Search papers, labs, and topics across Lattice.
ASTER, a novel unsupervised time-series anomaly detection framework, generates pseudo-anomalies directly in the latent space using a tailored decoder, circumventing the need for handcrafted anomaly injections. A pre-trained LLM enhances temporal and contextual representations within this latent space, which is then used to train a Transformer-based anomaly classifier. ASTER achieves state-of-the-art performance on three benchmark datasets, demonstrating the effectiveness of latent-space pseudo-anomaly generation and LLM-enhanced representations for TSAD.
Forget hand-crafted anomaly injections: ASTER leverages a latent-space decoder and LLM-enhanced representations to generate tailored pseudo-anomalies, setting a new SOTA for unsupervised time-series anomaly detection.
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.
