Information Sciences最新文献_第8页

Large-scale consensus in incomplete social network with non-cooperative behaviors and dimension reduction 具有非合作行为和降维的不完整社会网络中的大规模共识

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2024-10-17 DOI: 10.1016/j.ins.2024.121563

Wenxiu Ma , Jia Lv , Xiaoli Tian , Ondrej Krejcar , Enrique Herrera-Viedma

Obtaining a consensus solution is a formidable challenge for large-scale decision-making (LSDM) in a social network (SN). The reaching of large-scale consensus in SN is hindered by serious difficulties, including complex decision information, incomplete social relations, a multitude of decision-makers (DMs), and non-cooperative behaviors. This paper introduces a novel three-stage consensus framework that systematically addresses these challenges by data preprocessing, dimension reduction, and optimization modeling. Firstly, the cloud model is applied to convert the probabilistic linguistic information into numerical information, facilitating computational analysis. Meanwhile, an improved t-norm trust propagation method that incorporates the impact of opinion similarity is developed, ensuring the completeness of SN. Secondly, an improved Louvain algorithm is designed to divide large group into cohesive subgroups, enhancing the manageability of LSDM. On this basis, a three-stage consensus optimization that considers non-cooperative behaviors is proposed, which boasts threefold benefits: (i) Assures the synchronous achievement of local and global consensus. (ii) Implements self-adaptive management mechanism of non-cooperative behaviors. (iii) Provides acceptable adjusted opinions for subgroups and DMs. Finally, detailed numerical experiments and comparative analyses are given to demonstrate the effectiveness of the proposed method.

对于社会网络（SN）中的大规模决策（LSDM）而言，获得共识解决方案是一项艰巨的挑战。在社会网络中达成大规模共识面临着严重的困难，包括复杂的决策信息、不完整的社会关系、众多的决策者（DMs）以及非合作行为。本文介绍了一种新颖的三阶段共识框架，通过数据预处理、降维和优化建模系统地解决了这些难题。首先，应用云模型将概率语言信息转换为数字信息，从而方便计算分析。同时，还开发了一种改进的 t-norm 信任传播方法，该方法结合了意见相似性的影响，确保了 SN 的完整性。其次，设计了一种改进的卢万算法，将大型群组划分为具有凝聚力的子群组，增强了 LSDM 的可管理性。在此基础上，提出了一种考虑非合作行为的三阶段共识优化方法，它具有三方面的优点(i) 确保同步达成局部和全局共识。(ii) 实现非合作行为的自适应管理机制。(iii) 为分组和 DM 提供可接受的调整意见。最后，我们给出了详细的数值实验和比较分析，以证明所提方法的有效性。

{"title":"Large-scale consensus in incomplete social network with non-cooperative behaviors and dimension reduction","authors":"Wenxiu Ma , Jia Lv , Xiaoli Tian , Ondrej Krejcar , Enrique Herrera-Viedma","doi":"10.1016/j.ins.2024.121563","DOIUrl":"10.1016/j.ins.2024.121563","url":null,"abstract":"<div><div>Obtaining a consensus solution is a formidable challenge for large-scale decision-making (LSDM) in a social network (SN). The reaching of large-scale consensus in SN is hindered by serious difficulties, including complex decision information, incomplete social relations, a multitude of decision-makers (DMs), and non-cooperative behaviors. This paper introduces a novel three-stage consensus framework that systematically addresses these challenges by data preprocessing, dimension reduction, and optimization modeling. Firstly, the cloud model is applied to convert the probabilistic linguistic information into numerical information, facilitating computational analysis. Meanwhile, an improved t-norm trust propagation method that incorporates the impact of opinion similarity is developed, ensuring the completeness of SN. Secondly, an improved Louvain algorithm is designed to divide large group into cohesive subgroups, enhancing the manageability of LSDM. On this basis, a three-stage consensus optimization that considers non-cooperative behaviors is proposed, which boasts threefold benefits: (i) Assures the synchronous achievement of local and global consensus. (ii) Implements self-adaptive management mechanism of non-cooperative behaviors. (iii) Provides acceptable adjusted opinions for subgroups and DMs. Finally, detailed numerical experiments and comparative analyses are given to demonstrate the effectiveness of the proposed method.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"690 ","pages":"Article 121563"},"PeriodicalIF":8.1,"publicationDate":"2024-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142530745","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Hierarchical degradation-aware network for full-reference image quality assessment 用于全参考图像质量评估的分层降级感知网络

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2024-10-17 DOI: 10.1016/j.ins.2024.121557

Xuting Lan , Fan Jia , Xu Zhuang , Xuekai Wei , Jun Luo , Mingliang Zhou , Sam Kwong

Full-Reference Image Quality Assessment (FR-IQA) algorithms excel in evaluating perceptual distortions by comparing reference and distorted images. However, as the severity and quantity of distortions in datasets increase, existing FR-IQA methods struggle to capture complex nonlinear perceptual features. This limitation results in reduced adaptability and inaccurate assessments for images with more severe or multiple distortions. Recognizing the importance of understanding image degradation mechanisms, we propose a novel hierarchical degradation-aware network (HDaN) method. First, by exploring the degradation mechanisms from the reference image to the distorted image, our degradation network matches distortions that align more closely with the human visual system (HVS). Next, we design a convertor to project the matched features into multiple spaces, creating multidimensional feature representations that more comprehensively capture the complexity of image distortions rather than being confined to a single feature space. Then, we calculate a similarity matrix between the distorted and mapped features, selecting the most similar (top-k) features for merging. Finally, a regression network maps the merged features to quality scores, providing the final quality prediction. The experimental results demonstrate that our proposed HDaN method outperforms traditional deep learning-based FR-IQA methods. Specifically, the HDaN shows higher PLCC and SROCC metrics on benchmark datasets, significantly improving over existing methods. Moreover, the method exhibits better adaptability to images with varying degrees and types of distortions, thereby greatly enhancing the overall performance of IQA.

