Please wait a minute...

当期目录

    15 September 2022, Volume 8 Issue 5
    TOPIC: DATA CIRCULATION AND PRIVACY COMPUTING
    Data tenancy: a new paradigm for data circulation
    Wenqiang RUAN, Mingxin XU, Xinyu TU, Lushan SONG, Weili HAN
    2022, 8(5):  3-11.  doi:10.11959/j.issn.2096-0271.2022071
    Asbtract ( 525 )   HTML ( 139)   PDF (1582KB) ( 451 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Data is becoming a new type of factor of production.How to compliantly and audibly circulate data among multiple parties is very important for data value formation.A novel data circulation paradigm, namely data tenancy, was proposed from the perspective of privacy preservation and data utilization.The motivation of data tenancy was discussed, and five requirements that data tenancy should satisfy were identified.Finally, a secret sharing-based data tenancy technique was proposed.

    Threats and defenses of federated learning: a survey
    Jianhan WU, Shijing SI, Jianzong WANG, Jing XIAO
    2022, 8(5):  12-32.  doi:10.11959/j.issn.2096-0271.2022038
    Asbtract ( 1713 )   HTML ( 246)   PDF (2537KB) ( 1833 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    With the comprehensive application of machine learning technology, data security problems occur from time to time, and people’s demand for privacy protection is emerging, which undoubtedly reduces the possibility of data sharing between different entities, making it difficult to make full use of data and giving rise to data islands.Federated learning (FL), as an effective method to solve the problem of data islands, is essentially distributed machine learning.Its biggest characteristic is to save user data locally so that the models’ joint training process won’t leak sensitive data of partners.Nevertheless, there are still many security risks in federated learning in reality, which need to be further studied.The possible attack means and corresponding defense measures were investigated in federal learning comprehensively and systematically.Firstly, the possible attacks and threats were classified according to the training stages of federal learning, common attack methods of each category were enumerated, and the attack principle of corresponding attacks was introduced.Then the specific defense measures against these attacks and threats were summarized along with the principle analysis, to provide a detailed reference for the researchers who first contact this field.Finally, the future work in this research area was highlighted, and several areas that need to be focused on were pointed out to help improve the security of federal learning.

    Research on privacy data security sharing scheme based on blockchain and function encryption
    Yi LI, Jinsong WANG, Hongwei ZHANG
    2022, 8(5):  33-44.  doi:10.11959/j.issn.2096-0271.2022072
    Asbtract ( 452 )   HTML ( 100)   PDF (1752KB) ( 443 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Blockchain technology has provided new ideas for data validation, data traceability, data trustworthiness, and data availability in data sharing, but privacy data security in data sharing still faces many challenges.Firstly, the current status of blockchain-based data sharing research was reviewed.Then a secure sharing model of privacy data was proposed.By encrypting the privacy data through function cryptography, and generating the proof of computational correctness through zero-knowledge proof technology, a secure and reliable data sharing with “data available but not visible” was realized.The experimental results show that the sharing delay and economic overhead of the model are within the acceptable range, which demonstrates the security and feasibility of the model.

    Research on privacy preservation of member inference attacks in online inference process for vertical federated learning linear model
    Hongshu YIN, Xuhua ZHOU, Wenjun ZHOU
    2022, 8(5):  45-54.  doi:10.11959/j.issn.2096-0271.2022056
    Asbtract ( 326 )   HTML ( 49)   PDF (2195KB) ( 227 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    With the development of big data and the introduction of data security regulations, the awareness of privacy protection has gradually increased, and the phenomenon of data isolation has become more and more serious.Federated learning technology as one of the effective methods to solve this problem has become a hot spot of concern.In the online inference process of vertical federated learning, the current mainstream methods do not consider the protection of data identity, which is easy to leak user privacy.A privacy protection method for member inference attacks was proposed in the online inference process of the vertical federated linear model.A filter with a false positive rate was constructed to avoid the accurate positioning of data identity to ensure the security of data.Homomorphic encryption was used to realize the full encrypted state of the online inference process and protect the intermediate calculation results.According to the ciphertext multiplication property of homomorphic encryption, the random number multiplication method was used to mask data, which ensured the security of the final inference result.This scheme further improved the security of user privacy in the online inference process of vertical federated learning and had lower computation overhead and communication costs.

