检索结果-内蒙古大学图书馆

Runtime Backdoor Detection for Federated Learning via Representational Dissimilarity Analysis

IEEE Transactions on Dependable and Secure Computing 2025年

作者： Zhang, Xiyue Xue, Xiaoyong Du, Xiaoning Xie, Xiaofei Liu, Yang Sun, Meng University of Oxford Department of Computer Science United Kingdom Peking University Department of Information and Computing Science School of Mathematical Science China Monash University Department of Software Systems and Cybersecurity Australia Singapore Management University School of Computing and Information Systems Singapore Nanyang Technological University School of Computer Science and Engineering Singapore

Federated learning (FL), as a powerful learning paradigm, trains a shared model by aggregating model updates from distributed clients. However, the decoupling of model learning from local data makes FL highly vulnerable to backdoor attacks, where a single compromised client can poison the shared model. While recent progress has been made in backdoor detection, existing methods face challenges with detection accuracy and runtime effectiveness, particularly when dealing with complex model architectures. In this work, we propose a novel approach to detecting malicious clients in an accurate, stable, and efficient manner. Our method utilizes a sampling-based network representation method to quantify dissimilarities between clients, identifying model deviations caused by backdoor injections. We also propose an iterative algorithm to progressively detect and exclude malicious clients as outliers based on these dissimilarity measurements. Evaluations across a range of benchmark tasks demonstrate that our approach outperforms state-of-the-art methods in detection accuracy and defense effectiveness. When deployed for runtime protection, our approach effectively eliminates backdoor injections with marginal overheads. © 2025 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Enhancing Intrusion Detection systems Using RNN, LSTM, and Hybrid RNN-LSTM Models

Enhancing Intrusion Detection Systems Using RNN, LSTM, and H...

引用

Machine Learning and Autonomous systems (ICMLAS), International Conference on

作者： Sahar Ebadinezhad Nooshin Nooraei Nia Nasratullah Shirzad Nwabueze Kenneth Osemeha Department of Computer Information Systems Computer Information Systems Research and Technology Center (CISRTC) Near East University Nicosia Turkey Artificial Intelligence Near East University Nicosia North Cyprus Software Engineering Department Near East University Nicosia North Cyprus

ISBN: (数字)9798331505745

ISBN: (纸本)9798331505752

Due to The lack of comparison studies and practical applications of RNN, LSTM, and hybrid RNN-LSTM models for intrusion detection systems, especially when managing class imbalances in complex network datasets, represents a gap in the literature. This study investigates the application of deep learning techniques to improve the detection capabilities of IDS. For this study, we applied and assessed Recurrent Neural Network, Long Short-Term Memory, and Hybrid RNN-LSTM models on a UNSW-NB15 dataset, and to redress class imbalance we trained and validated our models using synthetic minority over-sampling. The RNN-LSTM model was shown to perform the best with a 9 4. 0 0 % accuracy in comparison to other models. These experimental results signify the possibility of enhancing IDS performance and a dependable method of detecting various network intrusions with the use of hybrid models comprised of RNN & LSTM.

关键词： Deep learning Recurrent neural networks Accuracy Computational modeling Intrusion detection computer architecture Network security Hyperparameter optimization Network intrusion Long short term memory

来源：评论

学校读者我要写书评

暂无评论

A novel facial expression recognition framework using deep learning based dynamic cross-domain dual attention network

引用

PeerJ computer Science 2025年 11卷 e2866-e2866页

作者： Alzahrani, Ahmed Omar Alghamdi, Ahmed Mohammed Ashraf, M. Usman Ilyas, Iqra Sarwar, Nadeem Alzahrani, Abdulrahman Alarood, Alaa Abdul Salam Department of Information Systems and Technology College of Computer Science and Engineering University of Jeddah Makkah Jeddah Saudi Arabia Department of Software Engineering College of Computer Science and Engineering University of Jeddah Makkah Jeddah Saudi Arabia Department of Computer Science Government College Women University Sialkot Punjab Sialkot Pakistan Department of Computer Science Bahria University Punjab Lahore Pakistan

