Evaluating anomaly detection methods in multivariate time series is challenging due to limited benchmark datasets with fine-grained annotations. Fun-TSG is introduced as a customizable time series generator to address this, enabling both automated and manual data generation with full transparency for rigorous evaluation.
This paper proposes a method for anomaly detection called Chimera Training, focusing on violations of semantic constraints given as logical rules over learned visual concepts. It employs a neural rule evaluator that compiles constraints into directed acyclic graphs, learning logical operators to calculate rule-satisfaction probabilities, even with scarce training data for actual violations.
This content discusses the application of unsupervised learning methods for anomaly detection within the Bitcoin network. It aims to identify unusual patterns or fraudulent transactions by leveraging AI techniques without prior labeled data.
The text discusses the need for explainable and causal AI in space operations, illustrating with a satellite incident where an automated correction violated data sovereignty regulations. It highlights the failure of traditional AI approaches to handle the complexity of technical constraints, operational priorities, and jurisdictional boundaries.
The author developed Torrix, a self-hosted LLM observability platform that logs calls, calculates costs, and flags anomalies automatically, to solve issues like surprise bills and silent model failures. A live demo with 30 days of simulated LLM traces across three projects was created to facilitate experimentation.
This content explores the detection of abnormal client behavior in federated learning environments. It aims to identify and mitigate unusual actions that could compromise the integrity of the global model.
ModSense is an AI-assisted moderation intelligence system, a production-grade prototype designed for large communities like Reddit. It combines real-time anomaly detection and graph-based community health modeling with an agentic AI layer (Gemini 3 Flash) to identify and respond to evolving issues like toxicity, brigading, and misinformation.
PhaseNet++ introduces a novel frequency-domain autoencoder for anomaly detection in Industrial Control Systems (ICS), addressing the overlooked phase spectrum in multivariate time series analysis. It utilizes a Phase Coherence Index to guide a graph attention network for enhanced detection of cyber-physical attacks.
This paper introduces a high-fidelity, physics-informed co-simulation of an aircraft main-fuel-pump system to generate data for anomaly detection and diagnosis algorithms. It addresses the inherent lack of data in critical cyber-physical systems and demonstrates feasibility using unsupervised RNN-VAE and SOM-VAE models.
This research presents a system for detecting suspicious performance patterns in athletics, using 1.6 million performances and eight methods including machine learning and trajectory analysis. It aims to complement traditional anti-doping programs by identifying potential violations through data analysis, with trajectory-based methods proving most effective.
This paper introduces a new non-parametric method for conditional anomaly detection using soft harmonic functions. It aims to identify unusual responses in clinical data, such as omitted lab tests, demonstrating its efficacy on real-world electronic health records.
SMT-AD é uma nova abordagem inspirada em computação quântica para detecção de anomalias, utilizando redes de tensores e embedding de características assistido por Fourier. O método se mostrou eficaz em datasets padrão, como transações de cartão de crédito, alcançando performance competitiva mesmo com configurações mínimas.
Industry 4.0 requires robust anomaly detection in Cyber-Physical Systems (CPS), but product-agnostic models have blind spots for subtle anomalies or targeted attacks. This work demonstrates this vulnerability and proposes a Product-Aware Autoencoder as a mitigation by restricting the learning domain.
A hierarchical ensemble pipeline is proposed for anomaly detection in multivariate telemetry data from the European Space Agency (ESA). This method, integrating various feature extraction and modeling techniques, shows strong generalization and effectiveness in detecting subtle anomalies in satellite telemetry.
This paper introduces Logic-GNN, a neuro-symbolic framework that leverages Temporal Graph Neural Networks and Graph Kolmogorov Complexity to detect data entry errors in clinical records. It identifies anomalies as "grammatical violations" in a latent logical grammar of medical interactions, achieving an F1-score of 0.94 on a large clinical dataset.
This paper proposes MODIAD, a novel framework for Multimodal Online Distributed Industrial Anomaly Detection, addressing limitations of existing methods in real-world industrial environments. It aims to leverage edge intelligence for distributed model training in industrial systems.
This paper introduces a new framework for detecting anomalies in maritime Automatic Identification System (AIS) datasets, focusing on abnormal vessel behaviors. It proposes a novel quality metric, MADQI, to evaluate the performance of unsupervised learning models in detecting these anomalies without requiring labeled data.
Within-dataset class-split evaluation for anomaly detection can be ill-posed when anomaly classes overlap with normal data, causing score instability or inversion. A new diagnostic, neighborhood class leakage, is introduced to predict this instability across various datasets and models.
Artificial Intelligence is revolutionizing cybersecurity, serving as both the ultimate weapon and the strongest shield in the digital world. AI-powered systems monitor vast networks and detect anomalies that surpass human capacity, operating ceaselessly.