Pub Date : 2024-03-01DOI: 10.53106/160792642024032502013
C. C. Charles Chen, Sai-Sai Shi Charles Chen, Sheng-Lung Peng Sai-Sai Shi
This paper mainly focuses on building the knowledge graph of semiconductor industry chain. The main research contents include knowledge extraction, knowledge storage, and construction of knowledge graph in semiconductor field. The crawler technology and character recognition technology are used to obtain semiconductor industry chain information from the Internet, magazines, and institutions to establish the original data set. Then, Lattice Long Short-Term Memory (Lattice-LSTM) model is used to implement the entity extraction and recognition. The piecewise convolutional neural network (PCNN) model based on the sentence-level attention mechanism is used to extract relationships and obtain entity triples. The semiconductor dictionary library is constructed through the obtained structured data. The dictionary library and Chinese natural language toolkit HanLP are combined to annotate unstructured text data for knowledge extraction. Neo4j graph database is used to store the extracted data of semiconductor industry chain. Finally, Spring Boot and Vue technology are used to create a knowledge graph system.
本文主要关注半导体产业链知识图谱的构建。主要研究内容包括半导体领域的知识提取、知识存储和知识图谱构建。利用爬虫技术和字符识别技术从互联网、杂志和机构中获取半导体产业链信息，建立原始数据集。然后，使用 Lattice Long Short-Term Memory（Lattice-LSTM）模型实现实体提取和识别。基于句子级关注机制的片断卷积神经网络（PCNN）模型用于提取关系并获得实体三元组。通过获得的结构化数据构建半导体词典库。字典库与中文自然语言工具包 HanLP 相结合，对非结构化文本数据进行注释，以提取知识。使用 Neo4j 图数据库存储提取的半导体产业链数据。最后，使用 Spring Boot 和 Vue 技术创建知识图谱系统。
{"title":"A Construction of Knowledge Graph for Semiconductor Industry Chain Based on Lattice-LSTM and PCNN Models","authors":"C. C. Charles Chen, Sai-Sai Shi Charles Chen, Sheng-Lung Peng Sai-Sai Shi","doi":"10.53106/160792642024032502013","DOIUrl":"https://doi.org/10.53106/160792642024032502013","url":null,"abstract":"\u0000 This paper mainly focuses on building the knowledge graph of semiconductor industry chain. The main research contents include knowledge extraction, knowledge storage, and construction of knowledge graph in semiconductor field. The crawler technology and character recognition technology are used to obtain semiconductor industry chain information from the Internet, magazines, and institutions to establish the original data set. Then, Lattice Long Short-Term Memory (Lattice-LSTM) model is used to implement the entity extraction and recognition. The piecewise convolutional neural network (PCNN) model based on the sentence-level attention mechanism is used to extract relationships and obtain entity triples. The semiconductor dictionary library is constructed through the obtained structured data. The dictionary library and Chinese natural language toolkit HanLP are combined to annotate unstructured text data for knowledge extraction. Neo4j graph database is used to store the extracted data of semiconductor industry chain. Finally, Spring Boot and Vue technology are used to create a knowledge graph system.\u0000 \u0000","PeriodicalId":442331,"journal":{"name":"網際網路技術學刊","volume":"49 44","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140400554","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Various levels of government across Taiwan are eager to promote the establishment of long-term care residential facilities to meet the significant caregiving needs arising from the wave of population aging. However, the successful establishment of an effective mechanism relies on proper supervision and guidance. Therefore, implementing a value assessment system for long-term care service quality management is of paramount importance. Using multi-criteria decision-making (MCDM) approach can provide effective conditions for the establishment of such system and enable a more comprehensive and objective evaluation of long-term care service quality. Using this system, decision-makers can incorporate different indicators based on various needs and weights to evaluate the quality and performance of long-term care services. This facilitates the determination of priorities and the formulation of improvement strategies, thereby enhancing the quality of long-term care services. This study develops an information-based assessment model for a platform that is win-win for both institutions and individuals. The model incorporates consumer reputation and environmental social governance (ESG) dimensions, in addition to indicators such as operational and management efficiency, professional care quality, safety and environmental facilities, and protection of individual rights and interests. Further, it integrates multiple indicator items and employs the analytic hierarchy process (AHP) to decompose and structure complex multi-dimensional issues, thereby aligning itself with current corporate evaluations. It aims to assist care service agencies in making key service quality decisions across different dimensions, and enhance the overall quality and competitiveness of those agencies, while increasing public trust and recognition in the evaluation of care service quality.
