检索结果-内蒙古大学图书馆

PriFace: a privacy-preserving face recognition framework under untrusted server

Journal of Ambient Intelligence and Humanized Computing 2023年第3期14卷 2967-2979页

作者： Zhao, Shan Zhang, Lefeng Xiong, Ping School of Information and Safety Engineering Zhongnan University of Economics and Law Wuhan China Center for Cyber Security and Privacy School of Computer Science University of Technology Sydney Sydney Australia

Alongside the rapid development of artificial intelligence and mobile Internet, face recognition has gradually become a mainstream intelligent terminal authentication technology that is widely applied in security, finance, and social networking, among other fields. However, face images usually contain sensitive personal information, which can have serious consequences for users if this information is leaked or misused. Most existing face privacy protection methods are based on data encryption techniques or differential privacy mechanisms, which make it difficult to strike a good balance between privacy and usability. In this paper, we propose PriFace, a new privacy-preserving method for face recognition based on eigenfaces and locality-sensitive hashing (LSH). PriFace aims to achieve accurate and efficent face recognition with privacy preserved. It applies eigenface technique to reduce the dimension of face images for improving retrieval performance. With the assumption that the server is untrustworthy, PriFace uses the locality-sensitive hashing technique to introduce randomness into the face information for privacy preservation so that the server, or other adversaries, can not reconstruct any user’s face image. Experiments on real-world datasets demonstrate that the proposed PriFace can effectively preserve the privacy of face images and further outperforms the traditional methods. © 2023, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.

关键词： Face recognition

来源：评论

学校读者我要写书评

暂无评论

Ventricular Vision: AI-Driven Heart Disease Detection with CNN-SVM-Based Classification

Ventricular Vision: AI-Driven Heart Disease Detection with C...

引用

2024 Asia Pacific Conference on Innovation in technology, APCIT 2024

作者： Suryavanshi, Ankita Mehta, Shiva Gupta, Richa Garg, Navin Jain, Vishal Chitkara University Institute of Engineering and Technology Chitkara University Punjab India Graphic Era Hill University Computer Science and Engineering Uttarakhand Dehradun India Graphic Era Deemed to Be University Computer Science and Engineering Uttarakhand Dehradun India Sharda University Computer Science and Engineering Sharda School of Engineering and Technology U.P. Greater Noida India

ISBN: (纸本)9798350361537

This study builds upon the CNN-SVM model as a sound diagnostic method for identifying five distinct heart diseases. We utilize the effectiveness of the deep learning procedures of various CNNs and the ability of the SVMs, which, in the end, might be customized and optimized to solve particular cardiac diseases. Two classifications examined are atrioventricular block, arrhythmia, bradycardia, tachycardia, second-degree heart block, and annexation. The appraisal of CNN-SVM model performance accomplished through various parameters was detailed. The precision of the correct classes arrayed from 94.89% to 97.00%, the recall ranged between 93.70% and 98.91%, and the F1-scores were also high, extending from 94.87% to 97.25%. A significant enhancement in the precision and accuracy of the model was registered, which demonstrated statistics close to 0.99 for Class 1, 0.98 for Classes 2 and 3, and 0.99 for Classes 4 and 5, highlighting its high performance. Various types of averages were used to evaluate the model's effectiveness. The superiority of the macro average in accuracy, recall, and F1-score was demonstrated. It is calculated for all classes equally. The results were 96.14%, 96.20%, and 96.16%. We applied the weighted average (taking into account class imbalance) and got 96.15% for the straight as well as for the recall and the F1-score. The micro average, which selects the best from among all classes with identical scores, is kept at 96.15% for all tips, showing a well-suited classifier. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Enhanced Play FAIR Cipher

Enhanced Play FAIR Cipher

引用

2023 International Conference on New Frontiers in Communication, Automation, Management and Security, ICCAMS 2023

作者： Syed Siraj Ahmed, N. Fathimath Nihala, T. Presidency University School of Computer Science and Engineering & Information Technology Karnataka Bangalore India

