检索结果-内蒙古大学图书馆

Robust federated contrastive recommender system against targeted model poisoning attack

Science China(Information Sciences) 2025年第4期68卷 50-65页

作者： Wei YUAN Chaoqun YANG Liang QU Guanhua YE Quoc Viet Hung NGUYEN Hongzhi YIN School of Electrical Engineering and Computer Science The University of Queensland School of Information and Communication Technology Griffith University Deep Neural Computing Company Limited

Federated recommender systems(FedRecs) have garnered increasing attention recently, thanks to their privacypreserving benefits. However, the decentralized and open characteristics of current FedRecs present at least two ***, the performance of FedRecs is compromised due to highly sparse on-device data for each client. Second, the system's robustness is undermined by the vulnerability to model poisoning attacks launched by malicious users. In this paper, we introduce a novel contrastive learning framework designed to fully leverage the client's sparse data through embedding augmentation, referred to as CL4FedRec. Unlike previous contrastive learning approaches in FedRecs that necessitate clients to share their private parameters, our CL4FedRec aligns with the basic FedRec learning protocol, ensuring compatibility with most existing FedRec implementations. We then evaluate the robustness of FedRecs equipped with CL4FedRec by subjecting it to several state-of-the-art model poisoning attacks. Surprisingly, our observations reveal that contrastive learning tends to exacerbate the vulnerability of FedRecs to these attacks. This is attributed to the enhanced embedding uniformity, making the polluted target item embedding easily proximate to popular items. Based on this insight, we propose an enhanced and robust version of CL4FedRec(rCL4FedRec) by introducing a regularizer to maintain the distance among item embeddings with different popularity levels. Extensive experiments conducted on four commonly used recommendation datasets demonstrate that rCL4FedRec significantly enhances both the model's performance and the robustness of FedRecs.

关键词： federated recommender system contrastive learning model poisoning attack and defense

来源：评论

学校读者我要写书评

暂无评论

Two-Stage Planning for Smart Buildings With Flexible Heating Load Considering Climate Change Induced Heat Waves

引用

IEEE Transactions on Smart Grid 2025年第3期16卷 2012-2025页

作者： Zhao, Tianyang Xu, Qianwen KTH Royal Institute of Technology School of Electrical Engineering and Computing Sciences Stockholm114 58 Sweden

Building energy planning is significantly challenged by climate change, particularly the increasing frequency of heat waves impacting heating and cooling demands. Current planning methodologies neglect the impacts of heat waves on energy consumption and do not accurately model the temperature-dependent performance of heat pumps (HPs). This paper addresses the critical issue of designing energy-efficient and climate-resilient buildings through optimal resource configuration under uncertain weather conditions. A two-stage stochastic optimization model for building energy system planning is proposed. In the first stage, the capacities of energy resources are optimized;in the second stage, operational strategies under various weather scenarios are determined. A novel long-term load forecasting method using morphing techniques is developed to generate scenario trees accounting for both normal conditions and heat waves, capturing the impact of climate change on energy demand. Additionally, a temperature-dependent HP model with finite partial output levels is introduced, improving upon existing fixed coefficient of performance models to reflect practical operational characteristics. Simulation results on a real educational building in Stockholm demonstrate the effectiveness of the approach, showing an 8.33% reduction in heating capacity requirements and a 62.14% decrease in solution time, enhancing both resilience and computational efficiency. © 2010-2012 IEEE.