Variations in domain targets have recently posed significant challenges for facial expression recognition tasks, primarily due to domain shifts. Current methods focus largely on global feature adoption to achieve domain-invariant learning;however, transferring local features across diverse domains remains an ongoing challenge. Additionally, during training on target datasets, these methods often suffer from reduced feature representation in the target domain due to insufficient discriminative supervision. To tackle these challenges, we propose a dynamic cross-domain dual attention network for facial expression recognition. Our model is specifically designed to learn domain-invariant features through separate modules for global and local adversarial learning. We also introduce a semantic-aware module to generate pseudo-labels, which computes semantic labels from both global and local features. We assess our model’s effectiveness through extensive experiments on the Real-world Affective Faces Database (RAF-DB), FER-PLUS, AffectNet, Expression in the Wild (ExpW), SFEW 2.0, and Japanese Female Facial Expression (JAFFE) datasets. The results demonstrate that our scheme outperforms the existing state-of-the-art methods by attaining recognition accuracies 93.18, 92.35, 82.13, 78.37, 72.47, 70.68 respectively. Copyright 2025 Alzahrani et al. Distributed under Creative Commons CC-BY 4.0

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Leveraging Predictive Maintenance for Photovoltaic systems 9th

Leveraging Predictive Maintenance for Photovoltaic Systems

引用

9th International Conference on Smart City Applications, SCA 2024

作者： Ledmaoui, Younes Raif, Mourad El Aroussi, Mohamed Saadane, Rachid Jakimi, Abdeslam Chehri, Abdellah Laboratory Engineering System Hassania School of Public Works Casablanca Morocco Software Engineering and Information Systems Engineering Team FST of Errachidia Moulay Ismail University Meknes Morocco Department of Mathematics and Computer Science Royal Military College of Canada Kingston K7K 7B4 ON Canada

ISBN: (纸本)9783031886522

This paper introduces a novel edge device architecture designed to optimize solar energy management systems. It integrates cutting-edge functionalities such as generation prediction, maintenance alerts, and anomaly detection into a unified framework. Through the utilization of edge computing, the system enables real-time analysis and decision-making at the network’s edge. Leveraging machine learning algorithms and predictive models, these edge devices provide highly accurate energy generation forecasts, facilitating efficient energy utilization and strategic planning. Additionally, the architecture incorporates anomaly detection techniques to proactively identify deviations from normal operation, minimizing downtime and enabling timely maintenance interventions. The seamless integration of these features within the edge devices significantly enhances the overall efficiency, reliability, and performance of energy monitoring systems. By adopting this advanced edge device architecture, energy stakeholders can reap benefits such as improved energy management, reduced operational costs, and enhanced system reliability. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Edge device Energy management system Energy monitoring Predictive maintenance Renewable energy Solar energy

来源：评论

学校读者我要写书评

暂无评论

Hardware Design and Security Needs Attention: From Survey to Path Forward

arXiv

引用

arXiv 2025年

作者： Ghimire, Sujan Chowdhury, Muhtasim Alam Latibari, Banafsheh Saber Mamun, Muntasir Carpenter, Jaeden Wolf Tan, Benjamin Pearce, Hammond Satam, Pratik Salehi, Soheil Systems and Industrial Engineering Department University of Arizona TucsonAZ United States Electrical and Computer Engineering Department University of Arizona TucsonAZ United States Department of Electrical and Software Engineering University of Calgary Alberta CalgaryAB Canada School of Computer Science and Engineering University of New South Wales Sydney Australia

Recent advances in attention-based artificial intelligence (AI) models have unlocked vast potential to automate digital hardware design while enhancing and strengthening security measures against various threats. This rapidly emerging field leverages Large Language Models (LLMs) to generate HDL code, identify vulnerabilities, and sometimes mitigate them. The state of the art in this design automation space utilizes optimized LLMs with HDL datasets, creating automated systems for register transfer level(RTL) generation, verification, and debugging and establishing LLM-driven design environments for streamlined logic designs. Additionally, attention-based models like graph attention have shown promise in chip design applications, including floorplanning. This survey investigates the integration of these models into hardware-related domains, emphasizing logic design and hardware security, with or without the use of IP libraries. This study explores the commercial and academic landscape, highlighting technical hurdles and future prospects for automating hardware design and security. Moreover, it provides new insights into the study of LLM-driven design systems, advances in hardware security mechanisms, and the impact of influential works on industry practices. Through the examination of 30 representative approaches and illustrative case studies, this paper underscores the transformative potential of attention-based models in revolutionizing hardware design while addressing the challenges that lie ahead in this interdisciplinary domain. © 2025, CC BY.

关键词： Hardware-software codesign

来源：评论

学校读者我要写书评

暂无评论

Highly Accurate Crow Search Optimized CNN Classifier for Cyclone Prediction

Highly Accurate Crow Search Optimized CNN Classifier for Cyc...