全参照图像质量评估（FR-IQA）算法通过比较参照图像和失真图像来评估感知失真，效果非常出色。然而，随着数据集中失真的严重程度和数量的增加，现有的全参照图像质量评估方法很难捕捉到复杂的非线性感知特征。这种局限性导致适应性降低，对具有更严重或多重失真的图像的评估不准确。认识到了解图像降解机制的重要性，我们提出了一种新颖的分层降解感知网络（HDaN）方法。首先，通过探索从参考图像到失真图像的降级机制，我们的降级网络可以匹配更接近人类视觉系统（HVS）的失真图像。接下来，我们设计了一个转换器，将匹配的特征投射到多个空间，创建多维特征表示，从而更全面地捕捉图像失真的复杂性，而不是局限于单一的特征空间。然后，我们计算失真特征和映射特征之间的相似性矩阵，选择最相似（前 k 个）的特征进行合并。最后，回归网络将合并后的特征映射到质量分数上，提供最终的质量预测。实验结果表明，我们提出的 HDaN 方法优于传统的基于深度学习的 FR-IQA 方法。具体来说，HDaN 在基准数据集上显示出更高的 PLCC 和 SROCC 指标，明显优于现有方法。此外，该方法对不同失真程度和类型的图像具有更好的适应性，从而大大提高了 IQA 的整体性能。

{"title":"Hierarchical degradation-aware network for full-reference image quality assessment","authors":"Xuting Lan , Fan Jia , Xu Zhuang , Xuekai Wei , Jun Luo , Mingliang Zhou , Sam Kwong","doi":"10.1016/j.ins.2024.121557","DOIUrl":"10.1016/j.ins.2024.121557","url":null,"abstract":"<div><div>Full-Reference Image Quality Assessment (FR-IQA) algorithms excel in evaluating perceptual distortions by comparing reference and distorted images. However, as the severity and quantity of distortions in datasets increase, existing FR-IQA methods struggle to capture complex nonlinear perceptual features. This limitation results in reduced adaptability and inaccurate assessments for images with more severe or multiple distortions. Recognizing the importance of understanding image degradation mechanisms, we propose a novel hierarchical degradation-aware network (HDaN) method. First, by exploring the degradation mechanisms from the reference image to the distorted image, our degradation network matches distortions that align more closely with the human visual system (HVS). Next, we design a convertor to project the matched features into multiple spaces, creating multidimensional feature representations that more comprehensively capture the complexity of image distortions rather than being confined to a single feature space. Then, we calculate a similarity matrix between the distorted and mapped features, selecting the most similar (top-k) features for merging. Finally, a regression network maps the merged features to quality scores, providing the final quality prediction. The experimental results demonstrate that our proposed HDaN method outperforms traditional deep learning-based FR-IQA methods. Specifically, the HDaN shows higher PLCC and SROCC metrics on benchmark datasets, significantly improving over existing methods. Moreover, the method exhibits better adaptability to images with varying degrees and types of distortions, thereby greatly enhancing the overall performance of IQA.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"690 ","pages":"Article 121557"},"PeriodicalIF":8.1,"publicationDate":"2024-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142530651","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Graph-based stock prediction with multisource information and relational data fusion 利用多源信息和关系数据融合进行基于图表的股票预测

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2024-10-17 DOI: 10.1016/j.ins.2024.121561

Qiuyue Zhang , Yunfeng Zhang , Fangxun Bao , Yang Ning , Caiming Zhang , Peide Liu

With the application of multisource information in different fields, the combination of different types of information, such as numerical data and text information, has become a favourable choice for performing stock market analyses. Despite the rich information provided by multisource data, building structured relationships remains challenging. In addition, some market relationship-based analysis methods use a predefined graph structure as a stock relationship graph, which makes it impossible to sensitively aggregate attribute features, and these methods cannot dynamically update market relationships or relationship strengths. In this paper, we propose a novel dynamic attribute-driven graph attention network incorporating sentiment (AGATS) information, transaction data, and text data. Inspired by behavioural finance, we separately extract sentiment information as a factor of technical indicators, and further realize the early fusion of technical indicators and textual data through tensor fusion. In particular, real-time intramarket dependencies and key attribute information are captured with graph networks, enabling dynamic relationship and relationship strength updates. Experiments conducted on real datasets show that our model is capable of ourperforming previously developed methods in prediction and trading.

