检索结果-内蒙古大学图书馆

Achieving dynamic privacy measurement and protection based on reinforcement learning for mobile edge crowdsensing of IoT

引用

Digital Communications and networks 2024年第2期10卷 380-388页

作者： Renwan Bi Mingfeng Zhao Zuobin Ying Youliang Tian Jinbo Xiong Fujian Provincial Key Laboratory of Network Security and Cryptology College of Computer and Cyber SecurityFujian Normal UniversityFuzhou350117China Guangxi Key Laboratory of Trusted Software Guilin University of Electronic TechnologyGuilin541004China Faculty of Data Science City University of Macao999078MacaoChina State Key Laboratory of Public Big Data College of Computer Science and TechnologyGuizhou UniversityGuiyang550025China

With the maturity and development of 5G field,Mobile Edge CrowdSensing(MECS),as an intelligent data collection paradigm,provides a broad prospect for various applications in ***,sensing users as data uploaders lack a balance between data benefits and privacy threats,leading to conservative data uploads and low revenue or excessive uploads and privacy *** solve this problem,a Dynamic Privacy Measurement and Protection(DPMP)framework is proposed based on differential privacy and reinforcement ***,a DPM model is designed to quantify the amount of data privacy,and a calculation method for personalized privacy threshold of different users is also ***,a Dynamic Private sensing data Selection(DPS)algorithm is proposed to help sensing users maximize data benefits within their privacy ***,theoretical analysis and ample experiment results show that DPMP framework is effective and efficient to achieve a balance between data benefits and sensing user privacy protection,in particular,the proposed DPMP framework has 63%and 23%higher training efficiency and data benefits,respectively,compared to the Monte Carlo algorithm.

关键词： Mobile edge crowdsensing Dynamic privacy measurement Personalized privacy threshold Privacy protection Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

LINKAGE: Listwise Ranking among Varied-Quality References for Non-Factoid QA Evaluation via LLMs

LINKAGE: Listwise Ranking among Varied-Quality References fo...

引用

2024 Conference on Empirical Methods in Natural Language Processing, EMNLP 2024

作者： Yang, Sihui Bi, Keping Cui, Wanqing Guo, Jiafeng Cheng, Xueqi CAS Key Laboratory of Network Data Science and Technology Institute of Computing Technology Chinese Academy of Sciences China University of Chinese Academy of Sciences Beijing China

ISBN: (纸本)9798891761681

Non-Factoid (NF) Question Answering (QA) is challenging to evaluate due to diverse potential answers and no objective criterion. The commonly used automatic evaluation metrics like ROUGE or BERTScore cannot accurately measure semantic similarities or answers from different perspectives. Recently, Large Language Models (LLMs) have been resorted to for NFQA evaluation due to their compelling performance on various NLP tasks. Common approaches include pointwise scoring of each candidate answer and pairwise comparisons between answers. Inspired by the evolution from pointwise to pairwise to listwise in learning-to-rank methods, we propose a novel listwise NFQA evaluation approach, that utilizes LLMs to rank candidate answers in a list of reference answers sorted by descending quality. Moreover, for NF questions that do not have multi-grade or any golden answers, we leverage LLMs to generate the reference answer list of various quality to facilitate the listwise evaluation. Extensive experimental results on three NFQA datasets, i.e., ANTIQUE, the TREC-DL-NF, and WebGLM show that our method has significantly higher correlations with human annotations compared to automatic scores and common pointwise and pairwise approaches. Our code and dataset can be found at https://***/babyyang525/LINKAGE-Listwise-NFQA-Evaluation. © 2024 Association for Computational Linguistics.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Blockchain and signcryption enabled asynchronous federated learning framework in fog computing

引用

Digital Communications and networks 2025年第2期11卷 442-454页

作者： Zhou Zhou Youliang Tian Jinbo Xiong Changgen Peng Jing Li Nan Yang State Key Laboratory of Public Big Data College of Computer Science and TechnologyGuizhou UniversityGuiyang 550025China Fujian Provincial Key Laboratory of Network Security and Cryptology College of Computer and Cyber SecurityFujian Normal UniversityFuzhou 350117China Guangxi Key Laboratory of Cryptography and Information Security Guilin University of Electronic TechnologyGuilin 541004China

