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

TL;DR

ASTER introduces a novel latent space pseudo-anomaly generation framework for unsupervised time-series anomaly detection, leveraging large language models and Transformer classifiers to improve detection accuracy.

Contribution

It presents a new method that generates pseudo-anomalies in latent space, eliminating the need for domain-specific anomaly synthesis and handcrafted injections.

Findings

Achieves state-of-the-art performance on three benchmark datasets.

Sets a new standard for LLM-based time-series anomaly detection.

Outperforms existing reconstruction and embedding-based methods.

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.