随着多源信息在不同领域的应用，数字数据和文本信息等不同类型信息的组合已成为进行股市分析的有利选择。尽管多源数据提供了丰富的信息，但建立结构化关系仍具有挑战性。此外，一些基于市场关系的分析方法使用预定义的图结构作为股票关系图，无法灵敏地聚合属性特征，而且这些方法无法动态更新市场关系或关系强度。在本文中，我们提出了一种新颖的动态属性驱动图注意力网络，其中包含情感（AGATS）信息、交易数据和文本数据。受行为金融学的启发，我们将情绪信息作为技术指标的一个因子单独提取出来，并通过张量融合进一步实现了技术指标和文本数据的早期融合。特别是通过图网络实时捕捉市场内的依赖关系和关键属性信息，实现动态关系和关系强度更新。在真实数据集上进行的实验表明，我们的模型能够在预测和交易方面优于之前开发的方法。

{"title":"Graph-based stock prediction with multisource information and relational data fusion","authors":"Qiuyue Zhang , Yunfeng Zhang , Fangxun Bao , Yang Ning , Caiming Zhang , Peide Liu","doi":"10.1016/j.ins.2024.121561","DOIUrl":"10.1016/j.ins.2024.121561","url":null,"abstract":"<div><div>With the application of multisource information in different fields, the combination of different types of information, such as numerical data and text information, has become a favourable choice for performing stock market analyses. Despite the rich information provided by multisource data, building structured relationships remains challenging. In addition, some market relationship-based analysis methods use a predefined graph structure as a stock relationship graph, which makes it impossible to sensitively aggregate attribute features, and these methods cannot dynamically update market relationships or relationship strengths. In this paper, we propose a novel dynamic attribute-driven graph attention network incorporating sentiment (AGATS) information, transaction data, and text data. Inspired by behavioural finance, we separately extract sentiment information as a factor of technical indicators, and further realize the early fusion of technical indicators and textual data through tensor fusion. In particular, real-time intramarket dependencies and key attribute information are captured with graph networks, enabling dynamic relationship and relationship strength updates. Experiments conducted on real datasets show that our model is capable of ourperforming previously developed methods in prediction and trading.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"690 ","pages":"Article 121561"},"PeriodicalIF":8.1,"publicationDate":"2024-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142445333","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Learnable self-supervised support vector machine based individual selection strategy for multimodal multi-objective optimization 基于可学习自监督支持向量机的多模态多目标优化个体选择策略

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2024-10-17 DOI: 10.1016/j.ins.2024.121553

Xiaochuan Gao , Weiting Bai , Qianlong Dang , Shuai Yang , Guanghui Zhang

Multimodal multi-objective optimization problem (MMOP) is a frontier research problem, which can provide decision makers with more choices without making trade-offs. Many multimodal multi-objective evolutionary algorithms (MMOEAs) have been proposed to solve MMOP. However, most MMOEAs tend to prioritize the objective dominance of individuals in the process of individual selection, and only individuals with the same objective dominance will be considered the diversity, which leads to the loss of many promising solutions. To solve the above problem, this paper proposes a learnable self-supervised support vector machine (SVM) based multimodal multi-objective optimization algorithm (SVMEA). Support vector machine can learn the knowledge about distinguishing the advantages and disadvantages of individuals from the data in the existing training set and select individuals, in which the objective dominance of individuals is as important as diversity. Moreover, a crowding distance calculation method based on Manhattan distance is designed. Compared with the traditional method using Euclidean distance to calculate crowding distance, it can better evaluate the diversity of individuals in the decision space and assist the selection of elite solutions. Experimental results show that the proposed SVMEA is competitive with seven other advanced MMOEAs on 34 benchmark problems and a practical application problem.

多模态多目标优化问题（MMOP）是一个前沿研究问题，它能为决策者提供更多选择，而无需做出权衡。许多多模态多目标进化算法（MMOEAs）被提出来解决多模态多目标优化问题。然而，大多数多目标进化算法在个体选择过程中倾向于优先考虑个体的目标优势，只有目标优势相同的个体才会被认为是多样性的，这就导致了许多有希望的解决方案的丢失。为了解决上述问题，本文提出了一种基于可学习自监督支持向量机（SVM）的多模态多目标优化算法（SVMEA）。支持向量机可以从现有训练集中的数据中学习区分个体优劣的知识，并选择个体，其中个体的目标主导性与多样性同等重要。此外，还设计了一种基于曼哈顿距离的拥挤距离计算方法。与使用欧氏距离计算拥挤距离的传统方法相比，它能更好地评估决策空间中个体的多样性，并帮助选择精英解。实验结果表明，在 34 个基准问题和一个实际应用问题上，所提出的 SVMEA 与其他七种先进的 MMOEA 相比具有很强的竞争力。