引用

Data Science, Agents & Artificial Intelligence (ICDSAAI), International Conference on

作者： R. Ramya Kavin K. S R. Anuja A. T. R. Krishna Priya D. Karthikeyan Kiruba Thangam Raja Department of Electronics and Communication Engineering Saveetha School of Engineering Saveetha Institute of Medical and Technical Sciences (SIMATS) Tamilnadu India Department of Electrical and Electronics Engineering Government College of Engineering Tirunelveli Tamilnadu India Department of Computer Science and Engineering Rohini College of Engineering and Technology Kanyakumari Tamilnadu India Department of Electrical and Electronics Engineering SRM Institute of Science and Technology Chennai India Department of Software and Systems Engineering School of Computer Science Engineering and Information Systems (SCORE) VIT University Vellore

ISBN: (数字)9798331537555

ISBN: (纸本)9798331537562

Accurate and timely prediction of cyclone events is essential for effective disaster management and risk mitigation. This study introduces a highly precise cyclone prediction model that combines Convolutional Neural Networks (CNNs) with the optimization capabilities of the Crow Search Algorithm (CSA). The resulting model, called Crow Search Optimized CNN (CSO-CNN), harnesses the feature extraction and classification strengths of CNNs, while the CSA is used to optimize the hyperparameters of the CNN architecture for optimal performance. The model is trained and validated on a comprehensive dataset that includes satellite imagery and meteorological data from a wide range of cyclone events. The CSO-CNN achieves remarkable classification accuracy, surpassing traditional CNN models and other leading techniques. Its ability to quickly and accurately assess the likelihood of cyclone formation provides crucial decision support for disaster management authorities, enabling timely and effective response strategies. By integrating the CSA optimization technique with the CNN framework, this study presents a novel approach to cyclone prediction. It highlights the potential of combining advanced Machine Learning (ML) algorithms to tackle complex environmental challenges and deepen our understanding of intricate atmospheric phenomena.

关键词： Accuracy Atmospheric modeling Disaster management Predictive models Prediction algorithms Feature extraction Cyclones Satellite images Convolutional neural networks Optimization

来源：评论

学校读者我要写书评

暂无评论

EcoWeedNet: A Lightweight and Automated Weed Detection Method for Sustainable Next-Generation Agricultural Consumer Electronics

arXiv

引用

arXiv 2025年

作者： Khater, Omar H. Siddiqui, Abdul Jabbar Hossain, M. Shamim Dhahran31261 Saudi Arabia SDAIAKFUPM Joint Research Center on Artificial Intelligence IRC for Intelligent Secure Systems Department of Computer Engineering KFUPM Saudi Arabia Department of Software Engineering College of Computer and Information Sciences King Saud University Riyadh12372 Saudi Arabia

Sustainable agriculture plays a crucial role in ensuring world food security for consumers. A critical challenge faced by sustainable precision agriculture is weed growth, as weeds share essential resources with the crops, such as water, soil nutrients, and sunlight, which notably affect crop yields. The traditional methods employed to combat weeds include the usage of chemical herbicides and manual weed removal methods However, these could damage the environment and pose health hazards. The adoption of automated computer vision technologies and ground agricultural consumer electronic vehicles in precision agriculture offers sustainable, low-carbon solutions. However, prior works suffer from issues such as low accuracy and precision and high computational expense. This work proposes EcoWeedNet, a novel model with enhanced weed detection performance without adding significant computational complexity, aligning with the goals of low-carbon agricultural practices. Additionally, our model is lightweight and optimal for deployment on ground-based consumer electronic agricultural vehicles and robots. The effectiveness of the proposed model is demonstrated through comprehensive experiments on the CottonWeedDet12 benchmark dataset reflecting real-world scenarios. EcoWeedNet achieves performance close to that of large models yet with much fewer parameters. (approximately 4.21% of the parameters and 6.59% of the GFLOPs of YOLOv4). This work contributes effectively to the development of automated weed detection methods for next-generation agricultural consumer electronics featuring lower energy consumption and lower carbon footprint. This work paves the way forward for sustainable agricultural consumer technologies. Copyright © 2025, The Authors. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Towards Fully-Automated Materials Discovery via Large-Scale Synthesis Dataset and Expert-Level LLM-as-a-Judge

arXiv

引用

arXiv 2025年

作者： Kim, Heegyu Jeon, Taeyang Choi, Seungtaek Hong, Jihoon Jeon, Dongwon Baek, Ga-Yeon Kwak, Kyung-Won Lee, Dong-Hee Bae, Jisu Lee, Chihoon Kim, Yunseo Choi, Sun-Jin Park, Jinsung Cho, Sungbum Cho, Hyunsouk Department of Artificial Intelligence Korea Republic of Department of Software and Computer Engineering Korea Republic of Department of Materials Science and Engineering Department of Energy Systems Research Ajou University Suwon16499 Korea Republic of Division of Materials Science and Engineering Hanyang University Seoul04763 Korea Republic of

