ASTER: Latent Pseudo-Anomaly Generation for Unsupervised Time-Series Anomaly Detection

2026-04-15

AI Summary

This work addresses the challenges of time series anomaly detection—namely, the rarity and heterogeneity of anomalies and the scarcity of labeled data—by proposing an unsupervised approach that automatically generates pseudo-anomalies in a latent space without requiring manual injection or domain knowledge. A latent-space decoder synthesizes tailored pseudo-anomalies to train a Transformer-based classifier, while a pre-trained large language model (LLM) is integrated to enhance temporal and contextual representations. By eliminating reliance on fixed distance metrics and handcrafted anomaly synthesis, the method achieves state-of-the-art performance across three benchmark datasets. Notably, it establishes a new paradigm by being the first to effectively incorporate LLMs into unsupervised time series anomaly detection.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Natural Language Processing: (Large) Language Models

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Time-series anomaly detection (TSAD) is critical in domains such as industrial monitoring, healthcare, and cybersecurity, but it remains challenging due to rare and heterogeneous anomalies and the scarcity of labelled data. This scarcity makes unsupervised approaches predominant, yet existing methods often rely on reconstruction or forecasting, which struggle with complex data, or on embedding-based approaches that require domain-specific anomaly synthesis and fixed distance metrics. We propose ASTER, a framework that generates pseudo-anomalies directly in the latent space, avoiding handcrafted anomaly injections and the need for domain expertise. A latent-space decoder produces tailored pseudo-anomalies to train a Transformer-based anomaly classifier, while a pre-trained LLM enriches the temporal and contextual representations of this space. Experiments on three benchmark datasets show that ASTER achieves state-of-the-art performance and sets a new standard for LLM-based TSAD.

Problem

• time-series anomaly detection
• unsupervised learning
• pseudo-anomaly generation
• latent space
• label scarcity

Innovation

• latent-space pseudo-anomaly generation
• unsupervised time-series anomaly detection
• Transformer-based classifier
• pre-trained LLM for time series
• domain-agnostic anomaly synthesis