{"title":"Learnable self-supervised support vector machine based individual selection strategy for multimodal multi-objective optimization","authors":"Xiaochuan Gao , Weiting Bai , Qianlong Dang , Shuai Yang , Guanghui Zhang","doi":"10.1016/j.ins.2024.121553","DOIUrl":"10.1016/j.ins.2024.121553","url":null,"abstract":"<div><div>Multimodal multi-objective optimization problem (MMOP) is a frontier research problem, which can provide decision makers with more choices without making trade-offs. Many multimodal multi-objective evolutionary algorithms (MMOEAs) have been proposed to solve MMOP. However, most MMOEAs tend to prioritize the objective dominance of individuals in the process of individual selection, and only individuals with the same objective dominance will be considered the diversity, which leads to the loss of many promising solutions. To solve the above problem, this paper proposes a learnable self-supervised support vector machine (SVM) based multimodal multi-objective optimization algorithm (SVMEA). Support vector machine can learn the knowledge about distinguishing the advantages and disadvantages of individuals from the data in the existing training set and select individuals, in which the objective dominance of individuals is as important as diversity. Moreover, a crowding distance calculation method based on Manhattan distance is designed. Compared with the traditional method using Euclidean distance to calculate crowding distance, it can better evaluate the diversity of individuals in the decision space and assist the selection of elite solutions. Experimental results show that the proposed SVMEA is competitive with seven other advanced MMOEAs on 34 benchmark problems and a practical application problem.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"690 ","pages":"Article 121553"},"PeriodicalIF":8.1,"publicationDate":"2024-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142445327","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A joint vehicular device scheduling and uncertain resource management scheme for Federated Learning in Internet of Vehicles 面向车联网联合学习的联合车辆设备调度和不确定资源管理方案

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2024-10-17 DOI: 10.1016/j.ins.2024.121552

Jianghui Cai , Bujia Chen , Jie Wen , Zhihua Cui , Jinjun Chen , Wensheng Zhang

Federated learning (FL) offers an effective framework for the efficient process in vehicular edge computing. However, FL encompasses the process of distributing and uploading model parameters, which are inevitably transmitted in a wireless network environment. Some challenges in FL-assisted Internet of Vehicles (IoV) sceneries gradually emerging, such as data heterogeneity, concerned device resources, and unstable communication environment, which necessitate intelligent vehicle selection schemes that accelerate training efficiency. Based on these, we consider a new scenario, specifically an FL-assisted IoV system under uncertain communication conditions, and develop an interval many-objective vehicle selection and bandwidth allocation (IMoVSBA) joint optimization scheme. This scheme takes into account computation latency, energy consumption, server utilization, and data quality, while meeting multi-criteria resource optimization requirements. Among these, server utilization is a new objective designed specifically for this joint optimization problem. For the proposed problem, a novel interval many-objective evolutionary algorithm with individual comprehensive indicator to control the evolution direction (IMaOEACI) is designed. Simulation results demonstrate that this method outperforms other schemes in terms of accuracy, training cost, and server utilization, effectively improving training efficiency in wireless channel environments and reasonably utilizing bandwidth resources. It provides significant scientific value and application potential in the field of the IoVs.

联合学习（FL）为车载边缘计算的高效流程提供了一个有效的框架。然而，FL 包括分发和上传模型参数的过程，而这些参数不可避免地要在无线网络环境中传输。FL辅助车联网（IoV）场景中的一些挑战逐渐显现，如数据异构性、相关设备资源和不稳定的通信环境等，这就需要能加快训练效率的智能车辆选择方案。在此基础上，我们考虑了一种新的场景，特别是在不确定通信条件下的 FL 辅助 IoV 系统，并开发了一种区间多目标车辆选择和带宽分配（IMoVSBA）联合优化方案。该方案兼顾了计算延迟、能耗、服务器利用率和数据质量，同时满足多标准资源优化要求。其中，服务器利用率是专门为联合优化问题设计的新目标。针对提出的问题，设计了一种新颖的区间多目标进化算法，用个体综合指标来控制进化方向（IMaOEACI）。仿真结果表明，该方法在精度、训练成本和服务器利用率等方面均优于其他方案，有效提高了无线信道环境下的训练效率，合理利用了带宽资源。它在物联网领域具有重要的科学价值和应用潜力。

{"title":"A joint vehicular device scheduling and uncertain resource management scheme for Federated Learning in Internet of Vehicles","authors":"Jianghui Cai , Bujia Chen , Jie Wen , Zhihua Cui , Jinjun Chen , Wensheng Zhang","doi":"10.1016/j.ins.2024.121552","DOIUrl":"10.1016/j.ins.2024.121552","url":null,"abstract":"<div><div>Federated learning (FL) offers an effective framework for the efficient process in vehicular edge computing. However, FL encompasses the process of distributing and uploading model parameters, which are inevitably transmitted in a wireless network environment. Some challenges in FL-assisted Internet of Vehicles (IoV) sceneries gradually emerging, such as data heterogeneity, concerned device resources, and unstable communication environment, which necessitate intelligent vehicle selection schemes that accelerate training efficiency. Based on these, we consider a new scenario, specifically an FL-assisted IoV system under uncertain communication conditions, and develop an interval many-objective vehicle selection and bandwidth allocation (IMoVSBA) joint optimization scheme. This scheme takes into account computation latency, energy consumption, server utilization, and data quality, while meeting multi-criteria resource optimization requirements. Among these, server utilization is a new objective designed specifically for this joint optimization problem. For the proposed problem, a novel interval many-objective evolutionary algorithm with individual comprehensive indicator to control the evolution direction (IMaOEACI) is designed. Simulation results demonstrate that this method outperforms other schemes in terms of accuracy, training cost, and server utilization, effectively improving training efficiency in wireless channel environments and reasonably utilizing bandwidth resources. It provides significant scientific value and application potential in the field of the IoVs.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"690 ","pages":"Article 121552"},"PeriodicalIF":8.1,"publicationDate":"2024-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142530609","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Defending against backdoor attack on deep neural networks based on multi-scale inactivation 基于多尺度失活防御深度神经网络后门攻击

