检索结果-内蒙古大学图书馆

Transfer learning-based few-shot sample deep learning-assisted template attacks

Physical Communication 2025年 72卷

作者： Feng, Zhiying Li, Lang Ou, Yu Deng, Lianrui College of Computer Science and Technology Hengyang Normal University Hengyang421002 China Hunan Provincial Key Laboratory of Intelligent Information Processing and Application Hengyang Normal University Hengyang421002 China

Deep learning-assisted template attack (DLATA) is a novel side-channel attack (SCA) method proposed by Lichao Wu at CHES2022. It utilizes a triplet network to assist template attacks (TA), avoiding the redundant training and hyperparameter tuning required in traditional DL-based SCA methods. However, the training of the triplet network requires a large number of power samples due to its unique structure. We propose a new optimization scheme, in which the transfer learning (TL) technology is used to train multiple models on several similar datasets with fewer power traces, to mitigation the problem. The approach allows us to leverage pre-trained models to product a new mode on the another target dataset by fine-tuning weights so that significantly reduce the training cost for the triplet network while maintaining attack effectiveness. We remould the structure and dimensionality of similar datasets so that the models trained on them can perform effective transfer learning for training on the target dataset. Concretely, some of parameters and features obtained from pretraining can be used directly for the target task, while the rest only require fine-tuning. Evaluation and experimental validation on the public ASCAD dataset demonstrate that our method achieves or even surpasses the performance of the original method with a 90% reduction in the training set. These findings highlight the effectiveness of the proposed TL strategy in achieving robust attack performance in low-sample training environments. © 2025 Elsevier B.V.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

DioR: Adaptive Cognitive Detection and Contextual Retrieval Optimization for Dynamic Retrieval-Augmented Generation

arXiv

引用

arXiv 2025年

作者： Guo, Hanghui Zhu, Jia Di, Shimin Shi, Weijie Chen, Zhangze Xu, Jiajie School of Computer Science and Technology Zhejiang Normal University China The Zhejiang Key Laboratory of Intelligent Education Technology and Application Zhejiang Normal University China School of Computer Science and Engineering Southeast University China Department of Computer Science and Engineering The Hong Kong University of Science and Technology Hong Kong School of Computer Science and Technology Soochow University China

Dynamic Retrieval-augmented Generation (RAG) has shown great success in mitigating hallucinations in large language models (LLMs) during generation. However, existing dynamic RAG methods face significant limitations in two key aspects: 1) Lack of an effective mechanism to control retrieval triggers, and 2) Lack of effective scrutiny of retrieval content. To address these limitations, we propose an innovative dynamic RAG method, DioR (Adaptive Cognitive Detection and Contextual Retrieval Optimization), which consists of two main components: adaptive cognitive detection and contextual retrieval optimization, specifically designed to determine when retrieval is needed and what to retrieve for LLMs is useful. Experimental results demonstrate that DioR achieves superior performance on all tasks, demonstrating the effectiveness of our work. 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

PoW-Based Sybil Attack Resistant Model for P2P Reputation Systems 2nd

PoW-Based Sybil Attack Resistant Model for P2P Reputation Sy...

引用

2nd International Conference on Blockchain and Trustworthy Systems, Blocksys 2020

作者： Li, Biaoqi Fu, Xiaodong Yue, Kun Liu, Li Liu, Lijun Feng, Yong Yunnan Provincial Key Laboratory of Computer Technology Application Kunming650500 China Faculty of Information Engineering and Automation Kunming University of Science and Technology Kunming650500 China School of Information Science and Engineering Yunnan University Kunming650091 China

ISBN: (纸本)9789811592126

Peer-to-peer and other distributed systems are known particularly vulnerable to Sybil attack, so is P2P reputation systems. Reputation helps users to make better decisions in P2P systems, but Sybil attacker can obtain multiple identities and pretends to be multiple. It can control more than "one vote" to let others’ reputation change just as the attacker wants. This paper presents a Sybil attacker resistant model for P2P reputation systems by introducing multiple rounds of PoW (Proof-of-Work) verification and dynamic difficulty adjustment. A savvy attacker usually doesn’t carry out attacks without positive returns, so we minimize the expected profit of attackers through the using of puzzles. The costs and benefits are quantified as effectiveness in our model. We explore the effectiveness in two types of comparison, one is in dynamic or static difficulty adjustment, the other is executing once or multiple rounds of verification, and show the effectiveness of our model both analytically and experimentally. © 2020, Springer Nature Singapore Pte Ltd.

