With the rapid development of smart grid technology, the demand for data integrity and privacy protection in grid systems is becoming increasingly prominent. To address the privacy protection needs, we propose an anonymous smart grid system based on revocable ring signatures. This system not only protects the user’s privacy and ensures the integrity of the user’s log data, but also allows the server to revoke log data after verifying the identity of the signer. This paper innovatively adds a revocable stage to the ring signature, combined with an interactive Schnorr authentication protocol to achieve identity authentication of the signer, meeting the needs of legitimate users for log revocation. We provide a detailed analysis of the security and performance of our scheme. The security analysis demonstrates the correctness and non-repudiation of the revocation stage authentication protocol, while the experimental results show the feasibility and practicality of the proposed solution, providing an innovative security management solution for grid systems.
The healthcare industry is one of the industries most affected by insider threats. At present, the mainstream abnormal behavior detection of internal threats has the problems of relatively single reference dimension of user behavior data and lack of correlation analysis between data of different dimensions. Therefore, this paper proposes an abnormal behavior detection method that integrates user behavior correlation feature sequence modeling (UBCFSM). This method fully considers the multidimension and correlation of user behavior data, and uses the time convolutional network to learn behavior sequence features and the bidirectional long and short term memory network to learn behavior habit features of the convolutional attention mechanism. In addition, the deep neural network is introduced to fuse the deviations between the predicted results of different models and the actual results, detect abnormal results, and improve the generalization ability and detection accuracy of the model. Finally, experiments were carried out on the CERT public data set, and the AUC score of 97.1% was obtained. Compared with other abnormal behavior detection models, the proposed model has achieved significant improvement, showing its effectiveness and competitiveness.
Time series anomaly detection often faces significant challenges due to the high dimensionality, interdependence, and label sparsity of the data. Traditional methods struggle to model the intricate relationships between multiple time series components and often fail to accurately detect anomalies due to insufficient data dependence mining. To address these issues, we introduce GCRNF, a novel Graph-Enhanced Standardized Flow Model that incorporates a Bayesian network for causality modeling. This model decomposes joint probabilities into conditional probabilities and utilizes a graph dependency encoder with hybrid CNN-RNN models and Graph Convolutional Networks (GCNs). Our experimental results demonstrate that GCRNF outperforms existing methods in density estimation, anomaly detection, and drift identification within time series distributions. Specifically, GCRNF showed a significant improvement in detection accuracy and robustness compared to traditional methods, providing a more effective solution for identifying anomalies in complex multivariate time series data.
Standardized big data, as a product of the combination of big data and standardization work, objectively reflects the status quo and trends of standardization work. An in-depth study of the distribution characteristics of standardized big data resources has important value for improving the efficiency of standardization. Therefore, this paper explores the characteristics of the distribution of standardization resources from the dimensions of spatial and temporal distribution, industry distribution, standard type distribution, and distribution of think tank experts through visual analysis methods. The study shows that: (1) the distribution of standardization resources in Shandong Province presents significant hierarchical and geographical characteristics; (2) from the viewpoint of the distribution of industry types and standard types, the distribution of standardization resources in Shandong Province presents a typical "long-tailed distribution"; (3) the distribution of think-tank experts presents obvious "agglomeration effect".
Prostate cancer is a prevalent form of cancer among men, with a high rate of occurrence but relatively lower mortality compared to other types of cancer. Early detection is crucial because it allows for earlier intervention, potentially preventing progression to more severe stages, and improving patient outcomes. However, human diagnosis is time-consuming and inefficient, an autonomous early detection system is essential for saving lives. To address this challenge, we propose a novel MoE-MRIViT framework that uses advanced machine learning and computer vision techniques to analyze MRI images and identify features indicative of prostate cancer. The system processes the data using a vision transformer to extract precise prostate cancer features from MRI scans, and uses attention mechanisms to focus on significant areas. Through the MoE-MRIViT intervention in two stages, we have achieved superior results compared to the state-of-the-art (SOTA) methods. Experimental results demonstrate that our model achieves an impressive accuracy of 94.84%, outperforming the ResNet-101 and the basic ViT models, which achieve accuracies of 89.4% and 92.31%, respectively.
Existing indoor visible light positioning methods, such as those based on fingerprint database algorithms, are complex to construct and consume significant resources. In response to this issue, the text proposes an end-to-end indoor visible light positioning method that samples continuous signals at a small number of random coordinate positions in indoor environments. Through a deep learning method based on attention mechanism, a nonlinear mapping between the received signal and spatial position coordinates is constructed, and the time series properties in the received signal are encoded as a feature in the data. The method is further improved by combining residual networks to improve accuracy. Through simulation experiments, it has been verified that the deep learning based localization method described in this article can achieve sub meter level accuracy on small sample datasets, and it has been proven that time series features have a significant impact on localization accuracy in indoor visible light positioning algorithms based on deep learning.
