检索结果-内蒙古大学图书馆

Adaptive Backdoor Attacks with Reasonable Constraints on Graph Neural Networks

IEEE Transactions on Dependable and Secure Computing 2025年

作者： Dong, Xuewen Li, Jiachen Li, Shujun You, Zhichao Qu, Qiang Kholodov, Yaroslav Shen, Yulong The School of Computer Science and Technology Xidian University The Engineering Research Center of Blockchain Technology Application and Evaluation Ministry of Education China The Shaanxi Key Laboratory of Blockchain and Secure Computing Xi’an710071 China The School of Computing Kent Interdisciplinary Research Centre in Cyber Security University of Kent CanterburyCT2 7NF United Kingdom Shenzhen Institute of Advanced Technology Chinese Academy of Sciences China The Intelligent Transportation Systems Lab Innopolis University Innopolis Russia The School of Computer Science and Technology Xidian University China

Recent studies show that graph neural networks (GNNs) are vulnerable to backdoor attacks. Existing backdoor attacks against GNNs use fixed-pattern triggers and lack reasonable trigger constraints, overlooking individual graph characteristics and rendering insufficient evasiveness. To tackle the above issues, we propose ABARC, the first Adaptive Backdoor Attack with Reasonable Constraints, applying to both graph-level and node-level tasks in GNNs. For graph-level tasks, we propose a subgraph backdoor attack independent of the graph’s topology. It dynamically selects trigger nodes for each target graph and modifies node features with constraints based on graph similarity, feature range, and feature type. For node-level tasks, our attack begins with an analysis of node features, followed by selecting and modifying trigger features, which are then constrained by node similarity, feature range, and feature type. Furthermore, an adaptive edge-pruning mechanism is designed to reduce the impact of neighbors on target nodes, ensuring a high attack success rate (ASR). Experimental results show that even with reasonable constraints for attack evasiveness, our attack achieves a high ASR while incurring a marginal clean accuracy drop (CAD). When combined with the state-of-the-art defense randomized smoothing (RS) method, our attack maintains an ASR over 94%, surpassing existing attacks by more than 7%. © 2004-2012 IEEE. All rights reserved.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

More than Encoder: Introducing Transformer Decoder to Upsample

More than Encoder: Introducing Transformer Decoder to Upsamp...

引用

2022 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2022

作者： Li, Yijiang Cai, Wentian Gao, Ying Li, Chengming Hu, Xiping South China University of Technology School of Computer Science and Engineering Guangzhou China Guangdong Prov. Key Laboratory of Artificial Intelligence in Medical Image Analysis and Application Guangdong Provincial People's Hospital Guangdong Academy of Medical Sciences Guangzhou China Sun Yat-sen University School of Intelligent Systems Engineering Shenzhen China Beijing Institute of Technology School of Medical Technology Beijing China

ISBN: (纸本)9781665468190

Medical image segmentation methods downsample images for feature extraction and then upsample them to restore resolution for pixel-level predictions. In such schema, upsample technique is vital in restoring information for better performance. However, existing upsample techniques leverage little information from downsampling paths. The local and detailed feature from the shallower layer such as boundary and tissue texture is crucial in segmentation, especially medical image segmentation. To this end, we propose a novel upsample approach for medical image segmentation, Window Attention Upsample (WAU), which upsamples features conditioned on local and detailed features from downsampling path in local windows by introducing attention decoders of Transformer. WAU could serve as a general upsample method and be incorporated into any segmentation model that possesses lateral connections. We first propose the Attention Upsample which consists of Attention Decoder (AD) and bilinear upsample. AD leverages pixel-level attention to model longrange dependency and global information for a better upsample. Bilinear upsample is introduced as the residual connection to complement the upsampled features. Moreover, considering the extensive memory and computation cost of pixel-level attention, we further design a window attention scheme to restrict attention computation in local windows instead of the global range. We evaluate our method (WAU) on classic UNet structure with lateral connections and achieve state-of-the-art performance on Medical Segmentation Decathlon (MSD) Brain and Automatic Cardiac Diagnosis Challenge (ACDC) datasets. We also validate the effectiveness of our method on multiple classic architectures and achieve consistent improvement. © 2022 IEEE.

关键词： Diagnosis

来源：评论

学校读者我要写书评

暂无评论

A Multi-Feature Fusion Method with Inception Architecture for Image Steganalysis

A Multi-Feature Fusion Method with Inception Architecture fo...