关键词： Peer to peer networks

来源：评论

学校读者我要写书评

暂无评论

Filter Pruning for Efficient CNNs via Knowledge-driven Differential Filter Sampler

arXiv

引用

arXiv 2023年

作者： Lin, Shaohui Huang, Wenxuan Xie, Jiao Zhang, Baochang Shen, Yunhang Yu, Zhou Han, Jungong Doermann, David The School of Computer Science and Technology East China Normal University Shanghai China The Key Laboratory of Advanced Theory and Application in Statistics and Data Science Ministry of Education China The Department of Automation School of Aerospace Engineering Xiamen University Xiamen China The Institute of Artificial Intelligence Beihang University Beijing China Zhongguancun Laboratory Beijing China The Youtu Lab Tencent Shanghai China The School of Statistics East China Normal University Shanghai China The Department of Computer Science University of Sheffield United Kingdom The University at Buffalo BuffaloNY United States

Filter pruning simultaneously accelerates the computation and reduces the memory overhead of CNNs, which can be effectively applied to edge devices and cloud services. In this paper, we propose a novel Knowledge-driven Differential Filter Sampler (KDFS) with Masked Filter Modeling (MFM) framework for filter pruning, which globally prunes the redundant filters based on the prior knowledge of a pre-trained model in a differential and non-alternative optimization. Specifically, we design a differential sampler with learnable sampling parameters to build a binary mask vector for each layer, determining whether the corresponding filters are redundant. To learn the mask, we introduce masked filter modeling to construct PCA-like knowledge by aligning the intermediate features from the pre-trained teacher model and the outputs of the student decoder taking sampling features as the input. The mask and sampler are directly optimized by the Gumbel-Softmax Straight-Through Gradient Estimator in an end-to-end manner in combination with global pruning constraint, MFM reconstruction error, and dark knowledge. Extensive experiments demonstrate the proposed KDFS’s effectiveness in compressing the base models on various datasets. For instance, the pruned ResNet-50 on ImageNet achieves 55.36% computation reduction, and 42.86% parameter reduction, while only dropping 0.35% Top-1 accuracy, significantly outperforming the state-of-the-art methods. The code is available at https://***/Osilly/KDFS. Copyright © 2023, The Authors. All rights reserved.

关键词： Machine learning

来源：评论

学校读者我要写书评

暂无评论

Autonomous Search Investigation for Radioactive Leaked Source Based on Updated Infotaxis Method During Nuclear Emergency Rescue

SSRN

引用

SSRN 2023年

作者： Chen, Liwei Zhou, Cong Wang, Yu Zong, Yiran Lu, Tingting Chen, Chunhua School of Computer and Artificial Intelligence Hefei Normal University Anhui Hefei230601 China Institute of Nuclear Energy Safety Technology Hefei Institutes of Physical Science Chinese Academy of Sciences Anhui Hefei230031 China Key Laboratory for the Application and Transformation of Traditional Chinese Medicine in the Prevention and Treatment of Major Pulmonary Diseases Anhui University of Chinese Medicine Anhui Province Anhui Hefei230012 China Department of Computer Science Hong Kong Baptist University 999077 Hong Kong

Identifying leakage location is one of key issues for off-site decision worker under nuclear accident condition, especially for early emergency rescue. In order to rapidly reduce the environmental radiation level of serious accidents in nuclear facilities caused by external events and identify radioactive residues outside containment, an autonomous search method for residue leakage source in nuclear emergency rescue based on updated Infotaxis method was proposed, taking into account radioactive decay and wet settlement caused by high-pressure sprinkling or precipitation. Sampling rate function was updated by radioactive decay term and washing factor in Infotaxis method, which improves the dilution effect of radiation environment and provides more accurate entropy information. The search trajectories were investigated under multi-scenarios, such as different source emission rate and wet deposition due to high pressure sprinkler or precipitation or not. The results showed that the proposed method could improve search efficiency by adding washing term, especially for improvement of the success rate of leakage source search under higher source emission rate, and it help nuclear emergency decision-makers to trace the location of leakage sources accurately during rescue process, so as to provide more scientific information for early emergency response and consequence assessment. © 2023, The Authors. All rights reserved.

关键词： Radioactivity

来源：评论

学校读者我要写书评

暂无评论

Nanodet-Ghost: A Lightweight Network for Quality Detection of Wheat Kernels Appearance

SSRN

引用

SSRN 2022年

作者： Wang, Yunfei Wang, Yanan Wang, Zheng Li, Rong Hua, Zhixin Zhang, Chuandong Zhang, Zuojing Song, Huaibo College of Mechanical and Electronic Engineering Northwest A&F University Yangling Shaanxi 712100 China Key Laboratory of Agricultural Internet of Things Ministry of Agriculture and Rural Affairs Yangling Shaanxi 712100 China School of Mathematics and Computer Application Technology Jining University Shandong Qufu273155 China