Cyber ranges are important scientific infrastructures and realistic experimental platforms that are highly valued by countries around the world. With continuous evolution of technologies, nowadays cyber ranges are facing new challenges such as large network emulation and security assessment problems. To find solutions to solve these problems, we need to conduct a detailed investigation of current research works and projects about cyber ranges. In this paper, we summarize two main challenges and analyze current typical projects by classifications to provide a global overview. We also summarize key techniques used in cyber ranges including scenario design works which are hotspots abroad recently. To our best knowledge, we are the first to survey scenario design works based on cyber ranges. This work hopes to provide a technical reference and innovative inspiration for cyber range researchers.
The categorization of plaintext and ciphertext constitutes a pivotal endeavour, necessitating the swift segregation of extant data and the evaluative determination of the value of potential classified information. Anchored in the tenets of the Random Forest algorithm, this study engineers a bidirectional Recurrent Neural Network (RNN) model endowed with multiple selectors. This innovation markedly augments the classification efficacy and generalisability of generic bidirectional models, capacitating them to manage high-dimensional and extensive datasets with proficiency. Furthermore, with a concentrated examination of the Long Short-Term Memory (LSTM) network and the Gated Recurrent Unit (GRU), a deep LSTM-SiTGRU ciphertext classification model is advanced. Through meticulous empirical examinations across diverse datasets and operational contexts, an exhaustive comparison of the LSTM-SiTGRU performance, merits, limitations, and suitable conditions in the explicit ciphertext classification endeavour is conducted. Experimental outcomes corroborate that the LSTM-SiTGRU model attains a classification accuracy reaching up to 86%, thereby offering both theoretical insights and pragmatic directions for the thorough investigation of plaintext categorisation and the deployment of deep learning methodologies within the domain of information security.
At present, the total number of disabled people in China is 85.914 million, and the number of elderly people over 60 is about 297 million, and the demand for rehabilitation and medical treatment is increasing day by day. However, there are still pain points, such as lack of cognition and shortage of talents in rehabilitation medicine, and there is a certain degree of disconnect with the huge demand. Therefore, it is very important to alleviate the imbalance between supply and demand of rehabilitation medical care through a new generation of information technology. However, due to the complexity and diversity of rehabilitation medicine literature and the differences among rehabilitation patients, the task of entity extraction is currently facing many challenges. At home and abroad, the research on medical entity extraction has achieved certain results. However, relatively little research has been done on this specific area of rehabilitation medicine. This study aims to fill this gap and provide an effective entity extraction method for knowledge management in rehabilitation medicine.(1)In order to solve the problems in the recognition of rehabilitation medicine terms, this paper carried out BIO standardized labeling of rehabilitation medicine data. Considering the complexity of rehabilitation medicine terminology, this paper introduces a named entity recognition framework that combines pre-training models, Bi-LSTM, and CRF to improve the performance of entity recognition tasks.(2)In order to evaluate the effectiveness of the named entity recognition model, contrast and ablation experiments were set up in this study. The results of comparative experiments show that the model has achieved remarkable performance improvement in the entity extraction task of rehabilitation medicine. Ablation experiments further verified the effectiveness of each component in the model.
Graph convolution methods have achieved great success in the field of skeleton-based action recognition, but they have limitations in some complex actions. In medical actions, there are many interactions between hands, hands and head, and hands and feet such as putting on gloves and taking off glasses. To effectively identify this type of proximal action, we propose a novel network model: Node Edge Fusion Gated Graph Convolution Network (NEF-GGCN), which fully extracts the information about the joints in the skeleton graph and all the connecting edges as the feature input of the model. Specifically, we design a multi-dependent adjacency matrix to enhance the correlation of joints and fuse it with the adjacency matrix through a gated convolution block which can not only improve the dependency between the end joints but also preserve the dependency between the original related nodes. Furthermore, we capture the motion information of joints and bones in the skeleton using a temporal attention mechanism, which improves the local feature representation of graph nodes and adapts to more near-end actions. In addition, we propose a novel dataset PPE-DD12 to complement the existing datasets. We validate our approach by conducting extensive experiments and the results show that the proposed method achieves remarkable gains over some well-known methods.
