Pub Date : 2026-01-29DOI: 10.1016/j.ipm.2026.104641
Xiaohan Wang , Xueting Liu , Wenxin Tai , Joojo Walker , Yong Wang , Kai Chen , Fan Zhou
Accurate IP geolocation plays a critical role in a wide range of location-aware applications, from cybersecurity to content delivery. While recent advances in deep learning have led to substantial improvements in geolocation accuracy, existing methods often fail to generalize under out-of-distribution (OOD) scenarios caused by distribution shifts. To address this challenge, we propose a novel framework-Graph Invariant Learning (GIL)-for IP geolocation, referred to as GILGeo. Our approach is designed to identify invariant structural patterns in IP graphs across diverse environments, thereby enhancing model generalizability. By dynamically recombining invariant and spurious features, GILGeo simulates a variety of environmental conditions during training. This promotes the learning of domain-invariant representations and leads to significantly improved performance in unseen OOD settings. Extensive experiments on three real-world datasets show that GILGeo outperforms state-of-the-art baselines, establishing a new benchmark for IP geolocation under distributional shift. Our anonymized code and datasets are publicly available at: https://github.com/xiaohanwang01/GILGeo.
{"title":"Invariant learning improves out-of-distribution generalization for IP geolocation","authors":"Xiaohan Wang , Xueting Liu , Wenxin Tai , Joojo Walker , Yong Wang , Kai Chen , Fan Zhou","doi":"10.1016/j.ipm.2026.104641","DOIUrl":"10.1016/j.ipm.2026.104641","url":null,"abstract":"<div><div>Accurate IP geolocation plays a critical role in a wide range of location-aware applications, from cybersecurity to content delivery. While recent advances in deep learning have led to substantial improvements in geolocation accuracy, existing methods often fail to generalize under out-of-distribution (OOD) scenarios caused by distribution shifts. To address this challenge, we propose a novel framework-Graph Invariant Learning (GIL)-for IP geolocation, referred to as GILGeo. Our approach is designed to identify invariant structural patterns in IP graphs across diverse environments, thereby enhancing model generalizability. By dynamically recombining invariant and spurious features, GILGeo simulates a variety of environmental conditions during training. This promotes the learning of domain-invariant representations and leads to significantly improved performance in unseen OOD settings. Extensive experiments on three real-world datasets show that GILGeo outperforms state-of-the-art baselines, establishing a new benchmark for IP geolocation under distributional shift. Our anonymized code and datasets are publicly available at: <span><span>https://github.com/xiaohanwang01/GILGeo</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104641"},"PeriodicalIF":6.9,"publicationDate":"2026-01-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081580","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-28DOI: 10.1016/j.ipm.2026.104644
Xiaohui Zhou , Yijie Wang , Hongzuo Xu , Yizhou Li
Time series anomaly detection is crucial in many fields, where the objective is to identify unusual patterns by learning normality from sequential observations. However, existing methods typically treat the entire training data as a single, homogeneous normal class, which disregards the normal diversity caused by distribution shifts over time. As a result, these methods are forced to learn a single, complex decision boundary that must enclose all variations of normal behavior, making it difficult to precisely distinguish subtle anomalies hidden within the normal patterns. Therefore, this paper tackles this challenge by explicitly modeling heterogeneous normality, which allows for learning simpler, localized decision boundaries to separate anomalies. Specifically, we propose a novel approach that decomposes the heterogeneous class space into multiple normal classes, adopting a two-stage coarse-to-fine training paradigm: (1) a Mixture of Experts (MoE) framework assigns pseudo-labels by routing input features to specialized experts for prediction, approximating the latent sub-class structure; (2) enhanced features are generated based on pseudo-labels and feature space is refined via spectral decomposition, which contracts class boundaries and better exposes anomalies. Extensive experiments on 23 univariate datasets and 17 multivariate datasets show that our approach significantly outperforms state-of-the-art competitors by 2.55%-21.76% in VUS-PR, validating the importance of modeling heterogeneous normality in time series anomaly detection.
