arXiv:2605.26446v1 Announce Type: new 
Abstract: Graph anomaly detection (GAD) aims to identify nodes or substructures whose behavior or attributes deviate significantly from the overall pattern in graph-structured data, with critical applications in financial risk control, social network analysis, and cybersecurity. However, existing GCN-based methods suffer from the fundamental problem of contamination propagation, where anomalous nodes pollute the representations of their neighbors through message passing, leading to degraded detection performance. In this paper, we propose DDGAD, a novel diffusion-based graph anomaly detection framework that leverages trajectory dynamics to distinguish normal and anomalous nodes. Our key insight is that normal nodes exhibit consistent and stable representation trajectories under the coupled effects of diffusion regularization and reliability-aware neighborhood consensus, while anomalous nodes exhibit unstable and conflicting dynamics due to the directional disagreement between the global manifold prior and locally contaminated message passing. To mitigate contamination propagation, we introduce a distributed reliability-aware consensus refinement mechanism and define three complementary anomaly signals: neighbor inconsistency, reliability weight, and dynamical conflict energy. We further provide a preliminary theoretical analysis on normal node stability under the coupled dynamics. These signals collectively characterize anomalous behaviors from the perspectives of local inconsistency, consensus reliability, and dynamical instability. Extensive experiments on five real-world datasets demonstrate the effectiveness of the proposed framework.

تم تقديم DDGAD مؤخرًا، وهو إطار جديد قائم على الانتشار للكشف عن الشذوذ في الرسوم البيانية (GAD)، يهدف إلى تحسين تحديد العقد الشاذة في البيانات المهيكلة على شكل رسوم بيانية من خلال الاستفادة من ديناميات المسارات. تتناول هذه الطريقة مشكلة انتشار التلوث السائدة في الأساليب المعتمدة على GCN، والتي غالبًا ما تؤدي إلى تدهور أداء الكشف.

La reciente introducción de DDGAD, un nuevo marco basado en la difusión para la detección de anomalías en grafos (GAD), tiene como objetivo mejorar la identificación de nodos anómalos en datos estructurados en grafos aprovechando la dinámica de trayectorias. Este enfoque aborda el problema de la propagación de contaminación que prevalece en los métodos basados en GCN, lo que a menudo conduce a un rendimiento de detección degradado.

La récente introduction de DDGAD, un nouveau cadre basé sur la diffusion pour la détection d'anomalies dans les graphes (GAD), vise à améliorer l'identification des nœuds anormaux dans les données structurées en graphes en s'appuyant sur la dynamique des trajectoires. Cette approche traite le problème de la propagation de la contamination, courant dans les méthodes basées sur GCN, qui conduit souvent à une dégradation des performances de détection.

The recent introduction of DDGAD, a novel diffusion-based framework for Graph Anomaly Detection (GAD), aims to enhance the identification of anomalous nodes in graph-structured data by leveraging trajectory dynamics. This approach addresses the contamination propagation issue prevalent in existing GCN-based methods, which often leads to degraded detection performance.

DDGAD: Trajectory Dynamics for Diffusion-Based Graph Anomaly Detection

arXiv:2606.11272v1 Announce Type: new 
Abstract: Federated Learning (FL) enables collaborative and privacy-preserving model training across distributed clients, but most existing FL systems implicitly assume data stationarity. In real-world settings-such as healthcare, industrial IoT (IIOT), cybersecurity, and smart cities-data streams are inherently non-stationary, leading classical FL methods to suffer from performance degradation, instability, and catastrophic forgetting.
  Continual Learning (CL) addresses learning under evolving data distributions but has been largely studied in centralized settings, overlooking key constraints of federated systems, including privacy, limited communication, and client heterogeneity. Federated Continual Learning (FCL) emerges at the intersection of FL and CL, aiming to support lifelong, adaptive, and privacy-aware learning over distributed and non-stationary data.
  This survey provides a comprehensive and systematic overview of FCL. We first present a formal definition of the FCL problem and clarify its distinctive characteristics. We then analyze the limitations of classical FL under non-stationary conditions, highlighting how CL principles support long-term adaptation. To organize the rapidly growing literature, we propose a multi-dimensional taxonomy of FCL approaches. Furthermore, we review representative application domains and data modalities, summarize commonly used evaluation metrics, and discuss experimental perspectives for assessing long-term performance and forgetting. Finally, we highlight key open challenges, including handling extreme heterogeneity under temporal drift, designing scalable and privacy-preserving memory mechanisms, and establishing standardized benchmarks. This survey aims to serve as a reference and a roadmap for advancing FCL toward robust and deployable real-world systems.

تم نشر مسح شامل حول التعلم المستمر الفيدرالي (FCL) الذي يتناول تحديات التعلم مدى الحياة والحفاظ على الخصوصية عبر البيانات الموزعة وغير الثابتة. يسلط هذا المسح الضوء على قيود أساليب التعلم الفيدرالي التقليدية (FL) التي تفترض غالبًا ثبات البيانات، ويؤكد على الحاجة إلى تقنيات التعلم التكيفي في تطبيقات العالم الحقيقي المختلفة مثل الرعاية الصحية والأمن السيبراني.

Se ha publicado una encuesta completa sobre el Aprendizaje Continu Federado (FCL), que aborda los desafíos del aprendizaje a lo largo de la vida y la preservación de la privacidad sobre datos distribuidos y no estacionarios. Esta encuesta destaca las limitaciones de los métodos tradicionales de Aprendizaje Federado (FL), que a menudo asumen la estacionariedad de los datos, y enfatiza la necesidad de técnicas de aprendizaje adaptativo en diversas aplicaciones del mundo real, como la atención médica y la ciberseguridad.

Une enquête complète sur l'apprentissage continu fédéré (FCL) a été publiée, abordant les défis de l'apprentissage tout au long de la vie et préservant la vie privée sur des données distribuées et non stationnaires. Cette enquête met en lumière les limitations des méthodes d'apprentissage fédéré (FL) traditionnelles, qui supposent souvent la stationnarité des données, et souligne la nécessité de techniques d'apprentissage adaptatif dans diverses applications du monde réel telles que la santé et la cybersécurité.

A comprehensive survey on Federated Continual Learning (FCL) has been published, addressing the challenges of lifelong and privacy-preserving learning over distributed and non-stationary data. This survey highlights the limitations of traditional Federated Learning (FL) methods, which often assume data stationarity, and emphasizes the need for adaptive learning techniques in various real-world applications such as healthcare and cybersecurity.

DDGAD: Trajectory Dynamics for Diffusion-Based Graph Anomaly Detection

Was this article worth reading? Share it

Ready to build your own newsroom?