    Exploration and practice of data quality governance in privacy computing scenarios
    Yan ZHANG, Yifan YANG, Ren YI, Shengmei LUO, Jianfei TANG, Zhengxun XIA
    2022, 8(5):  55-73.  doi:10.11959/j.issn.2096-0271.2022073
    Asbtract ( 497 )   HTML ( 127)   PDF (2649KB) ( 351 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Privacy computing is a new data processing technology, which can realize the transformation and circulation of a data value on the premise of protecting data privacy and security.However, the invisible feature of data in private computing scenarios poses a great challenge to traditional data quality management.There is still a lack of perfect solutions.To solve the above problems in the industry, a data quality governance method and process suitable for privacy computing scenarios were proposed.A local and multi-party data quality evaluation system was constructed, which could take into account the data quality governance of the local domain and the federal domain.At the same time, a data contribution measurement method was proposed to explore the long-term incentive mechanism of privacy computing, improve the data quality of privacy computing, and improve the accuracy of computing results.

    Exploration and practice of privacy preserving computing for vehicle-road collaboration system
    Ming LI, Abin LYU
    2022, 8(5):  74-87.  doi:10.11959/j.issn.2096-0271.2022069
    Asbtract ( 438 )   HTML ( 71)   PDF (2493KB) ( 643 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Based on the development of vehicle-road collaboration, the research progress of privacy computing, artificial intelligence, and other technologies for the vehicle-road collaboration scene was summarized.YITA-TFL platform was designed and implemented.A complete privacy protection scheme was provided for data management, model training, model management, and collaborative reasoning.And a model was established for the application of artificial intelligence combined with privacy computing in the transportation industry.

    STUDY
    Knowledge-enhanced policy-guided interactive reinforcement recommendation system
    Yuqi ZHANG, Xiaowen HUANG, Jitao SANG
    2022, 8(5):  88-105.  doi:10.11959/j.issn.2096-0271.2022033
    Asbtract ( 367 )   HTML ( 35)   PDF (3638KB) ( 380 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    The recommendation system is an important means to solve the problem of information overload in social media.To solve the problem that traditional recommendation systems cannot optimize the longterm user experience, researchers have proposed the interactive recommendation system and tried to use deep reinforcement learning to optimize the strategy of recommendation.However, the reinforcement recommendation algorithm faces problems such as sparse feedback, learning from zero which damages the user experience, and large item space.To solve the above problems, an improved interactive reinforcement recommendation model KGP-DQN was proposed.The model constructed a behavioral knowledge graph representation module, which combines user historical behavior and knowledge graph to solve the problem of sparse feedback.The model constructed a strategy initialization module to provide an initialization strategy for the reinforcement recommendation system based on user historical behaviors to solve the problem of learning from zero.The model constructed the candidate select module which creates candidates by dynamic clustering based on the item representation on the behavioral knowledge graph to solve the problem of large action space.The experiments were conducted on three real-world datasets.The experimental results show that the KGP-DQN method can quickly and effectively train the reinforcement recommendation system and its recommendation accuracy on three datasets is more than 80%.

    Extraction and visualization analysis of key elements of tax preferential policies
    Haishan GUAN, Yulong ZHENG, Bifan WEI, Zemin ZHANG, Hao YUE, Bin SHI, Bo DONG
    2022, 8(5):  106-123.  doi:10.11959/j.issn.2096-0271.2022035
    Asbtract ( 358 )   HTML ( 40)   PDF (3860KB) ( 362 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    With the rapid increase in the number of preferential tax policies, taxpayers face a large number of preferential tax policies, and it is difficult for taxpayers to quickly locate the preferential content related to them.As a result, many taxpayers do not enjoy the preferential policies they should enjoy.Based on the combination of pre-training language model BERT and rule processing, the visualization was realized of the characterization of preferential tax policies and regulations, the extraction of key elements, and the visual query of tax incentives, so that taxpayers can intuitively and quickly locate tax incentives related to themselves, and visualize the results.The experimental results show that the extraction performance of key elements is superior, and the query of preferential tax policies is quick and intuitive, which can effectively alleviate the problem of massive tax preferential information overload.