关键词： Stochastic systems

来源：评论

学校读者我要写书评

暂无评论

SGGAA: a robust adversarial generation based on synthesizing point cloud and mesh space

引用

Neural computing and Applications 2025年 1-23页

作者： Hu, Ruihan Yang, Rui Faculty of Computing Harbin Institute of Technology Harbin China School of Electrical and Information Engineering Tianjin University Tianjin China

3D vision recognition offers a significantly more robust tool for achieving machine cognition compared to traditional 2D vision techniques. However, similar to the vulnerabilities present in 2D vision, many 3D vision recognition applications, including point cloud classification models, are susceptible to degradation when confronted with imperceptible outliers. Outliers have a considerable impact on the performance and accuracy of models, compromising their overall effectiveness. Existing 3D point cloud attack and defense strategies neglect outliers and overlook the crucial impact of 3D printability in real-world scenarios. Neglecting this aspect hinders the consideration of potential manipulation or alteration of 3D objects in the physical realm, posing significant implications for the security and robustness of 3D vision systems. This manuscript presents a novel and effective targeted attack framework called synthesizing geometry and geodesic attack (SGGAA). It offers an alternative adversarial training algorithm that bridges the gap between the digital and physical domains. SGGAA introduces a novel generalized distance metric, facilitating the design of powerful and simple techniques for generating adversarial samples and improving the training process. It focuses on the regularizer loss object in synthesizing attack modes for mesh and point cloud spaces, utilizing Hausdorff distance, geometry distance, and geodesic distance losses. Comprehensive evaluations, including hyperparameter tuning, qualitative analysis, and quantitative experiments, were conducted to assess SGGAA’s performance. The results demonstrated its superiority, with high attack success rates, excellent geometric similarity scores, and outperformance of state-of-the-art attack strategies across various 3D defense techniques. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Mesh generation

来源：评论

学校读者我要写书评

暂无评论

CRIS-b: A High-Speed Unified Modulo Reduction Algorithm and Hardware Architecture for CRYSTALS-Kyber 38

CRIS-b: A High-Speed Unified Modulo Reduction Algorithm and ...

引用

38th International Conference on VLSI Design, VLSID 2025

作者： Majumdar, Alip Shrestha, Rahul School of Computing and Electrical Engineering Indian Institute of Technology Mandi India

ISBN: (纸本)9798331522445

This paper presents highly reliable algorithm and high-speed hardware architecture for a unified modulo reduction for CRYSTALS-Kyber. This new architecture for modulo reduction is capable of operating at a maximum clock frequency of 351.62 MHz on Artix-7 FPGA platform, consuming a latency of 5 clock cycles, which is approximately 35.24% faster than state-of-the-art implementations. In this work, we have alleviated the size expansion problem of intermediate value of the conventional Barrett algorithm (which is one of the fastest and most reliable algorithm for modulo reduction, reported in literature) by suggesting a new technique, referred as 'Conditional R-index Switching for b-value Minimization (CRIS-b)'. We propose a unified modulo reduction design handling both positive and negative spectrum numbers for all arithmetic modulo operations in a butterfly unit of Number Theoretic Transform(NTT). © 2025 IEEE.

关键词： Quantum cryptography

来源：评论

学校读者我要写书评

暂无评论

Context-Aware Selection of Machine Learning as a Service (MLaaS) in IoT Environments

Context-Aware Selection of Machine Learning as a Service ...

引用

PhD Symposium, Posters, Demos, and A Web for more inclusive, sustainable and prosperous societies, WEB-for-GOOD 2024 and 1st International Workshop on AI and Web Data Analytics, AIWDA 2024 form the 25th International Conference on Web Information Systems engineering, WISE 2024

作者： Patel, Keya Mistry, Sajib Kanneganti, Deepak Krishna, Aneesh School of Electrical Engineering Computing and Mathematical Sciences Curtin University Perth Australia

ISBN: (纸本)9789819614820

The integration of Machine Learning as a Service (MLaaS) into the Internet of Things (IoT) environments presents considerable opportunities for enhancing decision-making and automation. We propose a novel framework for context-aware selection of MLaaS in IoT settings, aimed at optimising the interaction between IoT users’ activities and machine learning services. Our framework considers various contextual dimensions, such as user preferences, locations, IoT device capabilities, and application requirements, to develop a dynamic selection process. By employing context-aware algorithms, our approach seeks to enhance the efficiency, accuracy, and responsiveness of IoT systems. We propose a context change analysis algorithm based on support vector machines (SVM). We develop a contextual bandits algorithm along with skyline queries to achieve optimal mapping between abstract MLaaS services and concrete MLaaS services for quality of service (QoS) attributes. Experiments conducted with real-world and simulated datasets demonstrate the effectiveness of our proposed methods. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Dual Averaging for Distributed Unbalanced Optimization with Delayed Information