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2024-10-17 DOI: 10.1016/j.ins.2024.121562

Anqing Zhang , Honglong Chen , Xiaomeng Wang , Junjian Li , Yudong Gao , Xingang Wang

Deep neural networks (DNNs) have excellent performance in various applications, especially for image classification tasks. However, DNNs also face the threat of backdoor attacks. Backdoor attacks embed a hidden backdoor into a model, after which the infected model can achieve correct classification on benign images, while incorrectly classify the images with the backdoor triggers as the target label. To obtain a clean model from a backdoor dataset, we propose a Kalman filtering based multi-scale inactivation scheme, which can effectively remove poison data in a poison dataset and obtain a clean model. Every sample in the suspicious training dataset will be judged by multi-scale inactivation and obtain a series of judging results, then data fusion is conducted using kalman filtering to determine whether it is a poison sample. To further improve the performance, a trigger localization and target determination based scheme is proposed. Extensive experiments are conducted to demonstrate the superior effectiveness of the proposed method. The results show that the proposed methods can remove poison samples effectively, and achieve greater than 99% recall rate, and the attack success rate of the retrained clean model is smaller than 1%.

深度神经网络（DNN）在各种应用中表现出色，尤其是在图像分类任务中。然而，深度神经网络也面临着后门攻击的威胁。后门攻击会在模型中嵌入隐藏的后门，受感染的模型可以对良性图像进行正确分类，而对以后门触发器为目标标签的图像进行错误分类。为了从后门数据集中获得干净的模型，我们提出了一种基于卡尔曼滤波的多尺度失活方案，它能有效去除有毒数据集中的有毒数据，获得干净的模型。对可疑训练数据集中的每个样本进行多尺度失活判断，得到一系列判断结果，然后利用卡尔曼滤波进行数据融合，判断其是否为中毒样本。为了进一步提高性能，提出了一种基于触发定位和目标判定的方案。为了证明所提方法的优越性能，我们进行了广泛的实验。实验结果表明，所提出的方法可以有效地去除有毒样本，召回率大于 99%，重新训练的干净模型的攻击成功率小于 1%。

{"title":"Defending against backdoor attack on deep neural networks based on multi-scale inactivation","authors":"Anqing Zhang , Honglong Chen , Xiaomeng Wang , Junjian Li , Yudong Gao , Xingang Wang","doi":"10.1016/j.ins.2024.121562","DOIUrl":"10.1016/j.ins.2024.121562","url":null,"abstract":"<div><div>Deep neural networks (DNNs) have excellent performance in various applications, especially for image classification tasks. However, DNNs also face the threat of backdoor attacks. Backdoor attacks embed a hidden backdoor into a model, after which the infected model can achieve correct classification on benign images, while incorrectly classify the images with the backdoor triggers as the target label. To obtain a clean model from a backdoor dataset, we propose a Kalman filtering based multi-scale inactivation scheme, which can effectively remove poison data in a poison dataset and obtain a clean model. Every sample in the suspicious training dataset will be judged by multi-scale inactivation and obtain a series of judging results, then data fusion is conducted using kalman filtering to determine whether it is a poison sample. To further improve the performance, a trigger localization and target determination based scheme is proposed. Extensive experiments are conducted to demonstrate the superior effectiveness of the proposed method. The results show that the proposed methods can remove poison samples effectively, and achieve greater than 99% recall rate, and the attack success rate of the retrained clean model is smaller than 1%.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"690 ","pages":"Article 121562"},"PeriodicalIF":8.1,"publicationDate":"2024-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142445330","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Obfuscation mechanism for simultaneous public event information release and private event information hiding in discrete event systems 离散事件系统中同时发布公共事件信息和隐藏私人事件信息的混淆机制

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2024-10-17 DOI: 10.1016/j.ins.2024.121554

Wei Duan , Christoforos N. Hadjicostis , Zhiwu Li

In this paper, we study the problem of simultaneous public event information release and private event information hiding in discrete event systems modeled by partially observed non-deterministic finite automata, where the public and private event information is respectively defined as unobservable fault and secret events. The notion of D-C-compossibility is introduced to characterize whether a system simultaneously conceals the occurrences of secret events while releasing the occurrences of fault events. When such a property does not hold, we investigate its enforcement through a defensive function. This function takes each observable event of a system as input and generates a suitably modified event sequence at the system's output using event deletion, insertion, or substitution. Specifically, our focus is on prioritizing the concealment of the secret events while maximizing the detection of the fault events through the use of defensive functions. The notion of D-C-enforceability that refers to the capability of a defensive function to employ a strategy for manipulating observations of a system to enforce D-C-compossibility is proposed. That is, a D-C-enforcing defensive function ensures that all occurrences of fault events can be revealed while concealing all occurrences of secret events. A D-C-diagnoser construction is proposed to enumerate all feasible defensive actions following system behavior. By taking advantage of the D-C-diagnoser, we obtain a necessary and sufficient condition for D-C-enforceability, along with a corresponding obfuscation strategy for defensive functions.