引用

International Conference on computer Supported Cooperative Work in Design

作者： Ming Jiang Shuo Zhang Feng Zhang Biao Guo Yun Li School of Computer Science and Information Security Guilin University Of Electronic Technology Guilin China Guangxi Metaverse Application Engineering Center Nanning China School of Information Engineering Guangxi Polytechnic of Construction Nanning China Guangxi Institute of Digital Technology Nanning China Guangxi Key Laboratory of Trusted Software Guilin University of Electronic Technology Guilin China

ISBN: (数字)9798350349184

ISBN: (纸本)9798350349191

Image steganography and image steganalysis is one of the research hotspots in the field of information security, the abuse of steganography causes many security risks. As an attack method of image steganography, the main purpose of image steganalysis is to detect whether there is secret information in the image. Recent works have demonstrated the efficacy of CNN in the realm of image steganalysis. However, due to the limitation of convolution kernel size and lack of the ability to model long-range dependencies, CNN-based steganalysis methods face challenges with capturing the global features of steganographic images which are beneficial for image steganalysis. This paper proposes a multi-branch feature fusion network, called MFFNet, which effective extraction and fusion of different features of steganographic images to achieve better detection performance. Firstly, several residual attention blocks (RAB) are incorporated with spatial rich model filters (SRM) in the preprocessing phase to enhance the signal-to-noise ratio (SNR) of steganographic signals. Secondly, in the feature extraction phase, a three-branch parallel fusion module (TFM) is designed to fuse different branches to rich feature representations of steganographic images, where the global branch captures global features by a hardware-friendly (DFC) attention mechanism. Finally, the channel split mechanism is exploited to prevent feature redundancies and decrease the number of parameters. The experiments reflect that the proposed network outperforms several state-of-the-art image steganalysis methods.

关键词： Steganography Filters Fuses Federated learning Redundancy Information security Interference

来源：评论

学校读者我要写书评

暂无评论

New Insights on Relieving Task-Recency Bias for Online Class Incremental Learning

arXiv

引用

arXiv 2023年

作者： Liang, Guoqiang Chen, Zhaojie Chen, Zhaoqiang Ji, Shiyu Zhang, Yanning Northwestern Polytechnical University 710072 China National Engineering Laboratory for Integrated Aero-Space-Ground-Ocean Big Data Application Technology China Key Laboratory of Big Data Storage and Management China School of Computer Science China School of Software Engineering China

—To imitate the ability of keeping learning of human, continual learning which can learn from a never-ending data stream has attracted more interests recently. In all settings, the online class incremental learning (OCIL), where incoming samples from data stream can be used only once, is more challenging and can be encountered more frequently in real world. Actually, all continual learning models face a stability-plasticity dilemma, where the stability means the ability to preserve old knowledge while the plasticity denotes the ability to incorporate new knowledge. Although replay-based methods have shown exceptional promise, most of them concentrate on the strategy for updating and retrieving memory to keep stability at the expense of plasticity. To strike a preferable trade-off between stability and plasticity, we propose an Adaptive Focus Shifting algorithm (AFS), which dynamically adjusts focus to ambiguous samples and non-target logits in model learning. Through a deep analysis of the task-recency bias caused by class imbalance, we propose a revised focal loss to mainly keep stability. By utilizing a new weight function, the revised focal loss will pay more attention to current ambiguous samples, which are the potentially valuable samples to make model progress quickly. To promote plasticity, we introduce a virtual knowledge distillation. By designing a virtual teacher, it assigns more attention to non-target classes, which can surmount overconfidence and encourage model to focus on inter-class information. Extensive experiments on three popular datasets for OCIL have shown the effectiveness of AFS. The code will be available at https://***/czjghost/AFS. Copyright © 2023, The Authors. All rights reserved.

关键词： Distillation

来源：评论

学校读者我要写书评

暂无评论

Match and Retrieval: Legal Similar Case Retrieval via Graph Matching Network

Match and Retrieval: Legal Similar Case Retrieval via Graph ...