ISBN: (纸本)9798350317060

Security plays a crucial role in the field of communication between any two devices. Design of complex unpredictable keys are essential to make the communication more secure in various dimensions of data such as small, medium, and large. However, modern portable and IoT peripheral devices have limited computational power that hinders the process of key generation. Further, their communication system does not have strong protection. To overcome this, we propose a model to encrypt and decrypt the messages which requires less computational power. This can be used in modern portable and IoT peripheral devices of various data lengths. The proposed method also provides additional security through octal and hexadecimal play fair cipher. The proposed model is relatively harder to crack when compared to play fair cipher. The principle of the proposed design is similar to that of play fair cipher, however it uses a completely different cipher matrix of dimensions 8x8, 16x16 and a different replacement rule as well. This model supports the design of a multi-character encryption cipher. Here, a plaintext is taken as input and is transformed into corresponding cipher text. The proposed design is applied to various communication systems with additional security compared to play fair cipher between devices of portable and IoT peripheral devices of different data lengths. Since it is implemented with octal number, hexadecimal number and hence named enhanced play fair cipher. © 2023 IEEE.

关键词： Cryptography

来源：评论

学校读者我要写书评

暂无评论

Exploration of C++ Teaching Reform Method Oriented by Ability Output 17th

Exploration of C++ Teaching Reform Method Oriented by Abilit...

引用

17th International Conference on computer science and Education, ICCSE 2022

作者： Zheng, Yunping School of Computer Science and Engineering South China University of Technology Guangzhou510006 China

ISBN: (纸本)9789819924455

Aiming at the main problems existing in traditional C+ + curriculum teaching, relying on the background of the "new engineering" era, a "five in one" C++ teaching innovation mode is proposed. The "five in one" means: a national quality resource sharing course, a system combining competition and teaching, a C++ intelligent examination system, an ideological and political classroom, a laboratory combining competition and teaching. The "five-in-one" teaching model highlights the "student-centered" teaching concept. According to the teaching objectives of each class, it is cleverly integrated into the curriculum ideological and political content with "problem solving" and "live-code" as the main teaching line. Information-based teaching is used as a way to promote the integration of students’ theory and practice with ACM/ICPC, organically integrate process assessment and summative assessment, and implement diversified assessment forms. The results of more than ten years of practical teaching show that the "five-in-one" teaching innovation mode effectively cultivates students’ computational thinking, and significantly improves students’ programming ability and innovation ability. In the previous ACM/ICPC, our students have repeatedly achieved good results. At present, the "five-in-one" teaching innovation mode has been effectively promoted and applied in the three campuses of our university, which has a certain leading and exemplary role in cultivating the strong professional programming skills, computational thinking, the family and country feelings, the global vision and the all-round development of the top-notch innovative talents in the computer field. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Students

来源：评论

学校读者我要写书评

暂无评论

Sentiment Analysis on Multilingual Data: Hinglish

Sentiment Analysis on Multilingual Data: Hinglish

引用

International Conference on Data Analytics and Management, ICDAM 2023

作者： Kumar, Ashish Mangotra, Abhishek Ailawadi, Ayush Jain, Rachna Arora, Monika School of Computer Science Engineering and Technology Bennett University Uttar Pradesh India Department of Information Technology Bhagwan Parshuram Institute of Technology Delhi India

ISBN: (纸本)9789819965465

In this study, sentiment analysis was done on a collection of bilingual Hindi and English tweets. Before performing sentiment analysis, we first preprocessed the dataset to improve its quality by removing hashtags, URLs, and mentions. To classify the tweets as good, negative, or neutral, we used a range of machine learning models, such as logistic regression, multinomial Naive Bayes, support vector machines, random forests, and decision trees. We compared the results and utilized the metrics of accuracy and precision to assess the performance of the models. The investigation showed that the hybrid model, which included logistic regression with support vector machines, produced the highest accuracy and precision scores. Hyperparameter adjustment was used to improve the logistic regression model’s performance even more. The best model was then used to predict sentiments on a new dataset of unlabeled Hindi-English tweets. The empirical findings demonstrated that the hybrid model outperformed the other machine learning models utilized in this study, achieving an accuracy of 61.7% and a precision of 62.1%. When compared with all forecasts that were truly positive, the precision score showed the percentage of true positives, while the accuracy score showed the percentage of accurate predictions. The results showed that a highly accurate and exact solution for sentiment analysis on Hindi-English tweets is a hybrid model combining logistic regression and a support vector model with hyperparameter tweaking. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd 2024.