{"title":"Modeling heterogeneous normality in time series anomaly detection","authors":"Xiaohui Zhou , Yijie Wang , Hongzuo Xu , Yizhou Li","doi":"10.1016/j.ipm.2026.104644","DOIUrl":"10.1016/j.ipm.2026.104644","url":null,"abstract":"<div><div>Time series anomaly detection is crucial in many fields, where the objective is to identify unusual patterns by learning normality from sequential observations. However, existing methods typically treat the entire training data as a single, homogeneous normal class, which disregards the normal diversity caused by distribution shifts over time. As a result, these methods are forced to learn a single, complex decision boundary that must enclose all variations of normal behavior, making it difficult to precisely distinguish subtle anomalies hidden within the normal patterns. Therefore, this paper tackles this challenge by explicitly modeling heterogeneous normality, which allows for learning simpler, localized decision boundaries to separate anomalies. Specifically, we propose a novel approach that decomposes the heterogeneous class space into multiple normal classes, adopting a two-stage <em>coarse-to-fine</em> training paradigm: (1) a Mixture of Experts (MoE) framework assigns pseudo-labels by routing input features to specialized experts for prediction, approximating the latent sub-class structure; (2) enhanced features are generated based on pseudo-labels and feature space is refined via spectral decomposition, which contracts class boundaries and better exposes anomalies. Extensive experiments on 23 univariate datasets and 17 multivariate datasets show that our approach significantly outperforms state-of-the-art competitors by 2.55%-21.76% in VUS-PR, validating the importance of modeling heterogeneous normality in time series anomaly detection.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104644"},"PeriodicalIF":6.9,"publicationDate":"2026-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081583","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-28DOI: 10.1016/j.ipm.2026.104649
Xinnan Liu , Chenxi Liu , Anran Yu , Bin Tang , Zhenyang Cao , Zhengxiong Long , Runqi Su , Heng-Yang Lu
Nowdays, news events increasingly exhibit cross-domain topicality, making multi-domain fake news detection a critical yet challenging task. A key challenge is enhancing model performance for fake news detection while mitigating domain bias caused by unbalanced dataset. To address this, we propose the Knowledge-assistance Knowledge-mining Knowledge-debiased Multi-domain Fake News Detection Framework (K3MDFEND) which introduces external knowledge and proposes multi-domain contrastive learning. In particular, we integrate Large Language Models (LLMs) through a novel argumentation-based prompt engineering framework to obtain reliable external knowledge. We design Quality-Aware Attention Fusion module that dynamically weights evidence credibility to handle reviews of varying quality while combining with our deep learning framework. To further distill key insights from the comments while preserving their inherent semantic integrity, we leverage feature alignment techniques on the comment features. To further mitigate domain bias, we propose multi-domain contrastive learning and successfully combine the spurious correlations between domains and news authenticity. Extensive experiments on Chinese and English datasets demonstrate that K3MDFEND achieves state-of-the-art performance in both detection performance and bias metric reduction. On Chinese and English datasets, F1 scores, increase 92.89% ∼ 95.13% and 83.59% ∼ 85.28%, bias metric, decrease 0.8522 ∼ 0.5612 and 0.2698 ∼ 0.1931.
{"title":"Break fake frontiers: A triple-knowledge approach to multi-domain fake news detection","authors":"Xinnan Liu , Chenxi Liu , Anran Yu , Bin Tang , Zhenyang Cao , Zhengxiong Long , Runqi Su , Heng-Yang Lu","doi":"10.1016/j.ipm.2026.104649","DOIUrl":"10.1016/j.ipm.2026.104649","url":null,"abstract":"<div><div>Nowdays, news events increasingly exhibit cross-domain topicality, making multi-domain fake news detection a critical yet challenging task. A key challenge is enhancing model performance for fake news detection while mitigating domain bias caused by unbalanced dataset. To address this, we propose the Knowledge-assistance Knowledge-mining Knowledge-debiased Multi-domain Fake News Detection Framework (K<sup>3</sup>MDFEND) which introduces external knowledge and proposes multi-domain contrastive learning. In particular, we integrate Large Language Models (LLMs) through a novel argumentation-based prompt engineering framework to obtain reliable external knowledge. We design Quality-Aware Attention Fusion module that dynamically weights evidence credibility to handle reviews of varying quality while combining with our deep learning framework. To further distill key insights from the comments while preserving their inherent semantic integrity, we leverage feature alignment techniques on the comment features. To further mitigate domain bias, we propose multi-domain contrastive learning and successfully combine the spurious correlations between domains and news authenticity. Extensive experiments on Chinese and English datasets demonstrate that K<sup>3</sup>MDFEND achieves state-of-the-art performance in both detection performance and bias metric reduction. On Chinese and English datasets, F1 scores, increase 92.89% ∼ <strong>95.13%</strong> and 83.59% ∼ <strong>85.28%</strong>, bias metric, decrease 0.8522 ∼ <strong>0.5612</strong> and 0.2698 ∼ <strong>0.1931</strong>.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104649"},"PeriodicalIF":6.9,"publicationDate":"2026-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081579","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-28DOI: 10.1016/j.ipm.2026.104655