{"title":"Designing a Multi-Criteria Decision-Making Framework to Establish a Value Ranking System for the Quality Evaluation of Long-Term Care Services","authors":"Lun-Ping Hung Lun-Ping Hung, Weidong Huang Lun-Ping Hung, Sheng-Tzong Cheng Weidong Huang, Zong-Jie Wu Sheng-Tzong Cheng, Syuan Ou Yang Zong-Jie Wu","doi":"10.53106/160792642024032502010","DOIUrl":"https://doi.org/10.53106/160792642024032502010","url":null,"abstract":"\u0000 Various levels of government across Taiwan are eager to promote the establishment of long-term care residential facilities to meet the significant caregiving needs arising from the wave of population aging. However, the successful establishment of an effective mechanism relies on proper supervision and guidance. Therefore, implementing a value assessment system for long-term care service quality management is of paramount importance. Using multi-criteria decision-making (MCDM) approach can provide effective conditions for the establishment of such system and enable a more comprehensive and objective evaluation of long-term care service quality. Using this system, decision-makers can incorporate different indicators based on various needs and weights to evaluate the quality and performance of long-term care services. This facilitates the determination of priorities and the formulation of improvement strategies, thereby enhancing the quality of long-term care services. This study develops an information-based assessment model for a platform that is win-win for both institutions and individuals. The model incorporates consumer reputation and environmental social governance (ESG) dimensions, in addition to indicators such as operational and management efficiency, professional care quality, safety and environmental facilities, and protection of individual rights and interests. Further, it integrates multiple indicator items and employs the analytic hierarchy process (AHP) to decompose and structure complex multi-dimensional issues, thereby aligning itself with current corporate evaluations. It aims to assist care service agencies in making key service quality decisions across different dimensions, and enhance the overall quality and competitiveness of those agencies, while increasing public trust and recognition in the evaluation of care service quality.\u0000 \u0000","PeriodicalId":442331,"journal":{"name":"網際網路技術學刊","volume":"221 4","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140402790","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
To protect the increasing cyberspace assets, attack detection systems (ADSs) as well as intrusion detection systems (IDSs) have been equipped in various network environments. Recently, with the development of big data, machine learning, deep learning, neural networks and other artificial intelligence (AI) technologies, more and more ADSs/IDSs based on Artificial Intelligence are presented in academia and industry. Particularly, depending on the outstanding performance and efficiency in recognizing and classifying images, computer vision algorithms have been employed to detect malicious software and malicious traffic. However, we found that in wireless networks, the results vary significantly depending on the mapping methods used to transform the original network traffic data into visual images. Therefore, in this paper, we propose an AI-based attack detection scheme (TV-ADS) by introducing a novel traffic-image mapping method, which segments the sequential network traffic into individual event cells and transforms variant images to a uniform standard size, and design a CNN model to recognize normal and malicious traffics with these visible network event images. Finally, the results of our experiments on the AWID3 dataset demonstrate that our TV-ADS outperforms the existing schemes in terms of accuracy, precision, recall, F1-score and efficiency.