Materials synthesis is vital for innovations such as energy storage, catalysis, electronics, and biomedical devices. Yet, the process relies heavily on empirical, trial-and-error methods guided by expert intuition. Our work aims to support the materials science community by providing a practical, data-driven resource. We have curated a comprehensive dataset of 17K expert-verified synthesis recipes from open-access literature, which forms the basis of our newly developed benchmark, AlchemyBench. AlchemyBench offers an end-to-end framework that supports research in large language models applied to synthesis prediction. It encompasses key tasks, including raw materials and equipment prediction, synthesis procedure generation, and characterization outcome forecasting. We propose an LLM-as-a-Judge framework that leverages large language models for automated evaluation, demonstrating strong statistical agreement with expert assessments. Overall, our contributions offer a supportive foundation for exploring the capabilities of LLMs in predicting and guiding materials synthesis, ultimately paving the way for more efficient experimental design and accelerated innovation in materials science. © 2025, CC BY-SA.

关键词： Characterization (materials science)

来源：评论

学校读者我要写书评

暂无评论

AI-Driven Cybersecurity and Functional Framework for Remote Laboratories: A software engineering Perspective

AI-Driven Cybersecurity and Functional Framework for Remote ...

引用

Computational Intelligence Approaches and Applications (ICCIAA), International Conference on

作者： Nesreen Otoum Enrique Arias Jesús Jover Alfaro Department of Software Engineering University of Petra Amman Jordan Department of Computer Systems University of Castilla-La Mancha Albacete Spain Albacete Research Institute of Informatics University of Castilla-La Mancha Albacete Spain

ISBN: (数字)9798331523657

ISBN: (纸本)9798331523664

Remote laboratories have reshaped the way engineering courses provide hands-on experimentation by enabling students to control physical lab equipment via online interfaces. nevertheless, their ease, However, these platforms face challenges such as security vulnerabilities, performance instability, and limited user support. This paper introduces a layered framework that integrates artificial intelligence (AI) into a secure, modular design. Using conventional software engineering techniques, our approach improves cybersecurity through intelligent threat detection, automated resource balancing, and context-aware user interaction. We provide a comprehensive explanation of the framework's architecture, which methodically addresses critical requirements. This includes a proactive cybersecurity apparatus that adapts to emergent threats, as well as real-time device control and data acquisition. Ultimately, this work aims to assist educators and institutions build or modernize remote lab environments that offer robust protection and seamless learning outcomes, even under rapidly shifting demands.

关键词： Remote laboratories computer architecture Threat assessment Real-time systems Resource management Reliability computer security Protection Standards software engineering

来源：评论

学校读者我要写书评

暂无评论

A multi-objective cuckoo search algorithm using generalized Lèvy flight and dissimilar egg identification for multispectral image thresholding

引用

Applied Soft Computing 2025年 175卷

作者： Pal, Ramen Roy, Pritam Mallick, Srijon Mukhopadhyay, Somnath Sarkar, Sunita Hinchey, Mike Department of Computer Science and Engineering Assam University Silchar Assam India Cognizant Technology Solutions India Pvt. Ltd. India Samsung Semiconductor India Research India Lero the SFI Research Centre for Software Department of Computer Science & Information Systems University of Limerick Limerick Ireland

Cuckoo Search (CS) stands as a highly efficient meta-heuristic optimization algorithm. Existing literature showcases the ability of CS in multi-objective scenarios, delineated by the three fundamental rules. However, the first rule of the algorithm incurs the cost of a generation to update a single nest, while the third rule necessitates hit-and-trial methods for parameter adjustment. To address these concerns, a multi-objective cuckoo search algorithm is proposed in this paper. The algorithm builds upon a generalized concept of Lèvy Flight for generating new solutions. Problem-specific, constraint-based strategies for identifying the best nest and dissimilar eggs are also introduced. The algorithm is further applied to solve multispectral remote sensing image thresholding problem. Prior studies have underscored the efficiency of entropy and clustering-based thresholding methods over other techniques. Nevertheless, most entropy-based approaches entail converting color images to grayscale before segmentation, potentially sacrificing crucial spectral information and consequently degrading segmentation algorithm's performance. To avoid these limitations, this research introduces an entropy-based thresholding method to segment a color image without converting it to grayscale. The experiments are carried out using very high resolution (VHR) and coarse resolution (CR) multispectral (MS) images from the satellite sensors Pl'eidas-1B and Sentinel-2b, respectively. The proposed methods undergo validation against four state-of-the-art techniques on benchmark functions and six clustering indexes, respectively. © 2025 Elsevier B.V.

关键词： Heuristic algorithms

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：