关键词： Sentiment analysis

来源：评论

学校读者我要写书评

暂无评论

SuperGlue-Based Deep Learning Method for Image Matching from Multiple Viewpoints 23

SuperGlue-Based Deep Learning Method for Image Matching from...

引用

8th International Conference on Mathematics and Artificial Intelligence, ICMAI 2023

作者： Li, Wotong School of Computer and Communication Engineering University of Science and Technology Beijing Beijing100083 China

ISBN: (纸本)9781450399982

With the ability to identify the same physical content from two or more images, image matching has always been a crucial and fundamental task in the fields of computer vision. Effective methods dealing with image matching are of great significance for 3D vision problems like Structure from Motion (SfM). In this work, a deep learning method is proposed based on SuperGlue to match images from multiple viewpoints with excellent performance. We utilize the SuperPoint network as the CNN backbone to extract keypoints and feature descriptors of an image pair. The descriptors are subsequently processed by SuperGlue using graph neural networks to identify then match the feature points. The matching results are finally optimized by the USAC_MAGSAC algorithm, which is a combination of MAGSAC++ implementation and degeneracy tests. The performance of our method is evaluated by the Kaggle competition Image Matching Challenge 2022, which is quantified by the mean Average Accuracy 0.84824/0.83940 on the public/private test set. This result places our team in the 15th position overall out of 643 teams and qualifies us for the silver medal in this competition. This work could have the potential to facilitate the development of generalized techniques of 3D vision tasks. © 2023 Copyright held by the owner/author(s)

关键词： Image matching

来源：评论

学校读者我要写书评

暂无评论

Poisoning attacks and countermeasures in intelligent networks:Status quo and prospects

引用

Digital Communications and Networks 2022年第2期8卷 225-234页

作者： Chen Wang Jian Chen Yang Yang Xiaoqiang Ma Jiangchuan Liu School of Computer Science and Information Engineering Hubei UniversityWuhanChina School of Electronic Information and Communications Huazhong University of Science and TechnologyWuhan430074China School of Computing Science at Simon Fraser University British ColumbiaCanada

Over the past years,the emergence of intelligent networks empowered by machine learning techniques has brought great facilitates to different aspects of human ***,using machine learning in intelligent networks also presents potential security and privacy threats.A common practice is the so-called poisoning attacks where malicious users inject fake training data with the aim of corrupting the learned *** this survey,we comprehensively review existing poisoning attacks as well as the countermeasures in intelligent networks for the first *** emphasize and compare the principles of the formal poisoning attacks employed in different categories of learning algorithms,and analyze the strengths and limitations of corresponding defense methods in a compact *** also highlight some remaining challenges and future directions in the attack-defense confrontation to promote further research in this emerging yet promising area.

关键词： Machine learning Poisoning attack Intelligent networks Security threat

来源：评论

学校读者我要写书评

暂无评论

Early and Automated Diagnosis of Dysgraphia Using Machine Learning Approach

引用

SN computer science 2023年第5期4卷 523页

作者： Agarwal, Basant Jain, Sonal Beladiya, Kishan Gupta, Yash Yadav, Avijit Singh Ahuja, Neelu Jyothi Department of Computer Science and Engineering Central University of Rajasthan Ajmer India PG Department of Computer Science and Technology Sardar Patel University Vallabh Vidhyanagar India Department of Computer Science and Engineering Indian Institute of Information Technology Kota Kota India Department of Computer Science School of Computer Science University of Petroleum and Energy Studies Dehradun 248007 India