{"title":"TV-ADS: A Smarter Attack Detection Scheme Based on Traffic Visualization of Wireless Network Event Cell","authors":"Zhiwei Zhang Zhiwei Zhang, Guiyuan Tang Zhiwei Zhang, Baoquan Ren Guiyuan Tang, Baoquan Ren Baoquan Ren, Yulong Shen Baoquan Ren","doi":"10.53106/160792642024032502012","DOIUrl":"https://doi.org/10.53106/160792642024032502012","url":null,"abstract":"\u0000 To protect the increasing cyberspace assets, attack detection systems (ADSs) as well as intrusion detection systems (IDSs) have been equipped in various network environments. Recently, with the development of big data, machine learning, deep learning, neural networks and other artificial intelligence (AI) technologies, more and more ADSs/IDSs based on Artificial Intelligence are presented in academia and industry. Particularly, depending on the outstanding performance and efficiency in recognizing and classifying images, computer vision algorithms have been employed to detect malicious software and malicious traffic. However, we found that in wireless networks, the results vary significantly depending on the mapping methods used to transform the original network traffic data into visual images. Therefore, in this paper, we propose an AI-based attack detection scheme (TV-ADS) by introducing a novel traffic-image mapping method, which segments the sequential network traffic into individual event cells and transforms variant images to a uniform standard size, and design a CNN model to recognize normal and malicious traffics with these visible network event images. Finally, the results of our experiments on the AWID3 dataset demonstrate that our TV-ADS outperforms the existing schemes in terms of accuracy, precision, recall, F1-score and efficiency.\u0000 \u0000","PeriodicalId":442331,"journal":{"name":"網際網路技術學刊","volume":"201 4","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140403411","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-01DOI: 10.53106/160792642024032502002
Hyeon Jo Hyeon Jo, Eun-Mi Baek Hyeon Jo
This study explores the key factors determining the success of online learning during the COVID-19 isolation. The analytical framework clarifies the role of social distancing attitude, social distancing intention, attitude towards online learning, and perceived value in developing user satisfaction. The research model was validated by fitting data gathered from 490 students in Asian countries through structural equation modeling (SEM). The results indicated that both social distancing attitude and social distancing intention do not have a significant impact on user satisfaction. The findings showed that both attitude and perceived value are significantly associated with user satisfaction. Risk perception affects social distancing attitude, attitude toward online learning, and perceived value. However, it does not impact social distancing intention. Cabin fever syndrome positively affects social distancing attitude while it negatively influences social distancing intention. The results would help to understand online learning success and to improve education strategies, explained by students’ perception of COVID-19.
{"title":"Navigating Online Learning Satisfaction in the Age of COVID-19: An Examination of Key Influencing Factors","authors":"Hyeon Jo Hyeon Jo, Eun-Mi Baek Hyeon Jo","doi":"10.53106/160792642024032502002","DOIUrl":"https://doi.org/10.53106/160792642024032502002","url":null,"abstract":"\u0000 This study explores the key factors determining the success of online learning during the COVID-19 isolation. The analytical framework clarifies the role of social distancing attitude, social distancing intention, attitude towards online learning, and perceived value in developing user satisfaction. The research model was validated by fitting data gathered from 490 students in Asian countries through structural equation modeling (SEM). The results indicated that both social distancing attitude and social distancing intention do not have a significant impact on user satisfaction. The findings showed that both attitude and perceived value are significantly associated with user satisfaction. Risk perception affects social distancing attitude, attitude toward online learning, and perceived value. However, it does not impact social distancing intention. Cabin fever syndrome positively affects social distancing attitude while it negatively influences social distancing intention. The results would help to understand online learning success and to improve education strategies, explained by students’ perception of COVID-19.\u0000 \u0000","PeriodicalId":442331,"journal":{"name":"網際網路技術學刊","volume":"83 2","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140405295","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-01DOI: 10.53106/160792642024032502009
Nai-Wei Lo Nai-Wei Lo, Chi-Ying Chuang Nai-Wei Lo, Jia-Ning Luo Chi-Ying Chuang, Chong-Long Yang Jia-Ning Luo
In recent years, there has been rapid development in vehicle safety technology, with the emergence of various active safety systems including blind spot information systems, adaptive cruise control, and front collision warning systems. Simultaneously, car manufacturers and technology companies are actively exploring technologies in the realm of autonomous driving. To facilitate such applications, vehicles are required to communicate with each other, exchanging vital information such as position, speed, and acceleration. However, this exchange of information poses a potential risk of drivers’ personal data being compromised. For safety purposes, vehicles must undergo appropriate authentication before engaging in communication with other vehicles. Ensuring this authentication process maintains the anonymity of the vehicles is crucial. Yet, striking a balance between protecting vehicle anonymity and enabling vehicle identification when necessary remains a challenging issue. This paper introduces a multi-tier Vehicular Ad-Hoc Network (VANET) framework designed to uphold the conditional anonymity and traceability of vehicles. The implementation of a group signature mechanism facilitates anonymous authentication, thereby enabling the realization of conditional anonymity and traceability. Moreover, comprehensive simulations and security analyses were conducted to validate the effectiveness of the proposed framework, demonstrating its efficiency while incorporating robust safety considerations.