Yaqian Zhou , Zhenghao Fang , Zhibin Gu , Song Yang , Yan Wang
Unsupervised 3D model retrieval and classification have obtained much attention due to widespread applications. However, existing methods focus only on global representations while ignoring local saliency learning, leading to redundant distraction and insufficient complementarity. In addition, they neglect intra- and inter-class contextual relevance during representation learning, leading to inaccurate embedding space partitioning and missing representative prototypes. To address these challenges, we present Instance and Prototype Contrastive Learning (IPCL), an unsupervised dual-network framework that simultaneously captures view-level local features and model-level semantic information. Specifically, we treat each view as an instance and employ inter-instance contrastive learning to extract discriminative local salient features, mitigating redundancy and enhancing cross-view complementarity. For global semantic modeling, we establish class prototypes for 3D models and propagate the semantic information to global features via a prototype-aware contrastive loss, strengthening class-level discriminability. Innovatively, we employ a bottom-up adaptive clustering algorithm called voting clustering, which mines deeper semantic correlations to refine prototype selection and embedding space structure. Comprehensive evaluations demonstrate the superiority of IPCL, e.g., IPCL outperforms most unsupervised methods, achieving classification accuracy improvements by 0.1% to 18.1% on ModelNet40 and 0.1% to 16.7% on ShapeNet55. IPCL achieves average retrieval gains of 15.2% in NN on ModelNet40 and 16.7% in mAP under the micro setting on ShapeNet55.
{"title":"Instance and prototype contrastive learning for multi-view 3D model retrieval and classification","authors":"Yaqian Zhou , Zhenghao Fang , Zhibin Gu , Song Yang , Yan Wang","doi":"10.1016/j.ipm.2026.104655","DOIUrl":"10.1016/j.ipm.2026.104655","url":null,"abstract":"<div><div>Unsupervised 3D model retrieval and classification have obtained much attention due to widespread applications. However, existing methods focus only on global representations while ignoring local saliency learning, leading to redundant distraction and insufficient complementarity. In addition, they neglect intra- and inter-class contextual relevance during representation learning, leading to inaccurate embedding space partitioning and missing representative prototypes. To address these challenges, we present Instance and Prototype Contrastive Learning (IPCL), an unsupervised dual-network framework that simultaneously captures view-level local features and model-level semantic information. Specifically, we treat each view as an instance and employ inter-instance contrastive learning to extract discriminative local salient features, mitigating redundancy and enhancing cross-view complementarity. For global semantic modeling, we establish class prototypes for 3D models and propagate the semantic information to global features via a prototype-aware contrastive loss, strengthening class-level discriminability. Innovatively, we employ a bottom-up adaptive clustering algorithm called voting clustering, which mines deeper semantic correlations to refine prototype selection and embedding space structure. Comprehensive evaluations demonstrate the superiority of IPCL, e.g., IPCL outperforms most unsupervised methods, achieving classification accuracy improvements by 0.1% to 18.1% on ModelNet40 and 0.1% to 16.7% on ShapeNet55. IPCL achieves average retrieval gains of 15.2% in NN on ModelNet40 and 16.7% in mAP under the micro setting on ShapeNet55.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104655"},"PeriodicalIF":6.9,"publicationDate":"2026-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081582","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-27DOI: 10.1016/j.ipm.2026.104654
Yunji Park, Doowon Jeong
Current blockchain-based digital evidence systems provide strong technical integrity but fail to adequately address the procedural legitimacy required for court admissibility, frequently omitting judicial authorization workflows, differentiated handling of voluntary versus compulsory evidence, and transparent destruction protocols. To address these gaps, we propose B-DEMS, a blockchain-based digital evidence management system that integrates the full evidence lifecycle–from registration to court-authorized destruction–while encoding jurisdiction-specific legal requirements across South Korea, the United States, the European Union, and China. B-DEMS implements multi-party authorization, conditional decryption, and transaction-based disposal to ensure auditability and procedural compliance. Experimental evaluation across 1950 workflow executions demonstrated that B-DEMS achieved a maximum throughput of 10,890 TPS, representing 51–219% improvement over state-of-the-art systems, while maintaining stable scalability with latency increasing only 2.7-fold under a 5-fold peer expansion. Security analysis confirmed a 0% attack success rate across 300 adversarial attempts, and cross-border cooperation scenarios exhibited consistent adherence to jurisdiction-specific approval workflows. By aligning evidentiary procedures with a scalable blockchain architecture, B-DEMS provides a technically robust and procedurally compliant foundation for practical deployment in multi-agency and international investigative environments.
