检索结果-内蒙古大学图书馆

LaserKey: Eavesdropping Keyboard Typing Leveraging Vibrational Emanations via Laser Sensing

IEEE Transactions on Mobile computing 2025年第6期24卷 4829-4844页

作者： Luo, Chengwen Xie, Zhuoqing Huang, Yuhan Chen, Gecheng Yao, Haiyi Zhang, Jin Cheng, Long Xu, Weitao Li, Jianqiang National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen518060 China School of Computing Clemson University United States City University of Hong Kong Hong Kong

Reconstructing keyboard input through side-channel attacks has posed significant threats to user security. While conventional keystroke eavesdropping attacks have demonstrated effectiveness using side channels such as acoustic signals, they are usually shorter in range and can be significantly affected by environmental noises. In this paper, we propose LaserKey, a novel keystroke eavesdropping technique that leverages the long-range and noise-resistant nature of lasers to achieve a more stealthy side-channel attack. We utilize laser sensors to accurately capture the subtle vibrations induced on laptop screens by keystrokes, and innovatively design a laser-driven deep learning-based keystroke recognition model with the inputs being the Mel-frequency Cepstral Coefficien (MFCC), Time Difference of Arrival (TDoA), and amplitude features extracted from such vibration signals. Through systematic experiments, we demonstrate that LaserKey achieves a 92.2% single-key recognition accuracy. By combining multiple single-key recognition capabilities based on this, we then realize the end-to-end word-level recognition. Moreover, to mitigate the recognition errors caused by the changes in keystroke positions, we introduce a meta-learning based domain generalization approach for achieving robust laser position calibration. Results show that LaserKey achieves as low as 3% character error rate (CER) for word-level recognition, proving its effectiveness for long-range and high-accuracy keystroke eavesdropping, and highlighting the necessity for countermeasures in the future. © 2002-2012 IEEE.

关键词： Systematic errors

来源：评论

学校读者我要写书评

暂无评论

Enhancing Connectivity of AODV in Urban Scenarios for VANETs: A QoS-Aware Approach 4

Enhancing Connectivity of AODV in Urban Scenarios for VANETs...

引用

4th International Conference on Sentiment Analysis and Deep Learning, ICSADL 2025

作者： Janet Kumari, J Jesy Thangam, S. Ghosh, Archishman Kolhe, Dhruv Vijay Ghosh, Soumish Amrita School of Computing Department of Computer Science and Engineering Bengaluru India

ISBN: (纸本)9798331523923

As cities expand, vehicles and congestion become more complex. Efficient vehicle-to-vehicle contact networks are needed for road safety and efficient traffic flow. Thus, Vehicular Ad Hoc Networks are needed to overcome all the issues. The Adhoc On-Demand Distance Vector (AODV) routing protocol is used in VANETs. However, this protocol has many challenges in urban settings due to high vehicle density, rapid topology changes, and non-line-of-sight conditions. This work improves AODV by integrating Quality of Service (QoS) metrics. The prioritizing factors such as overhead and end-to-end delay are used to create a more efficient protocol for urban traffic. The SUMO simulator is used for realistic traffic patterns and NS-2 for network analysis. An evaluation of the protocol against the classical AODV is done and important metrics for real-time communication are focused. The enhanced AODV showed marked improvements. It achieved lower overhead in route finding, an increase in the number of packets sent, and lower end-to-end delay. © 2025 IEEE.

关键词： Vehicular ad hoc networks

来源：评论

学校读者我要写书评

暂无评论

Towards an ontology on project portfolio management 16

Towards an ontology on project portfolio management

引用

16th International Conference on ENTERprise Information Systems, CENTERIS 2024 - 12th International Conference on Project MANagement, ProjMAN 2024 - 14th International Conference on Health and Social Care Information Systems and Technologies, HCist 2024

作者： Melo, Héctor Avila, Oscar Villamil, María Del Pilar Department of Systems and Computing Engineering School of Engineering Universidad de Los Andes Bogotá Colombia