With the development of display technology, people’s demand for high-definition video is increasing, and the requirements for video quality are also increasing. Nowadays, a large number of videos have emerged, and how to distinguish the quality of videos faces many challenges. Nowadays, deep learning is developing rapidly, and many video quality assessment (VQA) methods have emerged. These VQA methods can efficiently evaluate video quality and save a lot of manpower and material resources. To achieve the evaluation of UHD video quality, we propose an inter-frame downsampling UHD video quality assessment model (IFDU-VQA) based on deep learning. This model is a no-reference (NR) evaluation model that adopts a dual-branch structure: one branch uses a recursive network to deeply extract video inter-frame features, and combines inter-frame downsampling technology with a multi-information fusion (MIF) module to achieve different time and frequency information. Effective fusion; the other branch applies convolutional neural networks (CNN) to extract features from individual video frames. By fusing the outputs of the two branches, regression and pooling modules are used to calculate the final quality score of the video. Compared with multiple existing models, experimental results show that IFDU-VQA demonstrates superior performance on multiple standard data sets.
Blockchain technology, while revolutionary, presents significant challenges in terms of development complexity, interoperability, and security. This paper proposes a novel framework for blockchain application development that leverages microservice architecture (MSA) to address these challenges and aims to create a blockchain services marketplace. By abstracting blockchain-specific tasks into modular microservices, our framework simplifies the integration of blockchain functionalities, reduces development time, and enhances security. The framework promotes code reuse, improves flexibility, and separates blockchain and non-blockchain development tasks, enabling developers to focus on their areas of expertise. Furthermore, our approach allows for the flexible deployment of services, fostering innovation and the creation of unique applications across various blockchain platforms. The proposed blockchain services marketplace encourages blockchain developers to create and share well-crafted services that can be easily integrated by application developers. We demonstrate the framework’s effectiveness through a blockchain-based ideation and voting application, showcasing the ease of combining multiple blockchain services to create sophisticated use cases. Initial ongoing evaluations of the framework show promising results. These evaluations indicate that our framework can significantly advance blockchain application development by making it more accessible, efficient, and secure. By addressing the critical issues of interoperability and development complexity, our framework holds the potential to drive wider adoption and innovation in the blockchain space.
With the rapid development of Internet technology, leading to the evolution of malware variants, and malicious threats are also increasing. Therefore, rapid and accurate classification of malware becomes crucial. In response to this challenge, this paper proposes a malware classification method based on deep learning and visualization techniques. The method first extracts sample features, then converts malware binary samples into image format, thus transforming abstract binary data into images with visual information. Subsequently, color mapping technique is applied to convert the processed images into color images. Furthermore, to address imbalanced datasets, this paper introduces data augmentation methods to ensure more balanced and comprehensive training of the model across different sample categories. Finally, an improved convolutional neural network architecture is employed for the classification detection of malware families, achieving accurate classification and identification of different malware families through learning and analysis of sample features. To evaluate the effectiveness of the proposed method, experiments are conducted on the Google Code Jam (GCJ) benchmark dataset. The experimental results demonstrate an accuracy rate of 99.29% on the GCJ dataset. Additionally, we also compared the performance of the improved method before and after, and found that the proposed method can provide more comprehensive feature representation and significantly improve classification accuracy and detection efficiency.
With the globalization of the digital economy and the further development of cross-border trade, the cross-border flow of data is more and more frequent. In order to promote the secure sharing of cross-border data, this paper reviews the definition of cross-border data and the concepts of classification and grading, then gives the analysis and comparison of the current situation of cross-border data classification and grading in six main countries systematically, which is from the perspective of laws, regulations, and applications. With the construction requirements of the free trade ports in China, combined with existing laws and standards in China, a scheme proposal and its implementation process of the classification and grading of cross-border data in the free trade port are presented to improve the problems of weak operability and high costs of evaluation standards and existing technology deficiency for cross-border data. This study presents a comprehensive approach to the development of Chinese cross-border data regulations, with a particular focus on their role in facilitating the growth of the digital economy in free trade ports.
According to the needs and characteristics of the current medical care objects management mode, this project studies the construction of an intelligent management platform for medical care objects through the Internet of things(IoT) and digital twin technology. For the key technologies in the construction of the platform, the research method of technical route is proposed to effectively improve the management efficiency and quality, provide support for the behavior management, reasonable rehabilitation and accident alarm of medical care objects, and contribute a strong thrust to the sustainable development of intelligent management of medical care objects in China.