{"title":"Vehicle Identity Anonymity Framework with Accountability for VANET Environment","authors":"Nai-Wei Lo Nai-Wei Lo, Chi-Ying Chuang Nai-Wei Lo, Jia-Ning Luo Chi-Ying Chuang, Chong-Long Yang Jia-Ning Luo","doi":"10.53106/160792642024032502009","DOIUrl":"https://doi.org/10.53106/160792642024032502009","url":null,"abstract":"\u0000 In recent years, there has been rapid development in vehicle safety technology, with the emergence of various active safety systems including blind spot information systems, adaptive cruise control, and front collision warning systems. Simultaneously, car manufacturers and technology companies are actively exploring technologies in the realm of autonomous driving. To facilitate such applications, vehicles are required to communicate with each other, exchanging vital information such as position, speed, and acceleration. However, this exchange of information poses a potential risk of drivers’ personal data being compromised. For safety purposes, vehicles must undergo appropriate authentication before engaging in communication with other vehicles. Ensuring this authentication process maintains the anonymity of the vehicles is crucial. Yet, striking a balance between protecting vehicle anonymity and enabling vehicle identification when necessary remains a challenging issue. This paper introduces a multi-tier Vehicular Ad-Hoc Network (VANET) framework designed to uphold the conditional anonymity and traceability of vehicles. The implementation of a group signature mechanism facilitates anonymous authentication, thereby enabling the realization of conditional anonymity and traceability. Moreover, comprehensive simulations and security analyses were conducted to validate the effectiveness of the proposed framework, demonstrating its efficiency while incorporating robust safety considerations.\u0000 \u0000","PeriodicalId":442331,"journal":{"name":"網際網路技術學刊","volume":"12 4","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140404412","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-01DOI: 10.53106/160792642024032502014
Zheyi Wen Zheyi Wen, Chenlu Ye Zheyi Wen, Ming Zhao Chenlu Ye, Fang-Chuan Ou Yang Ming Zhao
Deep convolutional neural networks have achieved good performance in the application of computer vision, but there are also problems, such as a large amount of computation, time consuming, and high memory demand. In this paper, a depthwise separable convolution filter pruning method based on PCA is proposed. First, this paper uses depthwise separable convolution to replace the conventional convolution in ResNet to reduce the number of parameters and the amount of computation in the network. The specific operation process is to first use depthwise convolution to separate the spatial dimension to increase the network width and expand the range of feature extraction, and then use pointwise convolution to reduce the computational complexity of conventional convolution operation. Second, PCA is used to distinguish stacked similar filters and perform dimensionality reduction, which not only alleviates the dimensional disaster, but also achieves compression of data and minimizes information loss. Experimental results show that this method can significantly improve the calculation speed and accuracy of the deep convolutional neural network model, and further compress the model size. On the clinical Color Perception Test Chart, this method reduced the amount of model parameters and MACs on ResNet by about 91% while maintaining the test accuracy at about 95%. With almost no loss of accuracy, this method greatly shortened the running time of the model.