引用

IEEE Transactions on Signal and Information Processing over Networks 2025年 11卷 366-377页

作者： Huang, Qing Fan, Yuan Cheng, Songsong Anhui University Key Laboratory of Intelligent Computing and Signal Processing of Ministry of Education School of Electrical Engineering and Automation Anhui Hefei230601 China

In this paper, we study a category of distributed constrained optimization problems where each agent has access to local information, communicates with its neighbors, and cooperatively minimizes the aggregated cost functions over time-varying unbalanced graphs. To address the considered problems, we propose a distributed dual averaging algorithm based on a row-stochastic weighted matrix (DDAR), which improves the robustness of network topology compared to conventional push-sum algorithms. Moreover, we develop a modified version of DDAR with delayed information (DDARD), which considers the delays of both network communication and gradient calculation, enhancing the algorithm's flexibility in communication and iteration. Our analysis demonstrates that the DDAR and DDARD achieve the optimal value at rates of O(N/(1-λ)√T) and O(τ2N/(1-λ}})√T}), respectively. Finally, the theoretical results are confirmed by simulation on a logistic regression problem. © 2015 IEEE.

关键词： Cost functions

来源：评论

学校读者我要写书评

暂无评论

RAHBA-Net: Implementation of Rand Augment and Hybrid Bat Algorithm to Optimize Convolutional Neural Network Performance in Classifying Coffee Plant Diseases

RAHBA-Net: Implementation of Rand Augment and Hybrid Bat Alg...

引用

2025 International Conference on Advancement in Data Science, E-learning and Information System, ICADEIS 2025

作者： Putri, Yusnita Nuzeren, Prins Naval Kurniawan, Isman School of Electrical Engineering Telkom University Bandung Indonesia School of Computing Telkom University Bandung Indonesia

ISBN: (纸本)9798331513320

Coffee is one of the most popular beverages, with approximately 1.6 billion cups consumed daily worldwide. To meet global demand, global coffee production must increase by at least 50% by 2050. However, coffee plants in every country are susceptible to pests and diseases, posing a significant challenge for nations to improve both the quality and quantity of coffee production. Leaves are the primary source for detecting diseases in coffee plants. Rust, miner, and phoma are the most common diseases that affect coffee plants. Experts often use conventional methods to identify diseases through visual inspection, but these methods are time-consuming, costly, prone to errors, and impractical. Machine learning and image processing have also been widely used in classifying coffee plant diseases, yet there is still limited implementation of image processing, with many still relying on manual tuning. Therefore, this research proposes the implementation of RandAugment and HBA to optimize the performance of CNN architectures in classifying coffee plant diseases. The proposed model achieved the best accuracy of 98.81% in the first model and the best parameters of 0.376 million in the second model, ensuring both efficiency and effectiveness in classifying coffee plant diseases. © 2025 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Some Considerations for the Preservation of Endangered Languages Using Low-Resource Machine Translation 37th

Some Considerations for the Preservation of Endangered La...

引用

37th Australasian Joint Conference on Artificial Intelligence, AJCAI 2024

作者： Kho, Alastair Pham, Duc-Son Soon, Susannah Chan, Kit Yan School of Electrical Engineering Computing and Mathematical Sciences Curtin University BentleyWA Australia