Federated learning combines with fog computing to transform data sharing into model sharing,which solves the issues of data isolation and privacy disclosure in fog ***,existing studies focus on centralized single-layer aggregation federated learning architecture,which lack the consideration of cross-domain and asynchronous robustness of federated learning,and rarely integrate verification mechanisms from the perspective of *** address the above challenges,we propose a Blockchain and Signcryption enabled Asynchronous Federated Learning(BSAFL)framework based on dual aggregation for cross-domain *** particular,we first design two types of signcryption schemes to secure the interaction and access control of collaborative learning between ***,we construct a differential privacy approach that adaptively adjusts privacy budgets to ensure data privacy and local models'availability of intra-domain ***,we propose an asynchronous aggregation solution that incorporates consensus verification and elastic participation using ***,security analysis demonstrates the security and privacy effectiveness of BSAFL,and the evaluation on real datasets further validates the high model accuracy and performance of BSAFL.

关键词： Blockchain Signcryption Federated learning Asynchronous Fog computing

来源：评论

学校读者我要写书评

暂无评论

Underwater target recognition based on adaptive multi-feature fusion network

引用

Multimedia Tools and Applications 2025年第10期84卷 7297-7317页

作者： Pan, Xiaoying Sun, Jia Feng, TianHao Lei, MingZhu Wang, Hao Zhang, WuXia Shaanxi Key Laboratory of Network Data Analysis and Intelligent Processing Xi’an710121 China Xi’an Key Laboratory of Big Data and Intelligent Computing Xi’an710121 China School of Computer Science and Technology Xi’an University of Post & Telecommunications Xi’an710121 China

Due to the complexity of the underwater environment, underwater acoustic target recognition is more challenging than ordinary target recognition, and has become a hot topic in the field of underwater acoustics research. In recent years, deep learning has been widely used in underwater recognition due to its powerful feature learning capabilities. However, most deep learning-based underwater target recognition methods use only the time-frequency domain features of signals, ignoring the frequency domain image features of signals. Additionally, they only use simple summation or splicing strategies to fuse and discriminate the extracted features, which leads to limited improvement in the accuracy of underwater target recognition. To address these issues, this paper proposes an underwater target recognition method based on an adaptive multi-feature fusion network. The method consists of a data pre-processing module, a multi-dimensional feature extraction module, and an adaptive multi-feature fusion module. In the multi-dimensional feature extraction module, the long short time memory network (LSTM) and the one-dimensional convolutional neural network (1DCNN) are used to extract the time-frequency features of the underwater signal. Furthermore, the two-dimensional convolutional neural network (2DCNN) is used to extract the image features in the frequency domain. The adaptive multi-feature fusion module uses an attention mechanism for adaptive weighted fusion to make full use of the learned features. The effectiveness of the proposed method is validated on the ShipEar dataset, and the recognition accuracy reaches 94.92%, which is higher than other existing methods. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Generative Retrieval Meets Multi-Graded Relevance 38

Generative Retrieval Meets Multi-Graded Relevance

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Tang, Yubao Zhang, Ruqing Guo, Jiafeng de Rijke, Maarten Chen, Wei Cheng, Xueqi CAS Key Lab of Network Data Science and Technology ICT CAS China University of Chinese Academy of Sciences China University of Amsterdam Netherlands

Generative retrieval represents a novel approach to information retrieval. It uses an encoder-decoder architecture to directly produce relevant document identifiers (docids) for queries. While this method offers benefits, current approaches are limited to scenarios with binary relevance data, overlooking the potential for documents to have multi-graded relevance. Extending generative retrieval to accommodate multi-graded relevance poses challenges, including the need to reconcile likelihood probabilities for docid pairs and the possibility of multiple relevant documents sharing the same identifier. To address these challenges, we introduce a framework called GRaded Generative Retrieval (GR2). GR2 focuses on two key components: ensuring relevant and distinct identifiers, and implementing multi-graded constrained contrastive training. First, we create identifiers that are both semantically relevant and sufficiently distinct to represent individual documents effectively. This is achieved by jointly optimizing the relevance and distinctness of docids through a combination of docid generation and autoencoder models. Second, we incorporate information about the relationship between relevance grades to guide the training process. We use a constrained contrastive training strategy to bring the representations of queries and the identifiers of their relevant documents closer together, based on their respective relevance grades. Extensive experiments on datasets with both multi-graded and binary relevance demonstrate the effectiveness of GR. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

SmartZone: Runtime Support for Secure and Efficient On-Device Inference on ARM TrustZone