Project Portfolio Management (PPM) is essential for organizations aiming to align projects with strategic goals. Different organizations adopt diverse PPM frameworks and standards to manage project portfolios each employing its own terminology. This semantic heterogeneity leads to communication barriers, knowledge silos, and difficulty in integrating information across various platforms in an interorganizational context. This article proposes a PPM ontology to establish a common language encapsulating key concept, addressing this challenge. The methodology involves systematic revision of three prominent PPM standards-ISO 21504, PMI Standard for Portfolio Management, and AXELOS Management of Portfolios, and employs a novel algorithm to identify equivalences and containment relationships between terms across all three standards, resolving semantic ambiguities and enriching the ontology's expressiveness. This contribution benefits researchers, academics, portfolio managers, project managers, and PPM practitioners by providing a common vocabulary and framework for understanding and improving PPM practices, facilitates knowledge representation, improves communication and collaboration among stakeholders, and lays the groundwork for developing intelligent PPM systems capable of leveraging shared semantic understanding. © 2025 The Authors.

关键词： Project management

来源：评论

学校读者我要写书评

暂无评论

Analysis of the Whiplash gradient descent dynamics

Analysis of the Whiplash gradient descent dynamics

引用

作者： Bhattacharjee, Subhransu S. Petersen, Ian R. School of Engineering College of Engineering and Computing and Cybernetics The Australian National University CanberraACT Australia

In this paper, we propose the Whiplash inertial gradient dynamics, a closed-loop optimization method that utilizes gradient information. We introduce the symplectic asymptotic convergence analysis for the Whiplash system for convex functions. We also introduce relaxation sequences to explain the nonclassical nature of the algorithm and an exploring heuristic variant of the Whiplash algorithm to escape saddle points, deterministically. We provide a practical methodology for analyzing convergence rates using integral constraint bounds and a novel Lyapunov rate method. Our results demonstrate polynomial and exponential rates of convergence for quadratic costs. © 2023 The Authors. Asian Journal of Control published by John Wiley & Sons Australia, Ltd on behalf of Chinese Automatic Control Society.

关键词： Lyapunov methods

来源：评论

学校读者我要写书评

暂无评论

DeepLeuk: a convolutional neural network pre-trained model for microscopic cell images-Based leukemia Cancer analysis

引用

Multimedia Tools and Applications 2025年第14期84卷 13809-13842页

作者： Yenurkar, Ganesh Keshaorao Mal, Sandip Thakur, Nileshsingh Dhomne, Shrawani Dhurve, Merula Patel, Mayank Kulmeti, Karan Dhurve, Harsh Yeshwantrao Chavan College of Engineering Wanadongri Nagpur India School of Computing Science & Engineering VIT Bhopal University Bhopal India VIT Bhopal University Bhopal India

The blood and bone marrow are affected by leukemia, a deadly kind of cancer, that significantly impacts the quality of life of those diagnosed. Early identification and precise diagnosis are crucial for improving survival rates. Fortunately, recent advancements in medical image analysis, particularly deep learning-based techniques, have greatly improved the ability to distinguish leukemia cells from healthy ones through microscopic cell images. This research introduces a deep learning-based leukemia cancer classifier, specifically a CNN pre-trained model, utilizing microscopic cell images to detect malignant cells. Using pre-processing techniques such as picture scaling, Region of Interest (RoI) extraction, and Improved Anisotropic Filtering (IAF) and feature extraction, the blood cell image dataset is first cleaned. After that leukemia-affected and healthy cells are evaluated using various classification algorithms and neural networks, with optimal features identified to improve classifier performance. The results suggest that neural networks function well as a classifier algorithm to detect whether the person is cancerous or non-cancerous, with the proposed CNN pre-trained model providing precision of 98.9%, which is higher than any other method mentioned. The proposed model prioritizes recall, a key performance metric, to reduce the number of false negatives. Accurate diagnosis and treatment are critical, as misdiagnosing a patient with cancer as not having cancer can lead to severe consequences. With the main objective of minimizing inadvertent mistakes made by physicians, the proposed model performs better than kNN, Decision Trees, Random Forest, SVM, and Logistic Regression models. Using deep learning-based techniques to improve cancer diagnosis and treatment is essential. Improving survival rates and the quality of life for individuals with leukemia requires early identification and accurate diagnosis. This research can help doctors make more accurate diagnos

关键词： Image classification

来源：评论

学校读者我要写书评

暂无评论

Discovering Social Media Influencers Using a Deep Regression Analysis

引用

SN Computer Science 2025年第4期6卷 1-18页

作者： Ramya, G.R. Bagavathi Sivakumar, P. Department of Computer Science and Engineering Amrita School of Computing Amrita Vishwa Vidyapeetham Coimbatore India