    Research on trend prediction of time-coded LSTM based public opinion hot spots in universities
    Jie YI, Tengfei CAO, Mingfeng HUANG, Xiaohan HUANG, Zizhen ZHANG
    2022, 8(5):  124-138.  doi:10.11959/j.issn.2096-0271.2022034
    Asbtract ( 301 )   HTML ( 51)   PDF (3525KB) ( 542 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    With the development of Internet technology, network public opinion hot information can be quickly spread in a short time.Predicting the development trend of public opinion hot spots is helpful to the analysis and management of college students’ ideological health, and it is also an important issue in the field of network public opinion information research.Aiming at the public opinion information text in microblog, the hot spots trend prediction model of universities based on time-coded long short-term memory (LSTM) was constructed.Compared with the prediction effect of support vector machine, and recurrent neural network through experiments, the superiority of time-coded LSTM was verified.Finally, the prediction effect of time-coded LSTM was evaluated by using the real-time public opinion events of colleges and universities in microblogs, and the evaluation parameters were dynamically adjusted to optimize the performance of the evaluation, and the prediction effect was improved significantly.

    Classification algorithm for imbalance data of ECG based on PSOFS and TSK fuzzy system
    Xinhui LI, Qing SHEN, Xiongtao ZHANG
    2022, 8(5):  139-152.  doi:10.11959/j.issn.2096-0271.2022039
    Asbtract ( 217 )   HTML ( 22)   PDF (2124KB) ( 178 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    A new classification model of electrocardiogram (ECG) signal based on particle swarm optimization feature selection (PSOFS) and TSK (Takagi-Sugeno-Kang) fuzzy system was proposed, i.e., parallel ensemble fuzzy neural network based on PSOFS and TSK (PE-PT-FN), which was used for ECG prediction.Each class sample in the training set was randomly sampled, and the samples obtained by randomly sampled were added.Then, the feature selection method PSOFS was carried out independently and parallelly.In PSOFS, particles that were random initial positions represent different feature subsets and converge to the optimal positions after many iterations.Each subset had a corresponding feature subset.Several groups of TSK fuzzy neural network (TSK-FNN) were trained by each feature subset in parallel.Medical researchers could effectively find the correlation between ECG signal data and different types of disease through the interpretability of the fuzzy system and the feature subsets by the PSOFS algorithm.Experiments prove that PE-PT-FN greatly improves the macro-R to 92.35% while retaining interpretability.

    FORUM
    Construction of data rights system from information-data two-dimensional perspective
    Qin GU, Tao ZHOU, Shuli ZHONG, Zhimei QIN, Yaoyao ZHANG, Yi CHEN
    2022, 8(5):  153-169.  doi:10.11959/j.issn.2096-0271.2022074
    Asbtract ( 315 )   HTML ( 58)   PDF (1831KB) ( 471 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    The data rights confirmation system has become a fundamental issue in establishing a market-based system for data elements.From a dual-dimensional perspective, the distinctions between information and data regarding their concepts and characteristics were discussed, on this basis, a method was proposed to construct a data-right confirmation system by classifying information subjects and data management subjects.This procedure underlies the construction of the data rights confirmation system.Studies show that: (1) information is content substance while data is the carrier of information.Individuals, organizations, and other entities are the subjects of information content production, and individuals, and organizations are the owners of information; (2) Individuals or organizations have the rights and obligations of data management for the data formed by recording information in some form; (3) By the current legal framework and objective practice, data ownership shall belong to the State, while data management subjects have limited rights to possess, use, seek profits from and dispose of data, meanwhile ought to perform corresponding obligations to protect the rights and interests of information subjects.

    COLUMN: FOCUS ON LOCAL GOVERNMENT BIG DATA
    SCINECE POPULARIZATION
Most Download
Most Read
Most Cited