Quality detection of wheat grain appearance is a pre-step for acquiring high-throughput phenotyping and wheat breeding information. However, low portability brought by parameter redundancy is a challenge for general detection methods based on deep learning. To tackle the problem, a lightweight model (Nanodet-Ghost) was designed for this study. Feature extraction structure in GhostNet optimized by ELU was applied to extract features of wheat grains. Shallow and deep features were integrated to improve the distinguishability of wheat grains. DIoU loss was used to improve the detection accuracy and convergence speed. 6414 images of healthy, moldy, damaged, and germinated wheat grains were collected and classified into train set, verification set, and test set with a ratio of about 4:1:1. After 300 epochs of network training, the mAP (mean average accuracy), R (recall), and size of the network were 99.00%, 87.08%, and 11.1 MB. To further test the performance of Nanodet-Ghost, the network was compared with RetinaNet, Yolov5s, and CenterNet in the same data set and training parameters. Nanodet-Ghost achieved a respective increase of 13.24%, 10.10%, and 0.2the 0% in the mAP index compared with RetiaNet, Yolov5s, and CenterNet, and a decrease of 126.9 MB, 2.9 MB, and 65.9 MB in model size. Different grain sizes, adhesion conditions, and light sources were subsequently applied to test the model. The mAP detected by the proposed model were 98.51%, 94.87%, and 85.29% respectively. The overall results showed that Nanodet-Ghost was suitable for quality detection of wheat kernel appearance. © 2022, The Authors. All rights reserved.

关键词： Grain (agricultural product)

来源：评论

学校读者我要写书评

暂无评论

ResfEANet: ResNet-fused External Attention Network for Tuberculosis Diagnosis using Chest X-ray Images

引用

computer Methods and Programs in Biomedicine Update 2024年 5卷

作者： Ejiyi, Chukwuebuka Joseph Qin, Zhen Nnani, Ann O Deng, Fuhu Ejiyi, Thomas Ugochukwu Ejiyi, Makuachukwu Bennedith Agbesi, Victor Kwaku Bamisile, Olusola Network and Data Security Key Laboratory of Sichuan Province University of Electronic Science and Technology of China Chengdu China College of Environmental Science and Engineering Hohai University Nanjing China Department of Pure and Industrial Chemistry University of Nigeria Nsukka Enugu Nigeria Pharmacy Department University of Nigeria Nsukka Enugu Nigeria School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China Sichuan Industrial Internet Intelligent Monitoring and Application Engineering Technology Research Centre Chengdu University of Technology Chengdu China

Pulmonary tuberculosis (TB), the most prevalent form of TB, remains a major global public health concern, contributing to more than a million deaths each year. The accurate and timely diagnosis of this disease is of paramount importance for effective control and treatment. Chest X-ray (CXR) images have emerged as a valuable tool for screening lung diseases, including TB, owing to their cost-effectiveness and non-invasiveness. Despite advancements in technology, the challenges associated with interpreting CXR images persist, primarily due to the scarcity of trained radiologists. This underscores the pressing need for an automated and cost-effective computer-aided system capable of diagnosing TB, assisting medical practitioners in distinguishing between TB-positive and negative CXR scans. In response to this need, we introduce an innovative approach called ResNet-fused External Attention Network (ResfEANet). This network excels in accurately classifying TB from CXR images, achieving remarkable levels of accuracy and sensitivity. ResfEANet is built upon ResNet and incorporates an External Attention mechanism, albeit with fewer residual network blocks than ResNet-50 resulting in a relatively shallow network with fewer layers. This approach proves highly effective in feature extraction and yields competitive results in the classification of TB. Our method was employed to train a model that demonstrated an impressive accuracy rate of 97.59% and a remarkable sensitivity of 100% in binary classification tasks with optimal computational cost. These outcomes suggest that our proposed approach has the potential to serve as a valuable secondary tool in clinical decision-making, providing crucial assistance to radiologists and healthcare professionals. © 2023

关键词： Chest X-ray images External Attention ResfEANet ResNet Tuberculosis

来源：评论

学校读者我要写书评

暂无评论

Stable and Diverse: A Unified Approach for computerized Adaptive Testing

Stable and Diverse: A Unified Approach for Computerized Adap...

引用

IEEE International Conference on Cloud Computing and Intelligence Systems (CCIS)

作者： Yuting Ning Ye Liu Zhenya Huang Haoyang Bi Qi Liu Enhong Chen Dan Zhang Anhui Province Key Laboratory of Big Data Analysis and Application School of Computer Science and Technology University of Science and Technology of China School of Data Science University of Science and Technology of China iFLYTEK Research iFLYTEK CO. LTD

ISBN: (纸本)9781665441506

computerized Adaptive Testing (CAT), aiming to provide personalized tests for each examinee, is an emerging task in the intelligent education field. A CAT system selects questions step by step according to the knowledge states of each examinee, which are estimated by Cognitive Diagnosis Models (CDM). Most existing methods depend on the performance of a single CDM, which is often unstable. Besides, they may select similar questions to generate a test, which to some extent ignores the diversity of selected questions. To this end, in this paper, we propose a novel framework, namely Ensembled computerized Adaptive Testing (EnCAT). Specifically, EnCAT comprises two components, ensemble part and explore part. In the ensemble part, we ensemble multiple CDMs to determine whether a question is informative, which ensures the stability of CAT process. Then, in the explore part, we learn the question representation from the question content and design a mechanism to quantify the similarity of different questions, which avoids the selection of similar questions and is free of expensive human labeling. Finally, extensive experiments are conducted on a real-world dataset, where the experimental results demonstrate the effectiveness of our proposed EnCAT framework with good performance.