Highly influential users (IUs) play a vital role in disseminating information on online social networks (OSNs). Recognizing IUs is crucial for brand awareness, strategic marketing and consumer engagement. Researchers have developed the temporal influential model (TIM) and incremental logistic regression (ILLR) to find out the probability of a user being influential on Twitter. However, accurately identifying IUs requires considering multiple user and tweet attributes. This study proposes a model for predicting IUs by analyzing user and trend-based attributes from tweets, calculating a weighted associated influence (WAI) score for each user and enhancing IU prediction through user similarity, influence degree, and trustworthiness metrics. The goal is to combine the TIM with deep regression (TIM-DeepReg) to improve accuracy in predicting high IUs on OSNs while handling large amounts of data and attributes. The TIM-DeepReg model combines convolutional neural network (CNN) with regression to automatically learn user and tweet attributes and enhance IU prediction efficiency. Lastly, the experimental findings show that the TIM-DeepReg model surpasses other models in IU prediction for the Tariff, Service, and Punctuality topics of airline services on Twitter and Facebook, with 87.4%, 88.9%, and 88.8% of the most influential individuals predicted. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2025.

关键词： Deep regression Influencer on social networks Influencer prediction from Facebook and Twitter Temporal influential model Weighted associated influence score

来源：评论

学校读者我要写书评

暂无评论

A Transformer-Based Approach for Fake News and Spam Detection in Social Media Using RoBERTa

A Transformer-Based Approach for Fake News and Spam Detectio...

引用

2025 International Conference on Multi-Agent Systems for Collaborative Intelligence, ICMSCI 2025

作者： Pal, Ankit Abhijit Mondal, Sudin Kumar, C. Ashok Kumar, C. Jothi School of Computing College of Engineering and Technology Srm Institute of Science and Technology Department of Computing Technologies Kattankulathur Chennai India

ISBN: (纸本)9798331509828

The spread of misinformation and spam on social media has become a critical challenge, undermining information integrity and online security. Addressing this pressing issue, this study introduces an advanced solution utilizing RoBERTa, a transformer-based deep learning model, for the effective detection of fake news and spam. By employing sophisticated Natural Language Processing (NLP) techniques, the proposed method demonstrates exceptional accuracy in classifying unreliable content, making it a valuable tool in today's digital communication landscape. The approach is designed to provide real-time detection capabilities, ensuring swift identification and mitigation of misinformation and spam. Its adaptability allows seamless application across different domains, enhancing its versatility. Furthermore, the solution emphasizes sustainability by incorporating continuous monitoring and retraining mechanisms, which enable the model to maintain its performance even as the nature of misinformation evolves. This work underscores the transformative potential of deep learning models like RoBERTa in combating the widespread dissemination of online threats. By ensuring precise, scalable, and timely detection, the study contributes to creating a safer, more trustworthy digital ecosystem. In doing so, it addresses a critical need in the current era, where the reliability of information is paramount for societal and individual well-being. © 2025 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

From Buzz to Blockbuster: Predicting Movie Revenue Using a Hybrid Approach Combining Machine Learning and Sentiment Analysis

From Buzz to Blockbuster: Predicting Movie Revenue Using a H...

引用

2025 International Conference on Multi-Agent Systems for Collaborative Intelligence, ICMSCI 2025

作者： Vardhan, Senapathi Vishnu Balaji, Killemsetty Viswa Sai Kumar, C. Ashok Kumar, C. Jothi School of Computing College of Engineering and Technology Srm Institute of Science and Technology Department of Computing Technologies Kattankulathur Chennai India

ISBN: (纸本)9798331509828

This research investigates a hybrid approach for predicting movie revenue by integrating machine learning models with sentiment analysis. The growing influence of social media and online discussions offers a valuable opportunity to understand audience engagement and sentiment, which are critical in forecasting box office performance. By analyzing data from various social media platforms alongside movie metadata from sources like IMDb and TMDb, this study seeks to create a comprehensive model that can predict a movie's opening weekend revenue and classify it as a 'Hit,' 'Average,' or 'Flop.' The work highlights the value of audience sentiment, pre-release engagement, and historical box office trends in predicting financial outcomes. By using a combination of quantitative metrics (likes, shares, views) and qualitative data (sentiment scores, discussion themes), the study aims to refine revenue predictions and improve classification accuracy. The hybrid model's ability to incorporate both traditional and real-time data offers several advantages, including better responsiveness to audience perceptions, the ability to identify emerging trends, and more accurate revenue forecasts. This research provides actionable insights that can aid movie studios, marketers, and investors in decision-making processes, ultimately reducing financial risks associated with movie releases. The findings contribute to the growing field of predictive analytics in the entertainment industry, offering a novel and data-driven approach to understanding and forecasting movie success. © 2025 IEEE.