{"title":"A blockchain-based digital evidence management system: Integrating forensic procedures and multi-party authorization","authors":"Yunji Park, Doowon Jeong","doi":"10.1016/j.ipm.2026.104654","DOIUrl":"10.1016/j.ipm.2026.104654","url":null,"abstract":"<div><div>Current blockchain-based digital evidence systems provide strong technical integrity but fail to adequately address the procedural legitimacy required for court admissibility, frequently omitting judicial authorization workflows, differentiated handling of voluntary versus compulsory evidence, and transparent destruction protocols. To address these gaps, we propose B-DEMS, a blockchain-based digital evidence management system that integrates the full evidence lifecycle–from registration to court-authorized destruction–while encoding jurisdiction-specific legal requirements across South Korea, the United States, the European Union, and China. B-DEMS implements multi-party authorization, conditional decryption, and transaction-based disposal to ensure auditability and procedural compliance. Experimental evaluation across 1950 workflow executions demonstrated that B-DEMS achieved a maximum throughput of 10,890 TPS, representing 51–219% improvement over state-of-the-art systems, while maintaining stable scalability with latency increasing only 2.7-fold under a 5-fold peer expansion. Security analysis confirmed a 0% attack success rate across 300 adversarial attempts, and cross-border cooperation scenarios exhibited consistent adherence to jurisdiction-specific approval workflows. By aligning evidentiary procedures with a scalable blockchain architecture, B-DEMS provides a technically robust and procedurally compliant foundation for practical deployment in multi-agency and international investigative environments.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104654"},"PeriodicalIF":6.9,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-27DOI: 10.1016/j.ipm.2026.104653
Yue Fan , Hu Zhang , Ru Li , Guangjun Zhang , Yujie Wang , Hongye Tan , Yuanlong Wang , Xiaoli Li , Jiye Liang
Existing explainable question answering methods based on structured reasoning lack effective modeling of logical dependencies between steps and underutilize the potential of intermediate conclusions in structured reasoning. To address these challenges, we propose SRCR, a faithful Structured Reasoning method based on Curriculum Reinforcement learning. Specifically, we propose an easy-to-difficult reverse structured curriculum that gradually slides the initial state of reasoning from end to beginning, which fully captures the complex dependencies of multi-step reasoning. Moreover, we treat fact selection and deductive generation as a unified process and construct a faithfulness reward function to mine faithful reasoning steps during the model learning and exploring phases. Experimental results on the structured reasoning datasets EntailmentBank and STREET demonstrate that SRCR achieves state-of-the-art performance in factual accuracy and intermediate conclusion correctness, surpassing previous methods by 8.0% and 2.0%, respectively. Moreover, SRCR also improves answer accuracy by 2.6% to 8.3%, and extensive analysis shows that SRCR can generate more faithful structured explanations.
