Overview

Online anomaly detection studies how to flag unusual behavior in continuously arriving time series, logs, graphs, and sensor data, often before failures or attacks cause harm. Common approaches include prediction-residual and statistical change detectors, autoencoders and other learned representations, ensembles, and adaptive thresholds; newer systems also use drift detection, continual learning, or test-time adaptation. Work spans settings such as industrial telemetry, cloud logs, networks, and edge devices, where latency, memory, explanations, and operator feedback can matter as much as detection quality. A central tension is that adapting to changing data may help with drift but can add cost or false alarms, and recent benchmarks find that static methods can still outperform streaming ones, underscoring the need for chronological, realistic evaluations.

Latest papers

CardsList

65 papers5 papers in the last four weeks, down 37% on the four weeks before. 0.0% of all new papers.

  1. Self-Evolving Multi-Agent Network for Industrial IoT Predictive Maintenance

    Feb 17, 2026Rebin Saleh, Khanh Pham Dinh, Balázs Villányi +1Predictive MaintenanceIndustrial Anomaly Detection

  2. COMET: Codebook-based Online-adaptive Multi-scale Embedding for Time-series Anomaly Detection

    Feb 2, 2026Jinwoo Park, Hyeongwon Kang, Seung Hun Han +1Unsupervised Anomaly DetectionTime-Series Anomaly Detection

  3. A Novel GPT-Based Framework for Anomaly Detection in System Logs

    Oct 16, 2025Zeng Zhang, Wenjie Yin, Xiaoqi LiFocal LossAnomaly Detection

  4. Risk-Calibrated Bayesian Streaming Intrusion Detection with SRE-Aligned Decisions

    Sep 17, 2025Michel A. YoussefCost-Sensitive LearningProbability Calibration

  5. Joint Detection of Fraud and Concept Drift inOnline Conversations with LLM-Assisted Judgment

    May 7, 2025Ali Şenol, Garima Agrawal, Huan LiuOnline Anomaly DetectionConcept Drift