关键词： Predictive analytics

来源：评论

学校读者我要写书评

暂无评论

HybridViT: An Approach for Alzheimer’s Disease Classification with ADNI Neuroimaging Data

引用

SN Computer Science 2025年第4期6卷 1-18页

作者： Tekumudi, Vivek Sai Surya Chaitanya Ramanathan, Annamalai Department of Computer Science and Engineering Amrita School of Computing Amrita Vishwa Vidyapeetham Chennai India

Alzheimer’s disease (AD) is a prevalent neurological disorder characterized by progressive brain cell degeneration and atrophy, leading to a gradual decline in cognitive and functional abilities. Timely diagnosis is paramount in potentially delaying or preventing the course of this debilitating condition. Magnetic Resonance Imaging presents a non-invasive means for longitudinal monitoring and serves as a crucial biomarker for tracking disease progression. In particular, Structural Magnetic Resonance Imaging (sMRI) enables the quantification of atrophy, which is a reliable indicator for assessing the precise stage and severity of AD-related neurodegeneration. In this research, we employ advanced machine learning techniques to address the challenge of accurate and early diagnosis of AD. We utilize a comprehensive dataset comprising five stages of 2D sMRI Image data, encompassing AD, Cognitively Normal, Mild Cognitive Impairment, Early Mild Cognitive Impairment, and Late Mild Cognitive Impairment classes. To optimize the classification process, we explore a novel approach that combines a Vision Transformer with pre-trained convolutional neural networks. Our study includes binary and multi-class classification tasks, and the performance of 26 pre-trained Keras Deep Learning models is assessed. Notably, The DenseNet121+ViT model achieves 91.26% accuracy in multi-class classification. For binary classification, MobileNetV2+ViT, EfficientNetB4+ViT, and MobileNet+ViT achieve accuracies of 92.33%, 97.82%, and 94.81%, respectively. These results highlight our approach’s potential to improve AD diagnosis accuracy and underscore the importance of deep learning in early detection of neurological disorders. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2025.

关键词： Alzheimer’s disease (AD) Alzheimer’s disease neuroimaging initiative (ADNI) HybridViT Pre-trained models Vision transformer

来源：评论

学校读者我要写书评

暂无评论

An Innovative Deep Learning-Based Predictive Maintenance Model for Smart Automotive System

An Innovative Deep Learning-Based Predictive Maintenance Mod...

引用

2025 International Conference on Multi-Agent Systems for Collaborative Intelligence, ICMSCI 2025

作者： Indumathi, V. Ashokkumar, C. School of Computing College of Engineering and Technology Srm Institute of Science and Technology Department of Computing Technologies Kattankulathur Chennai India

ISBN: (纸本)9798331509828

This research presents an innovative deep learning-based predictive maintenance model designed for smart automotive systems, utilizing the EnsembleAE-Boost (EAE-Boost) algorithm. The primary objective of the proposed model is to enhance the prediction accuracy and fault detection capabilities of smart vehicles through advanced machine learning techniques. The EAE-Boost algorithm, an ensemble model integrating auto encoders and boosting, is compared with existing algorithms, namely K-Nearest Neighbors (KNN), Deep Neural Networks (DNN), and Extreme Gradient Boosting (XGBoost), based on key simulation metrics, including accuracy, precision, prediction time, and fault detection rate. Simulation results demonstrate that the EAE-Boost algorithm outperforms the existing algorithms, achieving superior performance in all metrics. Notably, the proposed method showed significant improvements in accuracy and fault detection rate, with a reduction in prediction time, highlighting its potential for real-time maintenance applications in smart automotive systems. The findings underscore the effectiveness of the proposed deep learning model in enhancing vehicle reliability, safety, and operational efficiency. © 2025 IEEE.

关键词： Predictive maintenance

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：