检索结果-内蒙古大学图书馆

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Qiankun Pi Jicang Lu Yepeng Sun Qinlong Fan Xukun Zhou Shouxin Shang State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou China School of Information Renmin University of China Beijing China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Stance detection, a natural language processing technique, captures user attitudes on controversial social media topics. However, the semantic ambiguity of social texts makes accurate stance determination challenging. Existing annotated data is often domain-specific, resulting in poor model generalization for unseen targets and cross-domain scenarios. To tackle this challenge, we introduce inference tasks related to stance detection as auxiliary tasks and use a multitask learning approach to improve the model’s understanding of textual semantics. Meanwhile, to alleviate the scarcity of annotated data, we use argumentation corpus with abundant resources as the source domain to train the basic stance detection model. We integrate this model with a large-scale framework to facilitate weakly supervised learning for machine annotation of unlabeled social texts, thereby boosting the model’s adaptability across different targets and domains. Experimental results on four Twitter datasets demonstrate that our method can significantly improve the model’s stance discriminative ability and generalization performance.

关键词： Adaptation models Weak supervision Social networking (online) Annotations Semantics Blogs Transfer learning Signal processing Data models Speech processing

来源：评论

学校读者我要写书评

暂无评论

OpenMP automatic translation framework for Sunway TaihuLight 3

OpenMP automatic translation framework for Sunway TaihuLight

引用

3rd IEEE International Conference on Communications, Information System and Computer engineering, CISCE 2021

作者： Li, Mingliang Pang, Jianmin Yue, Feng Wang, Qihan State Key Laboratory of Mathematical Engineering and Advanced Computing Henan Zhengzhou China

ISBN: (纸本)9780738112152

The Sunway processor is a unique heterogeneous many-core processor used by Sunway TaihuLight supercomputer. However, developing parallel programs on the Sunway processor is still complex. In this paper, a source-to-source translation framework for the Sunway heterogeneous many-core processors is designed and implemented, which can transform OpenMP programs into many-core accelerated programs amenable to the Sunway architecture. Considering differences between the OpenMP shared-memory programming model and the Sunway programming model, a translation model has been proposed for converting OpenMP programs. A set of runtime functions based on the Sunway architecture has been implemented, including a tree-based reduction algorithm and an effective distributed lock mechanism. The experimental results show that the translation framework can correctly translate OpenMP programs, and the proposed lock mechanism reduces the overhead by up to 90% against the baseline. © 2021 IEEE.

关键词： Supercomputers

来源：评论

学校读者我要写书评

暂无评论

Network Device Identification Based on MAC Boundary Inference 7th

Network Device Identification Based on MAC Boundary Inferenc...

引用

7th International Conference on Artificial Intelligence and Security, ICAIS 2021

作者： Guo, Xinmiao Li, Xiangnan Li, Ruixiang Wang, Xiuting Luo, Xiangyang State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou450000 China

ISBN: (纸本)9783030786205

Network device is an important part of cyberspace, and accurate identification of network device is the basis of network management and security analysis. The current identification method based on MAC address is rely on converting MAC address and other information into fingerprints and MAC address distance to identify network device eventually. However, this identification method based on MAC address distance has high false alarm rate. A method for network device identification based on MAC boundary inference is proposed. Considering the device manufacturers’ strategies of allocating MAC addresses in sequence for devices with same type, the relationship between the type and MAC address is built according known devices firstly. Then, MAC address aggregation rule is built to infer the MAC prefix for those known type, and the MAC boundary is obtained. Finally, the type of target network device is identified by matching target MAC with MAC prefix, or calculating the distance between target MAC with the MAC boundary. The experimental result in simulation dataset show that the identification method proposed in this paper is better than the identification method based on MAC address distance significantly, and then been less affected by the distribution of MACs of known devices. And the experimental result in Cisco device dataset show that our method increases the identification accuracy rate of the identification method based on MAC address distance by 11.9%. © 2021, Springer Nature Switzerland AG.

关键词： Computers

来源：评论

学校读者我要写书评

暂无评论

Research on key Technology of Industrial Network Boundary Protection based on Endogenous Security 4

Research on Key Technology of Industrial Network Boundary Pr...

引用

4th IEEE advanced Information Management, Communicates, Electronic and Automation Control Conference, IMCEC 2021

作者： Yu, Fei Wei, Qiang Geng, Yangyang Wang, Yunchao State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou450001 China

ISBN: (纸本)9781728185347

Industrial network boundary protection equipment faces threats from attackers when protecting the industrial control system network. The similarity and static characteristics caused by large-scale and long-term deployment determine that it could only defend against known attacks but could not deal with unknown APT threats, which leads to the breakthrough of one defense line is equivalent to the breakthrough of all defense lines and may bring challenges to industrial production safety. This paper proposes a mimic defense model of industrial isolation gateway based on endogenous security. With the dynamic scheduling mechanism to transform the attack surface, the gateway selects multiple heterogeneous filter executors to process the same packet simultaneously. By comparing the processing results of each executor, anomaly detection is carried out to realize the dynamic defense of the industrial isolation gateway. The experimental results show that the industrial isolation gateway based on mimic architecture can significantly increase the difficulty of backdoor utilization, such as paralysis, rule tampering, and information theft, and effectively defend the industrial control system from the threats caused by the backdoors and vulnerabilities of the isolation gateway while exerting the normal boundary protection function. © 2021 IEEE.