本文研究了部分观测非确定有限自动机建模的离散事件系统中同时发布公共事件信息和隐藏私人事件信息的问题，其中公共事件信息和私人事件信息分别定义为不可观测的故障事件和秘密事件。我们引入了 D-C-compossibility 概念来描述一个系统是否同时隐藏秘密事件的发生和释放故障事件的发生。当这种特性不成立时，我们通过防御函数来研究其执行情况。该函数将系统的每个可观测事件作为输入，并通过删除、插入或替换事件，在系统输出端生成一个经过适当修改的事件序列。具体来说，我们的重点是通过使用防御函数，在最大限度地检测故障事件的同时，优先隐藏秘密事件。我们提出了 D-C-enforceability 概念，指的是防御函数采用策略操纵系统观测结果以执行 D-C-compossibility 的能力。也就是说，一个 D-C 强制防御函数能确保所有故障事件的发生都能被揭示，同时隐藏所有秘密事件的发生。本文提出了一种 D-C 诊断器构造，用于根据系统行为枚举所有可行的防御行动。利用 D-C 诊断器，我们获得了 D-C 可执行性的必要条件和充分条件，以及相应的防御函数混淆策略。

{"title":"Obfuscation mechanism for simultaneous public event information release and private event information hiding in discrete event systems","authors":"Wei Duan , Christoforos N. Hadjicostis , Zhiwu Li","doi":"10.1016/j.ins.2024.121554","DOIUrl":"10.1016/j.ins.2024.121554","url":null,"abstract":"<div><div>In this paper, we study the problem of simultaneous public event information release and private event information hiding in discrete event systems modeled by partially observed non-deterministic finite automata, where the public and private event information is respectively defined as unobservable fault and secret events. The notion of <em>D-C-compossibility</em> is introduced to characterize whether a system simultaneously conceals the occurrences of secret events while releasing the occurrences of fault events. When such a property does not hold, we investigate its enforcement through a defensive function. This function takes each observable event of a system as input and generates a suitably modified event sequence at the system's output using event deletion, insertion, or substitution. Specifically, our focus is on prioritizing the concealment of the secret events while maximizing the detection of the fault events through the use of defensive functions. The notion of <em>D-C-enforceability</em> that refers to the capability of a defensive function to employ a strategy for manipulating observations of a system to enforce <em>D</em>-<em>C</em>-compossibility is proposed. That is, a <em>D</em>-<em>C</em>-enforcing defensive function ensures that all occurrences of fault events can be revealed while concealing all occurrences of secret events. A <em>D</em>-<em>C</em>-diagnoser construction is proposed to enumerate all feasible defensive actions following system behavior. By taking advantage of the <em>D</em>-<em>C</em>-diagnoser, we obtain a necessary and sufficient condition for <em>D</em>-<em>C</em>-enforceability, along with a corresponding obfuscation strategy for defensive functions.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"690 ","pages":"Article 121554"},"PeriodicalIF":8.1,"publicationDate":"2024-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142530820","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Minimal solutions of fuzzy relation equations via maximal independent elements 通过最大独立元素求模糊关系方程的最小解

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2024-10-17 DOI: 10.1016/j.ins.2024.121558

David Lobo , Jesús Medina , Timo Camillo Merkl , Reinhard Pichler

Fuzzy relation equations (FRE) are a useful formalism with a broad number of applications in different computer science areas. Testing if a solution exists and, if so, computing the unique greatest solution is straightforward. In contrast, the computation of minimal solutions is more complex. In particular, even in FRE with a very simple structure, the number of minimal solutions can increase exponentially. However, minimal solutions are immensely useful since, under mild conditions, they (together with the greatest solution) allow one to describe the entire space of solutions to an FRE. The main result of this work is a new method for enumerating the set of minimal solutions. It works by establishing a relationship between coverings of FRE and maximal independent elements of (hyper-)boxes. We can thus make efficient enumeration methods for maximal independent elements of (hyper-)boxes applicable also to our setting of FRE, where the operator considered in the composition of fuzzy relations only needs to preserve suprema of arbitrary subsets and infima of non-empty subsets. More specifically, we thus show that the enumeration of the minimal solutions of an FRE can be done with incremental quasi-polynomial delay.

模糊关系方程（FRE）是一种有用的形式主义，在不同的计算机科学领域有着广泛的应用。测试是否存在解，以及如果存在，计算唯一的最大解都很简单。相比之下，最小解的计算则更为复杂。特别是，即使在结构非常简单的 FRE 中，极小解的数量也会呈指数级增长。然而，极小解非常有用，因为在温和的条件下，极小解（连同最大解）可以让我们描述 FRE 的整个解空间。这项研究的主要成果是一种枚举最小解集合的新方法。它通过建立 FRE 的覆盖和（超）盒的最大独立元素之间的关系来实现。因此，我们可以使（超）盒的最大独立元素的高效枚举方法也适用于我们的 FRE 设置，在 FRE 设置中，模糊关系组成中考虑的算子只需要保留任意子集的上界和非空子集的下界。更具体地说，我们因此证明，枚举 FRE 的最小解可以用增量准多项式延迟来完成。