{"title":"SRCR: Faithful structured reasoning with curriculum reinforcement learning for explainable question answering","authors":"Yue Fan , Hu Zhang , Ru Li , Guangjun Zhang , Yujie Wang , Hongye Tan , Yuanlong Wang , Xiaoli Li , Jiye Liang","doi":"10.1016/j.ipm.2026.104653","DOIUrl":"10.1016/j.ipm.2026.104653","url":null,"abstract":"<div><div>Existing explainable question answering methods based on structured reasoning lack effective modeling of logical dependencies between steps and underutilize the potential of intermediate conclusions in structured reasoning. To address these challenges, we propose SRCR, a faithful <strong>S</strong>tructured <strong>R</strong>easoning method based on <strong>C</strong>urriculum <strong>R</strong>einforcement learning. Specifically, we propose an easy-to-difficult reverse structured curriculum that gradually slides the initial state of reasoning from end to beginning, which fully captures the complex dependencies of multi-step reasoning. Moreover, we treat fact selection and deductive generation as a unified process and construct a faithfulness reward function to mine faithful reasoning steps during the model learning and exploring phases. Experimental results on the structured reasoning datasets EntailmentBank and STREET demonstrate that SRCR achieves state-of-the-art performance in factual accuracy and intermediate conclusion correctness, surpassing previous methods by 8.0% and 2.0%, respectively. Moreover, SRCR also improves answer accuracy by 2.6% to 8.3%, and extensive analysis shows that SRCR can generate more faithful structured explanations.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104653"},"PeriodicalIF":6.9,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081585","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-27DOI: 10.1016/j.ipm.2026.104647
Min Zhu , Dengyin Zhang
The study presents the HERA-Net (Hierarchical Edge-aware Reinforcement Architecture) framework, which combines Hierarchical Motion Saliency (HMS) and Deep Reinforcement Learning (DRL) for adaptive crowd behavior recognition. The UCSD Ped2 dataset, comprising 32 surveillance clips (240 × 360 px), showed that HERA-Net improved generalization performance by 20 %, resilience to occlusion by 18 %, and recognition accuracy by 12–15 % compared to state-of-the-art models. In dynamic crowd situations, the HMS module hierarchically mixes local and global motion cues to maintain edge boundaries, while the DRL policy adaptively enhances recognition. A PPO-based DRL enables real-time adaptive behavior detection, and a unique edge-aware loss function ensures exact motion boundaries. Experimental results demonstrate that HERA-Net successfully balances precision and adaptability, making it a dependable, real-time system for intelligent surveillance, anomaly identification, and crowd monitoring.
{"title":"SPECTRA-Net: Spatiotemporal edge-preserving contextual reinforcement architecture for adaptive crowd behavior recognition","authors":"Min Zhu , Dengyin Zhang","doi":"10.1016/j.ipm.2026.104647","DOIUrl":"10.1016/j.ipm.2026.104647","url":null,"abstract":"<div><div>The study presents the HERA-Net (Hierarchical Edge-aware Reinforcement Architecture) framework, which combines Hierarchical Motion Saliency (HMS) and Deep Reinforcement Learning (DRL) for adaptive crowd behavior recognition. The UCSD Ped2 dataset, comprising 32 surveillance clips (240 × 360 px), showed that HERA-Net improved generalization performance by 20 %, resilience to occlusion by 18 %, and recognition accuracy by 12–15 % compared to state-of-the-art models. In dynamic crowd situations, the HMS module hierarchically mixes local and global motion cues to maintain edge boundaries, while the DRL policy adaptively enhances recognition. A PPO-based DRL enables real-time adaptive behavior detection, and a unique edge-aware loss function ensures exact motion boundaries. Experimental results demonstrate that HERA-Net successfully balances precision and adaptability, making it a dependable, real-time system for intelligent surveillance, anomaly identification, and crowd monitoring.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104647"},"PeriodicalIF":6.9,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146045205","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-27DOI: 10.1016/j.ipm.2026.104650
Qi Ouyang , Hongchang Chen , Shuxin Liu , Ran Li , Kai Wang , Yingle Li
Detecting group anomalies in spatiotemporal trajectories is vital for smart city security but remains challenging due to noise and complex dynamics. Existing diffusion models often suffer from “conditional contamination,” where observed anomalies inadvertently guide reconstruction, masking the very deviations they aim to detect. To address this, we propose Diffusion-Driven Group Anomaly Detection (DGAD). Specifically, we introduce an Interleaved Window Masking Strategy that segments data to enforce mutual supervision and reveal latent patterns. We pair this with an Unconditional Imputation Mechanism that conditions generation on forward noise instead of partial observations. This prevents anomaly leakage and significantly widens the divergence between normal and abnormal behaviors. Furthermore, a Denoising Weighted Voting module aggregates outputs across diffusion steps to mitigate uncertainty and enhance stability. Extensive experiments on synthetic and real-world datasets show that DGAD consistently outperforms state-of-the-art methods, improving F1-score by 1.1% on average and 1.8% in high-noise conditions. Code and case studies are available at [https://github.com/oyq-star/DAGD-main].