关键词： Anomaly detection

来源：评论

学校读者我要写书评

暂无评论

IPv6 Landmark Mining Based on Domain Name Screening and IPv4-IPv6 Connection 21

IPv6 Landmark Mining Based on Domain Name Screening and IPv4...

引用

5th International Conference on Electronic Information Technology and Computer engineering, EITCE 2021

作者： Liu, Minghao Sun, Yuchen Liu, Chong Zu, Shuodi Zhang, Fan State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou Henan China

ISBN: (纸本)9781450384322

Existing IP landmark mining technologies mainly focus on IPv4, and less attention is paid to the association between IPv4 and IPv6. The current IPv6 network has fewer reliable landmarks with limited accuracy, while the IPv4 network already has a large number of reliable landmarks. If IPv4 landmarks can be used in IPv6 landmark mining, the accuracy and efficiency of IPv6 landmark acquisition will be improved. To this end, this paper proposes an IPv6 landmark mining method based on domain name screening and IPv4-IPv6 connection. This method first uses the domain names of government agencies, top universities, and top websites as seed domain names, and continuously obtains a large number of domain names through web crawlers;then performs DNS resolution on the obtained domain names to obtain their corresponding IPv4 address and IPv6 address;and finally evaluate and analyze whether the two addresses are located in the same location, and determine whether to use it as an IPv6 landmark. This method successfully associates the IPv4 geolocation with the IPv6 geolocation, and uses the IPv4 landmarks to obtain and evaluate the IPv6 landmarks. Experimental results show that this method can obtain city-level IPv6 network landmarks more efficiently, and it also has a certain ability in street-level landmark mining. © 2021 ACM.

关键词： Mining

来源：评论

学校读者我要写书评

暂无评论

Anti-obfuscation Binary Code Clone Detection Based on Software Gene 7th

Anti-obfuscation Binary Code Clone Detection Based on Softwa...

引用

7th International Conference of Pioneering Computer Scientists, Engineers and Educators, ICPCSEE 2021

作者： Tang, Ke Liu, Fudong Shan, Zheng Zhang, Chunyan State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou450001 China

ISBN: (纸本)9789811659393

Information technology facilitates people’s lives greatly, while it also brings many security issues, such as code plagiarism, software in-fringement, and malicious code. In order to solve the problems, reverse engineering is applied to analyze abundant binary code manually, which costs a lot of time. However, due to the maturity of different obfuscation techniques, the disassembly code generated from the same function differs greatly in the opcode and control flow graph through different obfuscation options. This paper propose a method inspired by natural language processing, to realize the semantic similarity matching of binary code in basic block granularity and function granularity. In the similarity matching task of binary code obtained by different obfuscation options of LLVM, the indicator reaches 99%, which is better than the existing technologies. © 2021, Springer Nature Singapore Pte Ltd.

关键词： Genes

来源：评论

学校读者我要写书评

暂无评论

Tuning 2D magnetism in Fe3+XGeTe2 films by element doping

引用

National Science Review 2022年第6期9卷 163-170页

作者： Shanshan Liu Zihan Li Ke Yang Enze Zhang Awadhesh Narayan Xiaoqian Zhang Jiayi Zhu Wenqing Liu Zhiming Liao Masaki Kudo Takaaki Toriyama Yunkun Yang Qiang Li Linfeng Ai Ce Huang Jiabao Sun Xiaojiao Guo Wenzhong Bao Qingsong Deng Yanhui Chen Lifeng Yin Jian Shen Xiaodong Han Syo Matsumura Jin Zou Yongbing Xu Xiaodong Xu Hua Wu Faxian Xiu State Key Laboratory of Surface Physics and Department of Physics Fudan University Institute for Nanoelectronic Devices and Quantum Computing Fudan University College of Science University of Shanghai for Science and Technology Laboratory for Computational Physical Sciences(MOE) Fudan University Solid State and Structural Chemistry Unit Indian Institute of Science School of Electronic Science and Engineering Nanjing University Department of Physics University of Washington Department of Electronic Engineering Royal Holloway University of London Materials Engineering The University of Queensland Beijing Key Laboratory of Microstructure and Property of Advanced Materials Institute of Microstructure and Property of Advanced Materials Beijing University of Technology The Ultramicroscopy Research Center Kyushu University State Key Laboratory of ASIC and System School of Microelectronics Fudan University Collaborative Innovation Center of Advanced Microstructures Department of Applied Quantum Physics and Nuclear Engineering Kyushu University Centre for Microscopy and Microanalysis The University of Queensland Shanghai Research Center for Quantum Sciences