引用

IEEE International Conference on Data Mining Workshops (ICDM Workshops)

作者： Shang Gao Yanling Li Fengpei Ge Min Lin Haiqing Yu Sukun Wang Zhongyi Miao College of Computer Science and Technology Key Laboratory of Infinite-Dimensional Hamiltonian System and Its Algorithm Application (Ministry of Education) Inner Mongolia Normal University Hohhot China Library Beijing University of Posts and Telecommunications Beijing China College of Computer Science and Technology Inner Mongolia Normal University Hohhot China

Legal document retrieval is a crucial application in the field of legal Artificial Intelligence (AI). It involves retrieving the most relevant legal cases from a legal document database when inputting a query. Mainstream solutions typically employ pre-trained language models or large language models for retrieval. However, both of these methods rely on black-box models trained on unlabeled data, resulting in a lack of interpretability in the model's retrieval outcomes. Graph Neural Networks (GNNs) possess strong representational and interpretable capabilities. Therefore, representing legal cases using graph-structured data and performing the Legal Similar Case Retrieval (LSCR) task through GNNs is a promising approach. Nevertheless, typical GNNs primarily focus on node information within graph-structured data and overlook edge and inter-graph interaction information, leading to lower retrieval accuracy. To address these issues, in this paper, we propose a Legal case retrieval model utilizing a Graph Matching Network, called Legal-GMN. We evaluate the model on a dataset composed of real legal judgment cases. Experimental results demonstrate that Legal-GMN effectively enhances the retrieval accuracy for the LSCR task. Compared with baseline methods, Legal-GMN improves retrieval precision by approximately 15% and enhances retrieval efficiency by approximately 90%. Remarkably, while maintaining State-Of-The-Art (SOTA) performance, Legal-GMN can also generate visualized graphs of the retrieval process, significantly enhancing its interpretability.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Tensor Graph Convolutional Network for Dynamic Graph Representation Learning

arXiv

引用

arXiv 2024年

作者： Wang, Ling Yuan, Ye The School of Computer Science and Technology Chongqing University of Posts and Telecommunications Chongqing400065 China The Chongqing Key Laboratory of Big Data and Intelligent Computing Chongqing Engineering Research Center of Big Data Application for Smart Cities Chongqing Institute of Green and Intelligent Technology Chinese Academy of Sciences Chongqing400714 China The College of Computer and Information Science Southwest University Chongqing400715 China

Dynamic graphs (DG) describe dynamic interactions between entities in many practical scenarios. Most existing DG representation learning models combine graph convolutional network and sequence neural network, which model spatial-temporal dependencies through two different types of neural networks. However, this hybrid design cannot well capture the spatial-temporal continuity of a DG. In this paper, we propose a tensor graph convolutional network to learn DG representations in one convolution framework based on the tensor product with the following two-fold ideas: a) representing the information of DG by tensor form;b) adopting tensor product to design a tensor graph convolutional network modeling spatial-temporal feature simultaneously. Experiments on real-world DG datasets demonstrate that our model obtains state-of-the-art performance. Copyright © 2024, The Authors. All rights reserved.

关键词： Tensors

来源：评论

学校读者我要写书评

暂无评论

A Novel Binary Firefly Algorithm for the Minimum Labeling Spanning Tree Problem

引用

computer Modeling in Engineering & Sciences 2020年第10期125卷 197-214页

作者： Mugang Lin Fangju Liu Huihuang Zhao Jianzhen Chen College of Computer Science and Technology Hengyang Normal UniversityHengyang421002China Hunan Provincial Key Laboratory of Intelligent Information Processing and Application Hengyang421002China School of Computer Science University of South ChinaHengyang421001China

Given a connected undirected graph G whose edges are labeled,the minimumlabeling spanning tree(MLST)problemis to find a spanning tree of G with the smallest number of different *** is anNP-hard combinatorial optimization problem,which is widely applied in communication networks,multimodal transportation networks,and data *** approximation algorithms and heuristics algorithms have been proposed for the *** algorithm is a new meta-heuristic *** of its simplicity and easy implementation,it has been successfully applied in various ***,the basic firefly algorithm is not suitable for discrete *** this end,a novel discrete firefly algorithm for the MLST problem is proposed in this paper.A binary operation method to update firefly positions and a local feasible handling method are introduced,which correct unfeasible solutions,eliminate redundant labels,and make the algorithm more suitable for discrete *** results show that the algorithm has good *** algorithm can be extended to solve other discrete optimization problems.