关键词： Adaptation models Cloud computing Computational modeling Conferences Education Stability analysis Labeling

来源：评论

学校读者我要写书评

暂无评论

C²F²: Cross-Task Cross-Domain Feature Fusion for Semi-Supervised Change Detection

引用

IEEE Geoscience and Remote Sensing Letters 2024年 21卷

作者： Dongjie Zhang Yuting Hong Xiaojie Qiu Li Dong Diqun Yan Chengbin Peng Faculty of Electrical Engineering and Computer Science and the Key Laboratory of Mobile Network Application Technology of Zhejiang Province Ningbo University Ningbo China Zhejiang Cowain Automation Technology Company Ltd. Ningbo China

Semi-supervised learning for change detection (CD), which significantly reduces the labor costs associated with data annotation, has recently garnered substantial attention. In this study, we propose to enhance traditional semi-supervised learning frameworks by leveraging cross-task cross-domain (CTCD) models, which generate complementary features that differ from standard hidden features. The procedure is as follows. First, the standard features obtained from a traditional encoding–decoding structure are fused with attention-augmented complementary features. Second, a secondary decoder maps the fused heterogeneous features into the label space to obtain high-quality pseudo-labels, offering more precise guidance for semi-supervised learning on traditional structures. This approach improves pseudo-labels by leveraging the strength of CTCD models, including large pretrained models, to enhance the semi-supervised learning process of domain-specific and task-specific models. Experimental results on benchmark datasets demonstrate that our proposed approach surpasses state-of-the-art methods.

关键词： Standards Feature extraction Decoding Training Accuracy Semisupervised learning Supervised learning

来源：评论

学校读者我要写书评

暂无评论

An Incentive Mechanism Considering Privacy Metric for Personalized application

引用

IEEE Transactions on Consumer Electronics 2025年

作者： Xie, Yun Li, Peng Zhang, Jindan Attar, Razaz Waheeb Arya, Varsha Gupta, Brij B. Nanjing Vocational College of Information Technology Jiangsu Province Service Customization Network Application Engineering Research Center Jiangsu Province Service Customization Network Application Engineering Research Center Nanjing210023 China Nanjing University of Posts and Telecommunications School of Computer Science Nanjing210023 China Jiangsu High Technology Research Key Laboratory for Wireless Sensor Networks Nanjing210023 China Xianyang Polytechnic Institute China Princess Nourah bint Abdulrahman University College of Business Administration Management Department Riyadh11671 Saudi Arabia Hong Kong Metropolitan University Hong Kong Hong Kong Center for Interdisciplinary Research Dehradun India Asia University Department of Computer Science and Information Engineering Taichung413 Taiwan China Medical University Hospital China Medical University Department of Medical Research Taichung Taiwan Pune India Korea University School of Cybersecurity Seoul Korea Republic of Kyung Hee University Kyungheedae-ro Seoul26 Korea Republic of

Edge-based federated learning (EFL) is a solution for personalized applications. It addresses the limitations of cloud-based consumer service systems, such as data privacy and network latency. To further enhance the EFL system’s ability to protect data privacy, differential privacy (DP) is introduced. It is crucial to motivate high-quality collaboration among participants for optimal performance of the DP-EFL system. Researchers primarily focus on efficient learning algorithms, optimized resource scheduling, and user-oriented incentive mechanisms. Existing research exhibits notable limitations in modeling and analyzing the heterogeneity of resource costs (e.g., communication costs), particularly failing to comprehensively account for the additional costs imposed by participants’ privacy protection mechanisms‌. To address this gap, we propose a novel market transaction model based on Stackelberg game theory. This framework integrates incentive-driven reward allocation mechanisms to rigorously investigate the collaborative behaviors and strategic interactions between edge nodes (ENs) and the cloud server (CS)‌. We incorporate privacy costs into the multidimensional costs of participants with privacy metric, and based on this, derived collaboration strategies between ENs and CS, thus solving the Stackelberg game problem by theoretical analysis. Furthermore, numerical experiments validated the rationality of our proposed model, the impact of strategies on the performance of collaborators, and the advantages over other incentive mechanisms. © 2025 IEEE. All rights reserved.

关键词： Federated learning

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：