalysis methods based on deep learning use the powerful representation learning ability of the network to autonomously extract abnormal image features, which greatly reduces human participation and achieves good results. To promote the research of steganalysis technology based on deep learning, this study analyzes and summarizes the main methods and work in the field of steganalysis. Firstly, this study analyzes and compares the differences between traditional steganalysis and deep learning-based steganalysis. Furthermore, according to the different training methods, the steganalysis models based on deep learning are divided into two categories: semi-learning steganalysis model and full-learningsteganalysis model. The network structure and detection effect of various types of steganalysis based on deep learning are introduced in detail. In addition, the challenges that the adversarial samples pose to deep learning security are analyzed and summarized, the detection method of adversarial samples is expounded based on steganalysis. Finally, this study summarizes the pros and cons of existing steganalysis models based on deep learning and discusses its development trends.

Abstract:As an important topic in the field of surface deformation, flattening and folding has become a research hotspot in recent years. In order to satisfy the constraints of aesthetics and mechanics, it is necessary to design flattening and folding structure. Using computer technology to simulate the flattening and folding of objects, geometric structures that meet the constraints can be designed. At present, flattening and folding are widely used in industrial design, biomedicine, intelligent robot, furniture design, and other fields. This paper mainly introduces the research status of computer flattening and folding objects in recent years. Firstly, the algorithms of flattening and folding are classified, and the basic ideas of each method are briefly described. Then, the various methods are analyzed, and their advantages and limitations are summarized. Finally, the corresponding evaluation criteria are given for further comparison. The flattening and folding method of curved surface has been developed and innovated step by step, and tends to be mature. However, due to the complex practical application requirements, the deformation results may not be perfect. A review of the current flattening and folding methods can provide research direction for future work.