{"title":"A Compact Depth Separable Convolutional Image Filter for Clinical Color Perception Test","authors":"Zheyi Wen Zheyi Wen, Chenlu Ye Zheyi Wen, Ming Zhao Chenlu Ye, Fang-Chuan Ou Yang Ming Zhao","doi":"10.53106/160792642024032502014","DOIUrl":"https://doi.org/10.53106/160792642024032502014","url":null,"abstract":"\u0000 Deep convolutional neural networks have achieved good performance in the application of computer vision, but there are also problems, such as a large amount of computation, time consuming, and high memory demand. In this paper, a depthwise separable convolution filter pruning method based on PCA is proposed. First, this paper uses depthwise separable convolution to replace the conventional convolution in ResNet to reduce the number of parameters and the amount of computation in the network. The specific operation process is to first use depthwise convolution to separate the spatial dimension to increase the network width and expand the range of feature extraction, and then use pointwise convolution to reduce the computational complexity of conventional convolution operation. Second, PCA is used to distinguish stacked similar filters and perform dimensionality reduction, which not only alleviates the dimensional disaster, but also achieves compression of data and minimizes information loss. Experimental results show that this method can significantly improve the calculation speed and accuracy of the deep convolutional neural network model, and further compress the model size. On the clinical Color Perception Test Chart, this method reduced the amount of model parameters and MACs on ResNet by about 91% while maintaining the test accuracy at about 95%. With almost no loss of accuracy, this method greatly shortened the running time of the model.\u0000 \u0000","PeriodicalId":442331,"journal":{"name":"網際網路技術學刊","volume":"24 2","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140399773","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-01DOI: 10.53106/160792642024032502008
Kyusuk Han Kyusuk Han, Eiman Al Nuaimi Kyusuk Han, Shamma Al Blooshi Eiman Al Nuaimi, Rafail Psiakis Shamma Al Blooshi, Chan Yeob Yeun Rafail Psiakis
The drone swarm is a preferable way to deploy many drones for large-scale missions. Establishing secure communication among drones in drone swarms is essential as the fog drone controls all other edge drones in the swarm. Although many researchers proposed methods of authenticating drones, most of them are unsuitable for use in swarm environments as they require either the ground station during the authentication or expensive PKI-based crypto operations with limited flexibility and scalability. In this work, we propose an efficient and scalable authentication protocol for drone swarm environments, enabling mutual authentication between fog and edge drones without involving the ground station. Moreover, we show that the protocol enables the verification of the sender in group communication. Protocol evaluations show security requirements satisfaction while achieving 14 - 20 times less computation overhead as compared to PKI-based models.
{"title":"Scalable Authenticated Communication in Drone Swarm Environment","authors":"Kyusuk Han Kyusuk Han, Eiman Al Nuaimi Kyusuk Han, Shamma Al Blooshi Eiman Al Nuaimi, Rafail Psiakis Shamma Al Blooshi, Chan Yeob Yeun Rafail Psiakis","doi":"10.53106/160792642024032502008","DOIUrl":"https://doi.org/10.53106/160792642024032502008","url":null,"abstract":"\u0000 The drone swarm is a preferable way to deploy many drones for large-scale missions. Establishing secure communication among drones in drone swarms is essential as the fog drone controls all other edge drones in the swarm. Although many researchers proposed methods of authenticating drones, most of them are unsuitable for use in swarm environments as they require either the ground station during the authentication or expensive PKI-based crypto operations with limited flexibility and scalability. In this work, we propose an efficient and scalable authentication protocol for drone swarm environments, enabling mutual authentication between fog and edge drones without involving the ground station. Moreover, we show that the protocol enables the verification of the sender in group communication. Protocol evaluations show security requirements satisfaction while achieving 14 - 20 times less computation overhead as compared to PKI-based models.\u0000 \u0000","PeriodicalId":442331,"journal":{"name":"網際網路技術學刊","volume":"64 2","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140403358","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-01DOI: 10.53106/160792642024032502011
Yi-Jen Su Yi-Jen Su, Sheng-Yuan Yang Yi-Jen Su
The study focused on leveraging artificial intelligence (AI) for efficient energy conservation in scientific applications. The proposed cloud-based multi-agent system merges various intelligent technologies to swiftly gather high-quality cloud data for effective smart energy-saving. Incorporating case-based reasoning (CBR), big data analysis, and intelligent user interfaces as key functionalities, the system utilized Web services, ontology, open data, and data mining. It expanded on the practical advancements of the multi-agent Dr. What-Info system for information collection. A Web services platform seamlessly gathers cloud interactions among subagents processing energy-saving data. Rigorous performance and operational experiments were conducted to demonstrate the efficiency and effectiveness of the system interface, offering detailed insights into relevant R&D technologies and outcomes.