{"title":"Minimal solutions of fuzzy relation equations via maximal independent elements","authors":"David Lobo , Jesús Medina , Timo Camillo Merkl , Reinhard Pichler","doi":"10.1016/j.ins.2024.121558","DOIUrl":"10.1016/j.ins.2024.121558","url":null,"abstract":"<div><div>Fuzzy relation equations (FRE) are a useful formalism with a broad number of applications in different computer science areas. Testing if a solution exists and, if so, computing the unique greatest solution is straightforward. In contrast, the computation of minimal solutions is more complex. In particular, even in FRE with a very simple structure, the number of minimal solutions can increase exponentially. However, minimal solutions are immensely useful since, under mild conditions, they (together with the greatest solution) allow one to describe the entire space of solutions to an FRE. The main result of this work is a new method for enumerating the set of minimal solutions. It works by establishing a relationship between coverings of FRE and maximal independent elements of (hyper-)boxes. We can thus make efficient enumeration methods for maximal independent elements of (hyper-)boxes applicable also to our setting of FRE, where the operator considered in the composition of fuzzy relations only needs to preserve suprema of arbitrary subsets and infima of non-empty subsets. More specifically, we thus show that the enumeration of the minimal solutions of an FRE can be done with incremental quasi-polynomial delay.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"690 ","pages":"Article 121558"},"PeriodicalIF":8.1,"publicationDate":"2024-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142530608","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A distance-based network activity correlation framework for defeating anonymization overlays 基于距离的网络活动相关性框架，用于击败匿名重叠

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2024-10-17 DOI: 10.1016/j.ins.2024.121559

Ugo Fiore, Francesco Palmieri

As the effectiveness of modern Internet-based anonymization infrastructures grows, law enforcement agencies are experiencing a progressive erosion of their surveillance capabilities. This can severely undermine their efforts to prevent and investigate various types of unlawful activities, potentially increasing the impunity of organized criminal networks. Balancing the legitimate privacy needs of individuals with the imperative to maintain public safety and combat criminal behavior in the digital world remains a complex tradeoff for both policymakers and technologists who need to find a systematic and reliable way to link the traffic traces associated with criminal activities to their anonymized origins. Accordingly, this paper presents a simple but very effective de-anonymization approach capable of associating traffic traces captured at the edge of the overlay infrastructures, in correspondence with the true origins, to those captured in correspondence with the destinations. The approach is based on determining the minimum-distance pairs within a complete bipartite graph in which the traffic traces are the nodes. Experiments with different distance functions, applied in varied ways, show that the resulting framework appears to be a promising solution that is scalable and easily deployable on real-life network equipment.

随着基于互联网的现代匿名基础设施的有效性不断提高，执法机构的监控能力正在逐步削弱。这可能会严重削弱他们预防和调查各类非法活动的努力，有可能使有组织犯罪网络更加逍遥法外。平衡个人的合法隐私需求与维护公共安全和打击数字世界犯罪行为的必要性，对于政策制定者和技术专家来说仍然是一个复杂的权衡问题，他们需要找到一种系统可靠的方法，将与犯罪活动相关的流量痕迹与其匿名来源联系起来。因此，本文提出了一种简单但非常有效的去匿名化方法，能够将在重叠基础设施边缘捕获的与真实来源相对应的流量轨迹与捕获的与目的地相对应的流量轨迹联系起来。该方法的基础是确定一个完整的双向图中的最小距离对，其中的流量轨迹是节点。以不同方式应用不同距离函数的实验表明，由此产生的框架似乎是一种很有前途的解决方案，可在现实生活中的网络设备上进行扩展和轻松部署。

{"title":"A distance-based network activity correlation framework for defeating anonymization overlays","authors":"Ugo Fiore, Francesco Palmieri","doi":"10.1016/j.ins.2024.121559","DOIUrl":"10.1016/j.ins.2024.121559","url":null,"abstract":"<div><div>As the effectiveness of modern Internet-based anonymization infrastructures grows, law enforcement agencies are experiencing a progressive erosion of their surveillance capabilities. This can severely undermine their efforts to prevent and investigate various types of unlawful activities, potentially increasing the impunity of organized criminal networks. Balancing the legitimate privacy needs of individuals with the imperative to maintain public safety and combat criminal behavior in the digital world remains a complex tradeoff for both policymakers and technologists who need to find a systematic and reliable way to link the traffic traces associated with criminal activities to their anonymized origins. Accordingly, this paper presents a simple but very effective de-anonymization approach capable of associating traffic traces captured at the edge of the overlay infrastructures, in correspondence with the true origins, to those captured in correspondence with the destinations. The approach is based on determining the minimum-distance pairs within a complete bipartite graph in which the traffic traces are the nodes. Experiments with different distance functions, applied in varied ways, show that the resulting framework appears to be a promising solution that is scalable and easily deployable on real-life network equipment.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"690 ","pages":"Article 121559"},"PeriodicalIF":8.1,"publicationDate":"2024-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142445328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

EsmamDS: A more diverse exceptional survival model mining approach EsmamDS：更多样化的特殊生存模型挖掘方法