ISBN: (纸本)9789819603473

While widespread languages remain actively prevalent in digital mediums, endangered languages such as Indigenous Australian languages, are often scarce in textual resources and lack a substantial digital presence. This diminishes the survivability of their language and cultural heritages, making language preservation initiatives important. Neural Machine Translation (NMT) efforts for low-resource languages can help accelerate the digitisation and preservation of such languages by further enabling translation, information access, and second-language acquisition. This study explores the challenges and considerations with low-resource machine translation (MT), specifically focusing on primarily oral Indigenous Australian languages with minimally available textual resources. Additionally, we explore the existing challenges in the search for quality data and ethical research considerations in approaching Indigenous Cultural Intellectual Property (ICIP). As NMT performance often scales with the quality and quantity of multilingual corpora, we explore promising alternatives such as leveraging large language models (LLMs) to tackle severely low-resource MT as a few-shot prompting translation task. By employing a data imputation approach inspired by Continuous Bag-of-Words (CBOW) to strengthen a prompt’s contextual relevancy, we enhance translations generated by LLMs, achieving a chrF score of 37.3 on imputed data, compared to a baseline of 31.6 with GPT-3.5, and 39.3 compared to a baseline of 38.3 on GPT-4. Through our work, we hope to establish a foundation for future efforts in preserving Indigenous Australian languages. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Computer aided language translation

来源：评论

学校读者我要写书评

暂无评论

meMIA: Multilevel Ensemble Membership Inference Attack

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2025年第1期6卷 93-106页

作者： Ullah, Najeeb Aman, Muhammad Naveed Sikdar, Biplab National University of Singapore Department of Electrical and Computer Engineering 117583 Singapore University of Nebraska-Lincoln School of Computing LincolnNE68588 United States

Leakage of private information in machine learning models can lead to breaches of confidentiality, identity theft, and unauthorized access to personal data. Ensuring the safe and trustworthy deployment of AI systems necessitates addressing privacy concerns to prevent unintentional disclosure and discrimination. One significant threat, membership inference (MI) attacks, exploit vulnerabilities in target learning models to determine if a given sample was part of the training set. However, the effectiveness of existing MI attacks is often limited by the number of classes in the dataset or the need for diverse multilevel adversarial features to exploit overfitted models. To enhance MI attack performance, we propose meMIA, a novel framework based on stacked ensemble learning. meMIA integrates embeddings from a neural network (NN) and a long short-term memory (LSTM) model, training a subsequent NN, termed the meta-model, on the concatenated embeddings. This method leverages the complementary strengths of NN and LSTM models;the LSTM captures order differences in confidence scores, while the NN discerns probability distribution differences between member and nonmember samples. We extensively evaluate meMIA on seven benchmark datasets, demonstrating that it surpasses current state-of-the-art MI attacks, achieving accuracy up to 94.6% and near-perfect recall. meMIA's superior performance, especially on datasets with fewer classes, underscores the urgent need for robust defenses against privacy attacks in machine learning, contributing to the safer and more ethical use of AI technologies. © 2024 IEEE.

关键词： Information leakage

来源：评论

学校读者我要写书评

暂无评论

Robust facial expression recognition via lightweight reinforcement learning for rehabilitation robotics

引用

Optoelectronics Letters 2025年第2期21卷 97-104页

作者： CHEN Yifan FAN Weiming GAO Hongwei YU Jiahui JU Zhaojie School of Computing University of PortsmouthPO13HEUK School of Automation and Electrical Engineering Shenyang Ligong UniversityShenyang 110159China Department of Biomedical Engineering Zhejiang UniversityHangzhou 310027China

This paper proposes a lightweight reinforcement network (LRN) and auxiliary label distribution learning (ALDL)based robust facial expression recognition (FER) *** designed representation reinforcement (RR) network mainly comprises two modules,i.e.,the RR module and the auxiliary label space construction (ALSC) *** RR module highlights key feature messaging nodes in feature maps,and ALSC allows multiple labels with different intensities to be linked to one ***,LRN has a more robust feature extraction capability when model parameters are greatly reduced,and ALDL is proposed to contribute to the training effect of LRN in the condition of ambiguous training *** tested our method on FER-Plus and RAF-DB datasets,and the experiment demonstrates the feasibility of our method in practice during rehabilitation robots.

关键词： rehabilitation lightweight auxiliary

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：