{"title":"A User-friendly Cloud-based Multi-agent Information System for Smart Energy-saving","authors":"Yi-Jen Su Yi-Jen Su, Sheng-Yuan Yang Yi-Jen Su","doi":"10.53106/160792642024032502011","DOIUrl":"https://doi.org/10.53106/160792642024032502011","url":null,"abstract":"\u0000 The study focused on leveraging artificial intelligence (AI) for efficient energy conservation in scientific applications. The proposed cloud-based multi-agent system merges various intelligent technologies to swiftly gather high-quality cloud data for effective smart energy-saving. Incorporating case-based reasoning (CBR), big data analysis, and intelligent user interfaces as key functionalities, the system utilized Web services, ontology, open data, and data mining. It expanded on the practical advancements of the multi-agent Dr. What-Info system for information collection. A Web services platform seamlessly gathers cloud interactions among subagents processing energy-saving data. Rigorous performance and operational experiments were conducted to demonstrate the efficiency and effectiveness of the system interface, offering detailed insights into relevant R&D technologies and outcomes.\u0000 \u0000","PeriodicalId":442331,"journal":{"name":"網際網路技術學刊","volume":"94 S2","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140405599","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-01DOI: 10.53106/160792642024032502005
Qi Chen Qi Chen, Xinyi Gao Qi Chen, Renjie Li Xinyi Gao, Yong Zhang Renjie Li
This paper proposes a Mini Net lightweight model that can be used for real-time detection. This model works together with Mini Lower and Mini Higher, which greatly improves the detection efficiency while ensuring the accuracy. The Mini module designs both the batch normalization layer and the excitation function at the front end of the module, which realizes efficient convolution, greatly reduces the amount of parameters and computation, and introduces the nonlinearity brought by more layers in the spatial dimension, which can improve the performance of the module extraction capacity. Based on the Mini convolution module, a multi-stage training strategy is proposed. The first stage makes the system fast and stable. In order to improve the overfitting phenomenon of the system, the second and third stages use finer features to improve the detection of small targets, thereby improving the Model training efficiency and detection accuracy.