{"title":"Diffusion -driven group anomaly detection in spatiotemporal trajectories: Robust masked sequence imputation for enhanced pattern discovery","authors":"Qi Ouyang , Hongchang Chen , Shuxin Liu , Ran Li , Kai Wang , Yingle Li","doi":"10.1016/j.ipm.2026.104650","DOIUrl":"10.1016/j.ipm.2026.104650","url":null,"abstract":"<div><div>Detecting group anomalies in spatiotemporal trajectories is vital for smart city security but remains challenging due to noise and complex dynamics. Existing diffusion models often suffer from “conditional contamination,” where observed anomalies inadvertently guide reconstruction, masking the very deviations they aim to detect. To address this, we propose Diffusion-Driven Group Anomaly Detection (DGAD). Specifically, we introduce an Interleaved Window Masking Strategy that segments data to enforce mutual supervision and reveal latent patterns. We pair this with an Unconditional Imputation Mechanism that conditions generation on forward noise instead of partial observations. This prevents anomaly leakage and significantly widens the divergence between normal and abnormal behaviors. Furthermore, a Denoising Weighted Voting module aggregates outputs across diffusion steps to mitigate uncertainty and enhance stability. Extensive experiments on synthetic and real-world datasets show that DGAD consistently outperforms state-of-the-art methods, improving F1-score by 1.1% on average and 1.8% in high-noise conditions. Code and case studies are available at [<span><span>https://github.com/oyq-star/DAGD-main</span><svg><path></path></svg></span>].</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104650"},"PeriodicalIF":6.9,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081586","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-27DOI: 10.1016/j.ipm.2026.104651
Jia Yi , Fei Du , Yuchen Nie , Wencui Liang , Xiaoyu Zhou , Jinjuan Chen , Guangying Li , Mimi Liu , Yalan Lv , Wenlong Zhao , Xiaorong Hou
The rapid adoption of generative artificial intelligence (GAI) in healthcare has raised concerns about the quality of its outputs, yet existing health information assessment tools are not designed for AI-generated content. This study developed the GAI-HIQ assessment scale. Using Latent Dirichlet Allocation on 341 relevant publications, we initially identified three core dimensions and 15 secondary indicators for evaluating health information quality. A two-round Delphi consultation with 20 experts (100 % response rate) refined the framework to three core dimensions and 13 secondary indicators, achieving a statistically significant level of expert consensus (Kendall’s W = 0.288, P < 0.05). The analytic hierarchy process was then applied to calculate indicator weights, ensuring structured prioritization of quality dimensions (all judgment matrices yielded CR values below 0.1). The GAI-HIQ provides a consensus-based framework for evaluating health information generated by AI, offering practical tools for developers to optimize algorithms, for healthcare institutions to regulate applications, and for patients to assess information reliability.
生成式人工智能（GAI）在医疗保健领域的迅速采用引发了对其产出质量的担忧，但现有的健康信息评估工具并不是为人工智能生成的内容设计的。本研究编制了GAI-HIQ评估量表。通过对341份相关出版物的潜在狄利克雷分配，我们初步确定了评估健康信息质量的三个核心维度和15个次要指标。通过20位专家（100%回复率）的两轮德尔菲咨询，将框架细化为三个核心维度和13个次要指标，专家共识达到了具有统计学意义的水平（Kendall 's W = 0.288, P < 0.05）。然后应用层次分析法计算指标权重，确保质量维度的结构化优先级（所有判断矩阵的CR值均低于0.1）。AI- hiq为评估人工智能生成的健康信息提供了基于共识的框架，为开发人员优化算法、卫生保健机构规范应用程序和患者评估信息可靠性提供了实用工具。
{"title":"GAI-HIQ: Developing a health information quality assessment indicator system for generative artificial intelligence","authors":"Jia Yi , Fei Du , Yuchen Nie , Wencui Liang , Xiaoyu Zhou , Jinjuan Chen , Guangying Li , Mimi Liu , Yalan Lv , Wenlong Zhao , Xiaorong Hou","doi":"10.1016/j.ipm.2026.104651","DOIUrl":"10.1016/j.ipm.2026.104651","url":null,"abstract":"<div><div>The rapid adoption of generative artificial intelligence (GAI) in healthcare has raised concerns about the quality of its outputs, yet existing health information assessment tools are not designed for AI-generated content. This study developed the GAI-HIQ assessment scale. Using Latent Dirichlet Allocation on 341 relevant publications, we initially identified three core dimensions and 15 secondary indicators for evaluating health information quality. A two-round Delphi consultation with 20 experts (100 % response rate) refined the framework to three core dimensions and 13 secondary indicators, achieving a statistically significant level of expert consensus (Kendall’s <em>W</em> = 0.288, <em>P</em> < 0.05). The analytic hierarchy process was then applied to calculate indicator weights, ensuring structured prioritization of quality dimensions (all judgment matrices yielded CR values below 0.1). The GAI-HIQ provides a consensus-based framework for evaluating health information generated by AI, offering practical tools for developers to optimize algorithms, for healthcare institutions to regulate applications, and for patients to assess information reliability.</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104651"},"PeriodicalIF":6.9,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146081584","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-27DOI: 10.1016/j.ipm.2026.104645