引用

IEEE Transactions on Computers 2025年第6期74卷 2144-2158页

作者： Jian, Zhaolong Liu, Xu Dong, Qiankun Cheng, Longkai Xie, Xueshuo Li, Tao Nankai University College of Computer Science Tianjin300350 China Tianjin Key Laboratory of Network and Data Science Technology Tianjin300350 China Haihe Lab of ITAI Tianjin300450 China

On-device inference is a burgeoning paradigm that performs model inference locally on end devices, allowing private data to remain local. ARM TrustZone as a widely supported trusted execution environment has been applied to provide confidentiality protection for on-device inference. However, with the rise of large-scale models like large language models (LLMs), TrustZone-based on-device inference faces challenges in migration difficulties and inefficient execution. The rudimentary TEE OS on TrustZone lacks both the inference runtime needed for building models and the parallel support necessary to accelerate inference. Moreover, the limited secure memory resources on end devices further constrain the model size and degrade performance. In this paper, we propose SmartZone to provide runtime support for secure and efficient on-device inference on TrustZone. SmartZone consists three main components: (1) a trusted inference-oriented operator set, providing the underlying mechanisms adapted to the TrustZone’s execution mode for trusted inference of DNN models and LLMs. (2) the proactive multi-threading parallel support, which increases the number of CPU cores in the secure state via cross-world thread collaboration to achieve parallelism, and (3) the on-demand secure memory management method, which statically allocates the appropriate secure memory size based on pre-execution resource analysis. We implement a prototype of SmartZone on the Raspberry Pi 3B+ board and evaluate it on four well-known DNN models and llama2 LLM. Extensive experimental results show that SmartZone provides end-to-end protection for on-device inference while maintaining excellent performance. Compared to the origin trusted inference, SmartZone accelerates the inference speed by up to 4.26× and reduces energy consumption by 65.81%. © 1968-2012 IEEE.

关键词： Memory management

来源：评论

学校读者我要写书评

暂无评论

A Linked List Encryption Scheme for Image Steganography without Embedding

引用

Computer Modeling in Engineering & sciences 2024年第10期141卷 331-352页

作者： Pengbiao Zhao Qi Zhong Jingxue Chen Xiaopei Wang Zhen Qin Erqiang Zhou Network and Data Security Key Laboratory of Sichuan Province University of Electronic Science and Technology of ChinaChengdu610054China Faculty of Data Science City University of MacaoMacao999078China Department of Computer Science and Engineering University of CaliforniaRiversideCA 92521USA

Information steganography has received more and more attention from scholars nowadays,especially in the area of image steganography,which uses image content to transmit information and makes the existence of secret information *** enhance concealment and security,the Steganography without Embedding(SWE)method has proven effective in avoiding image distortion resulting from cover *** this paper,a novel encrypted communication scheme for image SWE is *** reconstructs the image into a multi-linked list structure consisting of numerous nodes,where each pixel is transformed into a single node with data and pointer *** employing a special addressing algorithm,the optimal linked list corresponding to the secret information can be *** receiver can restore the secretmessage fromthe received image using only the list header position *** scheme is based on the concept of coverless steganography,eliminating the need for any modifications to the cover *** boasts high concealment and security,along with a complete message restoration rate,making it resistant to ***,this paper proposes linked-list construction schemeswithin theproposedframework,which caneffectively resist a variety of attacks,includingnoise attacks and image compression,demonstrating a certain degree of *** validate the proposed framework,practical tests and comparisons are conducted using multiple *** results affirm the framework’s commendable performance in terms of message reduction rate,hidden writing capacity,and robustness against diverse attacks.

关键词： Steganography encryption steganography without embedding coverless steganography

来源：评论

学校读者我要写书评

暂无评论

Pretraining data Detection for Large Language Models: A Divergence-based Calibration Method

Pretraining Data Detection for Large Language Models: A Dive...

引用

2024 Conference on Empirical Methods in Natural Language Processing, EMNLP 2024

作者： Zhang, Weichao Zhang, Ruqing Guo, Jiafeng de Rijke, Maarten Fan, Yixing Cheng, Xueqi CAS Key Lab of Network Data Science and Technology ICT CAS Beijing China University of Chinese Academy of Sciences Beijing China Zhongguancun Laboratory Beijing China University of Amsterdam Amsterdam Netherlands