According to the characteristics of signal timing in industrial control system, a fault warning method based on TCN-MKELM is proposed. Firstly, an online reduction model based on time convolutional network (TCN) was established by using the historical operation data of the industrial control system, and the online prediction of each key signal of the industrial control system was carried out, and the residual data was generated; Secondly, a multi kernel extreme learning machine (MKELM) was constructed, and a fault early warning model based on MKELM is established by combining the residual data; Finally, the test verification is carried out by taking the operation data of the boiler temperature control system of a thermal power plant as an example. The experimental results show that the prediction model based on the TCN network has a smaller prediction error than the traditional recurrent neural network; Compared with the method using the original fault data for direct fault early warning, the MKELM fault early warning model based on the residual data has a higher accuracy rate.
In the era of mobile Internet and cloud computing, the rise of big data has profoundly impacted diverse industries and sectors, drawing significant scholarly attention. The objective of this paper is to delineate an overview of the international research landscape of big data, identifying key research hotspots and emerging trends. To gain a comprehensive overview of the current state of research in the field of big data, we employed a range of analytical techniques, including co-word analysis, clustering analysis, social network analysis, and dynamic strategy coordinate mapping. Our review encompassed articles published between 2017 and 2019, spanning a three-year period. The data were sourced from the SCI, SSCI, SPCI, and CPCI-SSH databases. The literature was subjected to a sysytematic analysis using Bibexcel, SPSS, and Gephi software, with the objective of uncovering distinct patterns and trends. The findings reveal 11 thematic clusters within international big data research. These are text mining, data fusion, distributed computing, Industry 4.0, data privacy, business intelligence, precision medicine, bioinformatics, and data mining techniques such as classification and clustering. It is noteworthy that Industry 4.0, data privacy, and text mining emerged as significant research hotspots. Furthermore, the fields of artificial intelligence, natural language processing, smart grids, energy big data, and wireless sensor networks are regarded as emerging areas of interest. It is anticipated that the convergence of bioinformatics and other pivotal areas of big data research, will serve as the impetus for future research hotspots, necessitating sustained scholarly investigation and discourse.
Secure two-party Newton interpolation refers to a process that two parties privately compute a polynomial over their coordinate pairs using Newton interpolation method. However, current solutions such as fully homomorphic encryption or garbled circuit remain inefficient. In this paper, we propose a specific solution to this problem based on oblivious polynomial evaluation. To improve the efficiency, we separate the coefficient construction phase and the observation value computing phase. By this means, the communication cost of computing one observation value is reduced to a constant level. And through comparison to solutions based on other interpolation method, we prove that our solution performs well in computing multiple values of the polynomial.
Aiming at filling missing triple structures, Knowledge Graph Completion (KGC) is a crucial task in knowledge graph reasoning. Especially in the judicial domain, it plays a significant role not only in enhancing the completeness of legal knowledge graphs but also in improving the accuracy of legal supervisory inference. Due to extensive legal expertise, complex entity relationships, and strict requirements for interpretable reasoning, traditional methods, such as embedding-based methods and those that utilize pre-trained language models (e.g., BERT), often fail to delve into deep semantic information or overlook the interconnections among triples, which leads to significant shortcomings in generalizability and capabilities in practical applications. Motivated by the great generation power of Large Language Models (LLMs), we propose a legal knowledge graph completion model based on Retrieval Augmented Generation (RAG), which we have named RA-KG-LLM. This model effectively combines the generative capabilities of LLMs with retrieval-augmented technology to enhance the semantic information and interrelations mining within knowledge graphs. The retrieval framework within our model utilizes text embeddings and vector similarity matching to provide the LLM with relevant triples, while fine-tuning techniques are employed to infuse the semantic information from the knowledge graph into the LLM. Extensive experiments on five real datasets demonstrate the effectiveness and potentiality in the judicial field of the proposed model. Specifically, on the legal domain dataset Cail2022, it achieves better Hits@1 score in the relation prediction task than the state-of-the-art related works.
In recent years, with the rapid development of network and computer technologies, the Governments of the United States, Japan, the United Kingdom and the European Union, as well as influential organizations, have successively carried out the construction of cyberspace ranges targeting the verification of new network technologies. New technology validation based on cyberspace ranges can solve the problems of high cost, difficult to manage iterative versions, and difficult to measure the accuracy of results. Based on the above advantages, this paper, in order to solve the problem of autonomous driving of cars in the park, which is difficult to drive safely under the influence of environmental uncertainties, completes the hardware design of intelligent cars in the cyberspace target range and proposes a Bayesian network-based design method of steering redundancy for hardware circuits, which divides the steering data of the two cars under different environmental factors into training set and test set. The Bayesian network parameters are predicted by the training set, and the prediction effect of the Bayesian network is tested by the test set. The prediction results show that the steering accuracy prediction of the smart car under different environmental factors is within the 85% confidence interval of the Bayesian network prediction, which provides a feasible way to design a safe and efficient underlying controller for smart cars.