Two-dimensional（2D） ferromagnetic materials have been discovered with tunable magnetism and orbital-driven nodal-line features. Controlling the 2D magnetism in exfoliated nanoflakes via electric/magnetic fields enables a boosted Curie temperature（TC） or phase transitions. One of the challenges, however, is the realization of high TC2D magnets that are tunable, robust and suitable for large scale fabrication. Here, we report molecular-beam epitaxy growth of wafer-scale Fe3+XGeTe2films with TCabove room temperature. By controlling the Fe composition in Fe3+XGeTe2, a continuously modulated TCin a broad range of 185–320 K has been achieved. This widely tunable TCis attributed to the doped interlayer Fe that provides a 40% enhancement around the optimal composition X = 2. We further fabricated magnetic tunneling junction device arrays that exhibit clear tunneling signals. Our results show an effective and reliable approach, i.e. element doping, to producing robust and tunable ferromagnetism beyond room temperature in a large-scale 2D Fe3+XGeTe2fashion.

关键词： 2D ferromagnetic material Fe3+XGeTe2 film element doping above room temperature TC tunability

来源：评论

学校读者我要写书评

暂无评论

Multi Association Semantics-Based User Matching Algorithm Without Prior Knowledge

Multi Association Semantics-Based User Matching Algorithm Wi...

引用

International Conference on Image, Vision and Intelligent Systems, ICIVIS 2021

作者： Jiang, Qiuyan Gong, Daofu State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou450001 China College of Science and Information Qingdao Agricultural University Qingdao266001 China

ISBN: (纸本)9789811669620

Cross-network user matching is one of the basic issues for realizing social network data integration. Existing research based on structure features provides a good matching method for nodes with high-degree, but ignores the matching of nodes with low values. This paper proposes an unsupervised cross-network user matching algorithm based on association semantics to solve the problem of cross-network user matching. It can effectively solve the low-degree user matching problem when the quality of the text type attributes of network users cannot be guaranteed. First, build a social network graph based on multiple types of user behaviors, and define different association semantics according to different behaviors;then, walk based on the association semantics to obtain user node sequences with association semantics, and then combine the network embedding model to learn the user feature vector representation is obtained;finally, the user similarity is measured based on the user vector similarity, and then cross-network user matching is realized. The highlights of this paper are two folds. (1) We treat the association as nodes. (2) Social network user graphs with association semantics can enrich structure feature;walks based on association semantics can obtain walks that are closer to natural language short sentences. The experimental results show that the proposed method can effectively match user with similar behavior characteristics without any prior knowledge, and can effectively achieve the matching of low-degree nodes in heterogeneous networks. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Directed Association-based User Identification Algorithm 2

Directed Association-based User Identification Algorithm

引用

2nd International Conference on Computer, Big Data and Artificial Intelligence, ICCBDAI 2021

作者： Jiang, Qiuyan Gong, Daofu Liu, Fenlin State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou450001 China College of Science and Information Qingdao Agricultural University Qingdao266001 China

This paper studies the social network with directed association semantics, and proposes an unsupervised user identification algorithm (DAUM-P). The algorithm construct a user associations graph based on multiple types of directed user behaviors, and define directed associations;then, walk based on directed associations to obtain a walking sequence, and then combine the network embedding model to learn user characteristics vector representation;finally, the identification result is given based on the representation vector distance. Finally, the experiments show that the user identification effect is better than that of the baseline method. © 2022 Institute of Physics Publishing. All rights reserved.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Malicious Traffic Detection Model Refactor Method Based on Adaptive Sample Generation 5

Malicious Traffic Detection Model Refactor Method Based on A...

引用

5th IEEE advanced Information Management, Communicates, Electronic and Automation Control Conference, IMCEC 2022

作者： Jiao, Liang Fei, Jinlong Zhao, Min State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou China Tianjin University of Science and Technology School of Electronic Information and Automation Tianjin China

ISBN: (纸本)9781665479677

As machine learning models are widely used in malicious traffic detection, model security has become an important research topic. Aiming at the problems that the model refactor attack methods need a large amount of prior knowledge of the target model, rely on a large amount of real data, and have a narrow scope of application, a model refactor method based on adaptive sample generation was proposed. By means of an adaptive sample generation algorithm based on graph retrieval, a small number of real samples are used to quickly generate generated samples uniformly distributed in the input domain. The boundary of the refactored model is corrected according to the decision of the target model, the refactored model is iteratively trained by combining the network traffic feature selection method and the model hyperparameter optimization method, and the model refactor attack is finally realized. The refactor attack experiment against five malicious traffic detection machine learning models shows that the refactor rate and accuracy of the refactored model increase gradually with the increase of iteration rounds, and the refactor rate is more than 90%, using only the target model output label. © 2022 IEEE.

关键词： Iterative methods

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：