本文提出了一种可用于实时检测的 Mini Net 轻量级模型。该模型与 Mini Lower 和 Mini Higher 配合使用，在保证准确性的同时大大提高了检测效率。Mini 模块在模块前端同时设计了批量归一化层和激励函数，实现了高效卷积，大大减少了参数量和计算量，并在空间维度上引入了多层带来的非线性，可以提高模块提取能力的表现。在迷你卷积模块的基础上，提出了多阶段训练策略。第一阶段使系统快速稳定。为了改善系统的过拟合现象，第二和第三阶段使用更精细的特征来提高对小目标的检测能力，从而提高模型训练效率和检测精度。
{"title":"Optimized Object Detection Based on The Improved Lightweight Model Mini Net","authors":"Qi Chen Qi Chen, Xinyi Gao Qi Chen, Renjie Li Xinyi Gao, Yong Zhang Renjie Li","doi":"10.53106/160792642024032502005","DOIUrl":"https://doi.org/10.53106/160792642024032502005","url":null,"abstract":"\u0000 This paper proposes a Mini Net lightweight model that can be used for real-time detection. This model works together with Mini Lower and Mini Higher, which greatly improves the detection efficiency while ensuring the accuracy. The Mini module designs both the batch normalization layer and the excitation function at the front end of the module, which realizes efficient convolution, greatly reduces the amount of parameters and computation, and introduces the nonlinearity brought by more layers in the spatial dimension, which can improve the performance of the module extraction capacity. Based on the Mini convolution module, a multi-stage training strategy is proposed. The first stage makes the system fast and stable. In order to improve the overfitting phenomenon of the system, the second and third stages use finer features to improve the detection of small targets, thereby improving the Model training efficiency and detection accuracy.\u0000 \u0000","PeriodicalId":442331,"journal":{"name":"網際網路技術學刊","volume":"21 84","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140406031","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-01DOI: 10.53106/160792642024032502003
Sunjun Lee Sunjun Lee, Yonggu Shin Sunjun Lee, Minseong Choi Yonggu Shin, Haehyun Cho Minseong Choi, Jeong Hyun Yi Haehyun Cho
A lot of the recently reported malware is equipped with the anti-analysis techniques (e.g., anti-emulation, anti-debugging, etc.) for preventing from being the analyzed, which can delay detection and make malware alive for a longer period. Therefore, it is of the great importance of developing automated approaches to defeat such anti-analysis techniques so that we can handle and effectively mitigate numerous malware. In this paper, by analyzing 1,535 malicious applications, we found that 18.31% of them equipped with anti-analysis techniques. Next, we propose a novel, dynamic analyzer, named DOOLDA, for automatically invalidating anti-analysis techniques through dynamic instrumentation. DOOLDA monitors executions of Android applications’ entire code layers (i.e., bytecode and native code). Based on monitoring results, DOOLDA finds the code related to anti-analysis techniques and invalidates the anti-analysis techniques by instrumenting it. To demonstrate the effectiveness of DOOLDA, we show that it can invalidate all known anti-analysis techniques. Also, we compare DOOLDA with other dynamic analyzers.
{"title":"Hybrid Dynamic Analysis for Android Malware Protected by Anti-Analysis Techniques with DOOLDA","authors":"Sunjun Lee Sunjun Lee, Yonggu Shin Sunjun Lee, Minseong Choi Yonggu Shin, Haehyun Cho Minseong Choi, Jeong Hyun Yi Haehyun Cho","doi":"10.53106/160792642024032502003","DOIUrl":"https://doi.org/10.53106/160792642024032502003","url":null,"abstract":"\u0000 A lot of the recently reported malware is equipped with the anti-analysis techniques (e.g., anti-emulation, anti-debugging, etc.) for preventing from being the analyzed, which can delay detection and make malware alive for a longer period. Therefore, it is of the great importance of developing automated approaches to defeat such anti-analysis techniques so that we can handle and effectively mitigate numerous malware. In this paper, by analyzing 1,535 malicious applications, we found that 18.31% of them equipped with anti-analysis techniques. Next, we propose a novel, dynamic analyzer, named DOOLDA, for automatically invalidating anti-analysis techniques through dynamic instrumentation. DOOLDA monitors executions of Android applications’ entire code layers (i.e., bytecode and native code). Based on monitoring results, DOOLDA finds the code related to anti-analysis techniques and invalidates the anti-analysis techniques by instrumenting it. To demonstrate the effectiveness of DOOLDA, we show that it can invalidate all known anti-analysis techniques. Also, we compare DOOLDA with other dynamic analyzers.\u0000 \u0000","PeriodicalId":442331,"journal":{"name":"網際網路技術學刊","volume":"52 3","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140400317","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：