Ziqi Li , Zhilin Chen , Tingting Guo , Yonghong Zhang , Xiaoning Song , Tianyang Xu
Nonnegative representation-based classification (NRC) enhances the discriminability of samples to some extent through nonnegativity constraints. However, this approach often fails to fully exploit local feature structures and class-discriminative information, which limits its overall classification performance. To overcome this limitation, we introduce the positive and negative neighbor dual-flexible nonnegative representation (PN2DFNR) classifier. This method integrates both the positive and negative neighbor sets of the query sample to effectively capture location information. Specifically, the positive neighbor set enforces spatial consistency, ensuring that the estimated representation of the query sample remains well-aligned with its neighboring samples. In contrast, the negative neighbor set introduces inverse constraints to suppress inter-class interference. Furthermore, a flexibility factor is incorporated to formulate a weighted flexible constraint strategy, which enhances the representational capacity of the correct class while adaptively attenuating the contribution of incorrect classes. To evaluate PN2DFNR’s performance, extensive experiments are conducted on facial recognition, handwritten digit classification, and large-scale datasets with diverse feature characteristics. The results demonstrate that PN2DFNR achieves superior classification performance, with the maximum improvement reaching approximately 2%. The model code will be available on the author’s homepage (https://github.com/li-zi-qi/PN2DFNR).
{"title":"Positive and negative neighbor dual-flexible nonnegative representation method for image classification","authors":"Ziqi Li , Zhilin Chen , Tingting Guo , Yonghong Zhang , Xiaoning Song , Tianyang Xu","doi":"10.1016/j.ipm.2026.104645","DOIUrl":"10.1016/j.ipm.2026.104645","url":null,"abstract":"<div><div>Nonnegative representation-based classification (NRC) enhances the discriminability of samples to some extent through nonnegativity constraints. However, this approach often fails to fully exploit local feature structures and class-discriminative information, which limits its overall classification performance. To overcome this limitation, we introduce the positive and negative neighbor dual-flexible nonnegative representation (PN<sup>2</sup>DFNR) classifier. This method integrates both the positive and negative neighbor sets of the query sample to effectively capture location information. Specifically, the positive neighbor set enforces spatial consistency, ensuring that the estimated representation of the query sample remains well-aligned with its neighboring samples. In contrast, the negative neighbor set introduces inverse constraints to suppress inter-class interference. Furthermore, a flexibility factor is incorporated to formulate a weighted flexible constraint strategy, which enhances the representational capacity of the correct class while adaptively attenuating the contribution of incorrect classes. To evaluate PN<sup>2</sup>DFNR’s performance, extensive experiments are conducted on facial recognition, handwritten digit classification, and large-scale datasets with diverse feature characteristics. The results demonstrate that PN<sup>2</sup>DFNR achieves superior classification performance, with the maximum improvement reaching approximately 2%. The model code will be available on the author’s homepage (<span><span>https://github.com/li-zi-qi/PN2DFNR</span><svg><path></path></svg></span>).</div></div>","PeriodicalId":50365,"journal":{"name":"Information Processing & Management","volume":"63 5","pages":"Article 104645"},"PeriodicalIF":6.9,"publicationDate":"2026-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146045206","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"管理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号