ISBN: (纸本)9798891761643

As the scale of training corpora for large language models (LLMs) grows, model developers become increasingly reluctant to disclose details on their data. This lack of transparency poses challenges to scientific evaluation and ethical deployment. Recently, pretraining data detection approaches, which infer whether a given text was part of an LLM's training data through black-box access, have been explored. The Min-K% Prob method, which has achieved state-of-the-art results, assumes that a non-training example tends to contain a few outlier words with low token probabilities. However, the effectiveness may be limited as it tends to misclassify non-training texts that contain many common words with high probabilities predicted by LLMs. To address this issue, we introduce a divergence-based calibration method, inspired by the divergence-from-randomness concept, to calibrate token probabilities for pretraining data detection. We compute the cross-entropy (i.e., the divergence) between the token probability distribution and the token frequency distribution to derive a detection score. We have developed a Chinese-language benchmark, PatentMIA, to assess the performance of detection approaches for LLMs on Chinese text. Experimental results on English-language benchmarks and PatentMIA demonstrate that our proposed method significantly outperforms existing methods. Our code and PatentMIA benchmark are available at https://***/zhang-wei-chao/DC-PDD. © 2024 Association for Computational Linguistics.

关键词： Calibration

来源：评论

学校读者我要写书评

暂无评论

Sensing-Communication-Computation Integration for Federated Edge Learning with Controllable Model Dropout

引用

IEEE Internet of Things Journal 2025年第12期12卷 19767-19781页

作者： Jiao, Xiang Zhu, Guangxu Jiang, Wei Chen, Li Luo, Wu Wen, Dingzhu State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Peking University Beijing China Shenzhen Research Institute of Big Data Shenzhen China The State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Peking University Beijing China The CAS Key Laboratory of Wireless Optical Communication University of Science and Technology of China Hefei China Network Intelligence Center School of Information Science and Technology ShanghaiTech University Shanghai China

FEderated Edge Learning (FEEL) is an advanced paradigm in edge artificial intelligence, enabling privacy-preserving collaborative model training through periodic communication between edge devices and a central server. FEEL involves three key processes: sensing, computation, and communication for data acquisition, processing, and exchange, respectively. Due to limited system resources, optimizing each process individually may lead to suboptimal learning performance. This challenge has sparked research into Integrated Sensing-Computation-Communication (ISCC) design for enhanced FEEL. While previous work has optimized general learning parameters such as batch size and computing frequency, there is a lack of customized designs considering the neural network architecture as an optimizable variable in ISCC for FEEL. To close this gap, we introduce a novel design where each device generates a sub-model through controllable weight dropout, adding flexibility by directly manipulating the learning process and reducing computation and communication overhead. To guide ISCC resource allocation in this new setting, we present a comprehensive convergence analysis, revealing the tight coupling of sensing, computation, and communication across devices and their impact on FEEL convergence. Building on these theoretical insights, we formulate an ISCC problem aiming to maximize the FEEL convergence rate through joint optimization of variables such as batch size, sensing power, dropout rate, and communication power. This non-convex problem is decomposed into two sub-problems via alternating optimization: one controls batch size using a sorting algorithm, while the other focuses on ISCC device parameters, transformable into a convex problem solved by successive convex approximation. Extensive experiments using human motion recognition datasets demonstrate the superiority of the proposed design over baseline schemes. © 2014 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

ADLT: a novel method for effective detection of adversarial attacks in image retrieval

ADLT: a novel method for effective detection of adversarial ...

引用

2024 International Conference on Computer Vision and Image Processing, CVIP 2024

作者： Xu, Shiyue Zhang, Yu College of Cyber Science Tianjin Key Laboratory of Network and Data Security Technology Nankai University Tianjin300350 China

ISBN: (纸本)9781510688384

The versatile application of deep learning has made it prone to vulnerabilities, particularly adversarial attacks, which have been increasingly prevalent as its scope expands. Attack methods based on adversarial queries pose significant challenges to image retrieval models, disrupting their functionality. While numerous defense methods have been proposed for adversarial query attacks in image classification, approaches to detecting or defending against such attacks in image retrieval remain limited due to inherent challenges in adapting these methods. Based on these observations, we introduce the Augmentation Detection based on Top-K-List (ADTL) method, the first effective approach for detecting adversarial attacks in image retrieval. ADTL applies image augmentation to a query and determines its adversarial nature by calculating the divergence in the Top-K list before and after augmentation. This paper presents a method that employs suitable image augmentation techniques and K-Reciprocal clustering for pre-classifying images in the gallery to ease divergence calculations, which has been tested on the Oxford5k and Paris6k datasets and has shown remarkable effectiveness with a high detection accuracy of 87.3%. © 2025 SPIE.

关键词： Adversarial perturbation Attack detection Image augmentation Image retrieval

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：