IF 8.1 1区计算机科学 0 COMPUTER SCIENCE, INFORMATION SYSTEMS

Information Sciences

Pub Date : 2024-10-16 DOI: 10.1016/j.ins.2024.121549

Renato Vimieiro , Juliana Barcellos Mattos , Paulo S.G. de Mattos Neto

In this work we present an Ant Colony Optimization heuristic to find subgroups with exceptional behavior in time-to-event data. The area of time-to-event or survival data analysis has its basis in statistics, where the main goal is to predict if and when an event will happen. In other words, the main goal in survival analysis has long been to build global models able to predict the time for the occurrence of an event. Nevertheless, very often predictive models are used to compare stratified data in order to evaluate whether a variable is associated or not with the outcome. For instance, patients might be stratified according to a treatment variable (placebo or not) to compare models (survival curves) and decide on the effectiveness of the treatment. Although this is an effective approach if the variable of interest is already known, it does not provide an alternative for the cases where specialists do not know how to stratify the data, that is, if they do not know which variable could be related to the outcome. Our approach targets exactly this. Our method seeks combinations of variables that are associated, i.e. describe, subgroups of individuals with unexpected or exceptional survival curves. In this sense, we complement the literature with a descriptive approach that is able to find and characterize those groups for specialists. Our method is based on the framework of exceptional model mining. It improves on a preliminary version presented in a conference. The main enhancement was to redesign our heuristic to retrieve interesting and diverse subgroups while minimizing three aspects of redundancy: coverage; description; and model. Our second extension regards how the quality function is applied. We now allow users to control whether the quality measure compares subgroups against the population, or against individuals that do not satisfy the descriptive rule. Third, we conduct further experiments to compare the performance of our approach to state of the art algorithms with real world benchmark data sets. Finally, we also present a case study showing a possible application of our method in the bioinformatics/health domain.

在这项工作中，我们提出了一种蚁群优化启发式方法，用于在时间到事件数据中寻找具有特殊行为的子群。时间到事件或生存数据分析领域的基础是统计学，其主要目标是预测事件是否发生以及何时发生。换句话说，长期以来，生存分析的主要目标是建立能够预测事件发生时间的全局模型。然而，预测模型通常用于比较分层数据，以评估变量是否与结果相关。例如，可以根据治疗变量（安慰剂或非安慰剂）对患者进行分层，以比较模型（生存曲线）并决定治疗的有效性。虽然在相关变量已知的情况下，这是一种有效的方法，但在专家不知道如何对数据进行分层的情况下，也就是在专家不知道哪个变量可能与结果相关的情况下，这种方法并不能提供替代方案。我们的方法正是针对这种情况。我们的方法寻求与之相关的变量组合，即描述具有意外或特殊生存曲线的个体亚群。从这个意义上说，我们用一种描述性方法对文献进行了补充，这种方法能够为专家找到并描述这些群体。我们的方法基于特殊模型挖掘框架。它改进了在一次会议上提出的初步版本。主要的改进是重新设计了我们的启发式，以检索有趣且多样化的子群，同时最大限度地减少冗余的三个方面：覆盖范围、描述和模型。我们的第二个扩展涉及如何应用质量函数。现在，我们允许用户控制质量度量是将子群与总体进行比较，还是与不符合描述规则的个体进行比较。第三，我们进行了进一步的实验，用现实世界的基准数据集比较了我们的方法与最先进算法的性能。最后，我们还介绍了一个案例研究，展示了我们的方法在生物信息学/健康领域的可能应用。

{"title":"EsmamDS: A more diverse exceptional survival model mining approach","authors":"Renato Vimieiro , Juliana Barcellos Mattos , Paulo S.G. de Mattos Neto","doi":"10.1016/j.ins.2024.121549","DOIUrl":"10.1016/j.ins.2024.121549","url":null,"abstract":"<div><div>In this work we present an Ant Colony Optimization heuristic to find subgroups with exceptional behavior in time-to-event data. The area of time-to-event or survival data analysis has its basis in statistics, where the main goal is to predict <em>if</em> and <em>when</em> an event will happen. In other words, the main goal in survival analysis has long been to build global models able to predict the time for the occurrence of an event. Nevertheless, very often predictive models are used to compare stratified data in order to evaluate whether a variable is associated or not with the outcome. For instance, patients might be stratified according to a treatment variable (placebo or not) to compare models (survival curves) and decide on the effectiveness of the treatment. Although this is an effective approach if the variable of interest is already known, it does not provide an alternative for the cases where specialists do not know how to stratify the data, that is, if they do not know which variable could be related to the outcome. Our approach targets exactly this. Our method seeks combinations of variables that are associated, i.e. describe, subgroups of individuals with unexpected or exceptional survival curves. In this sense, we complement the literature with a descriptive approach that is able to find and characterize those groups for specialists. Our method is based on the framework of exceptional model mining. It improves on a preliminary version presented in a conference. The main enhancement was to redesign our heuristic to retrieve interesting and diverse subgroups while minimizing three aspects of redundancy: coverage; description; and model. Our second extension regards how the quality function is applied. We now allow users to control whether the quality measure compares subgroups against the population, or against individuals that do not satisfy the descriptive rule. Third, we conduct further experiments to compare the performance of our approach to state of the art algorithms with real world benchmark data sets. Finally, we also present a case study showing a possible application of our method in the bioinformatics/health domain.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"690 ","pages":"Article 121549"},"PeriodicalIF":8.1,"publicationDate":"2024-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142530822","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0