关键词： Minimum labeling spanning tree problem binary firefly algorithm meta-heuristics discrete optimization

来源：评论

学校读者我要写书评

暂无评论

An Asymptotically Consistent Shell Model for Compressible Biological Structures with Finite-Strain Deformation

SSRN

引用

SSRN 2024年

作者： Yu, Xiang Chen, Xiaoyi Department of Mathematics School of Computer Science and Technology Dongguan University of Technology Dongguan523808 China Guangdong Provincial Key Laboratory of Interdisciplinary Research and Application for Data Science Department of Applied Mathematics Beijing Normal University-Hong Kong Baptist University BNU-HKBU United International College Zhuhai519087 China

We derive an asymptotically consistent morphoelastic shell model to describe the finite deformations of biological tissues using an energy approach. Biological materials may exhibit remarkable compressibility when under large deformations, and we take this factor into account for accurate predictions of their morphoelastic changes. The morphoelastic shell model combines the growth model of Rodriguez et al. and a novel shell model developed by us. We start from the three-dimensional (3D) morphoelastic model and construct the optimal shell energy based on a series expansion around the middle surface. A two-step variational method is applied that retains the leading-order expansion coefficient while eliminating the higher-order ones. The main outcome is a two-dimensional (2D) shell energy depending on the stretching and bending strains of the middle surface. The derived morphoelastic shell model is asymptotically consistent with three-dimensional morphoelasticity and can recover various shell models in literature. Several examples are shown for the verification and illustration. © 2024, The Authors. All rights reserved.

关键词： Shells (structures)

来源：评论

学校读者我要写书评

暂无评论

Efficiently Computing Geodesic Loop for Interactive Segmentation of a 3D Mesh

引用

Recent Advances in computer Science and Communications 2021年第8期14卷 2477-2488页

作者： Meng, Yun Zhu, Shaojun Liu, Bangquan Sun, Dechao Liu, Li Yang, Weihua College of Digital Technology and Engineering Ningbo University of Finance & Economics Ningbo China Zhejiang Province Key Laboratory of Smart Management & Application of Modern Agricultural Resources Huzhou University Huzhou China College of Electronics and Computer Zhejiang Wanli University Ningbo China First People’s Hospital of Huzhou Huzhou China

Introduction: Shape segmentation is a fundamental problem of computer graphics and geometric modeling. Although the existence of segmentation algorithms of shapes have been widely studied in the mathematics community, little progress has been made on how to compute them on polygonal surfaces interactively using geodesic loops. Methods: We compute the geodesic distance fields with the improved Fast March Method (FMM) proposed by Xin and Wang. A new algorithm is proposed to compute geodesic loops over a triangulated surface as well as a new interactive shape segmentation manner. Results: The average computation time on the 50K vertices model is less than 0.08s. Discussion: In the future, we will use an accurate geodesic algorithm and parallel computing techniques to improve our algorithm to obtain a better smooth geodesic loop. Conclusion: A large number of experimental results show that the algorithm proposed in this paper can effectively achieve high precision geodesic loop paths, and this method can also be used for interactive shape segmentation in real-time. © 2021 Bentham Science Publishers.

关键词： Geodesy

来源：评论

学校读者我要写书评

暂无评论

Characteristic analysis and similarity calculation of molecular compounds of traditional chinese medicine 2020

Characteristic analysis and similarity calculation of molecu...

引用

3rd International Conference on Data Science and Information technology, DSIT 2020

作者： Hong, Juxing Wang, Lu Zhang, Shoude Xie, Fenfen Computer Technology and Application Qinghai University Xining China State Key Laboratory Qinghai University Xining China

ISBN: (纸本)9781450376044

To be able to extract corresponding features of molecular compounds of traditional Chinese medicine, and to use the extracted features to analyze physical and chemical properties and similarities. In this study, Python technology is used to process the TCM molecular library file into an effective TCM molecular compound file, and the chemical drug molecular compound analysis method is used to extract the characteristics between TCM molecular compounds and analyze the physical and chemical properties contained in it. Besides, the four similarity calculation methods such as the Tanimoto coefficient and the Dice coefficient in data mining techniques combined with molecular fingerprint to analyze the similarity and physicochemical properties between traditional Chinese medicine molecular compounds. The experimental results of this study show that the analysis method of chemical drug molecular compounds can better extract the characteristics of traditional Chinese medicine molecular compounds. The combination of data mining techniques and the analysis method to analyze the similarity and characteristics of traditional Chinese medicine molecular compounds can play an important role in the field of drug discovery and biological activity research of traditional Chinese medicine. It can play an important role in the screening of drugs for the clinical treatment of diseases. © 2020 ACM.

关键词： Data mining

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：