Dysgraphia is a handwriting problem that impairs a person’s ability to write. Even the diagnosis of this condition is challenging, and there is currently no cure. Researchers from all over the world have studied this issue and offered several solutions. Motivation to work on this problem did arise after meeting with a few students struggling in achieving performance despite putting in sincere efforts. This paper also discusses the various forms of dysgraphia and its associated symptoms and proposes machine-learning models to detect dysgraphia. Unsupervised machine learning techniques are used to detect dysgraphia-related handwriting impairment. To accomplish the goal, a fresh handwriting dataset is created by conducting handwriting exercises and a wide variety of features are extracted to represent various handwriting characteristics. Results indicate that Random forest returns the best accuracy but scores less while detecting dysgraphic samples correctly. One class SVM has been tried to deal with the issue of the availability of dysgraphic samples required to train machines. Results indicate good hope in identification with a scope of improvement with increase in sample size for machine training. This paper also seeks to raise awareness of the dysgraphia issue and its effects on society. © 2023, The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd.

关键词： Dysgraphia Learning difficulties Machine Motor ability OCC One class SVM Random forest

来源：评论

学校读者我要写书评

暂无评论

Image based and Point Cloud based Methods for 3D View Reconstruction in Real-time Environment 14

Image based and Point Cloud based Methods for 3D View Recons...

引用

2024 IEEE 14th Annual Computing and Communication Workshop and Conference, CCWC 2024

作者： Agrawal, Arya Sharma, Teena Verma, Nishchal K. Vasikarla, Shantaram Vellore Institute of Technology School of Computer Science and Engineering Vellore India Indian Institute of Technology Mehta Family School of Data Science and Artificial Intelligence Guwahati India Indian Institute of Technology Dept. of Electrical Engineering Kanpur India California State University Dept. of Computer Science NorthridgeCA United States

ISBN: (纸本)9798350360134

3D view reconstruction in a real-time unknown environment poses several challenges to vision-based applications. Herein, this paper outlines real-time experimental results and analysis of 3D view reconstruction methods. Since real-time experimentation requires inputs from different sensors, this paper outlines the results of methods from the following categories: image based methods and point cloud based methods. This paper explains the detailed steps involved in 3D view reconstruction followed by an explanation of the methods. It provides detailed discussions, advantages, limitations, and empirical results, along with presenting differences between algorithms. For analysis of experimental results, two benchmarked datasets i.e., KITTI 3D benchmark suite and ShapeNet are considered. Furthermore, the paper explores potential future scopes, including the utilization of generative models, unsupervised learning, advanced sensors, and many more. © 2024 IEEE.

关键词： Network coding

来源：评论

学校读者我要写书评

暂无评论

Enhanced Hybrid Intrusion Detection System with Attention Mechanism using Deep Learning

引用

SN computer science 2024年第5期5卷 534页

作者： Chavan, Pundalik Hanumanthappa, H. Satish, E.G. Manoli, Sunil Supreeth, S. Rohith, S. Ramaprasad, H.C. School of Computer Science and Engineering REVA University Bengaluru 560064 India Nitte Meenakshi Institute of Technology Bengaluru 560064 India Department of ECE Nagarjuna College of Engineering and Technology Bengaluru India

The introduction of the Attention mechanism by the Internet of Things—or WSN-IoT—in the sector has greatly enhanced the intrusion detection mechanism capabilities, whereas the deep learning techniques, together with the attention mechanism, enhance the efficacy and efficiency of IDS within WSNs. The proposed “Enhanced Hybrid Intrusion Detection System with Attention Mechanism” (EHID-SCA) underlying insight of this is that the characteristics of WSN data validate this including Convolutional Neural Networks (CNNs) and other deep learning models into an effective and coherent architecture. The design of the deep hybrid network with attention incorporates Channel Attention and Spatial Attention as some of the main components. The model biased capacity is enlarged to choose and stress important spatial and channel information from the sensor input. The method is based on the consideration of these things and it cuts noise. Therefore, the proposed technique can be made to pave the way for enabling the Internet of Things intrusion detection systems to do the automatic extraction of useful information from sensor data through the utilization of deep learning along with the Attention mechanism. The attack might therefore be better situated for mitigation in the network design that allows analysis of the geographical and temporal context of events, which are then more properly termed as intrusion events. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2024.

关键词： Attention mechanism Deep learning Enhanced Hybrid Intrusion Detection Wireless sensor networks

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：